FEDERAL COURT OF AUSTRALIA

Sanda v PTTEP Australasia (Ashmore Cartier) Pty Ltd (No 7) [2021] FCA 237

File number:

NSD 1245 of 2016

Judge:

YATES J

Date of judgment:

19 March 2021

Catchwords:

REPRESENTATIVE PROCEEDINGS – common law negligence – oil spill from an oil well within the offshore area of the Territory of Ashmore and Cartier Islands – where the respondent was the holder of a petroleum production licence for an area covering the Montara oil field where the well was located – where the respondent had the responsibility to exercise control over the suspension and operation of the well – whether oil from the spill reached the coastal areas of the Regencies of Kupang and Rote in Indonesia – whether oil from the spill caused or materially contributed to the death and loss of seaweed crops in those areas – where the applicant and Group Members are seaweed farmers – whether the respondent owed the applicant and Group Members a duty of care – whether the respondent breached its duty of care – whether the applicant has established that he suffered loss and damage – assessment of the applicant’s damages

Legislation:

Environment Protection and Biodiversity Conservation Act 1999 (Cth)

Federal Court of Australia Act 1976 (Cth)

Offshore Petroleum Act 2006 (Cth)

Offshore Petroleum and Greenhouse Gas Storage Act 2006 (Cth) s 569

Petroleum (Submerged Lands) (Management of the Environment) Regulations 1999 (Cth) reg 14(8)

Limitation Act 1981 (NT) s 44

Cases cited:

Bryan v Maloney (1995) 182 CLR 609

Caltex Refineries (QLD) Pty Ltd v Stavar [2009] NSWCA 258; 75 NSWLR 649

Chapman v Hearse (1961) 106 CLR 112

Fink v Fink (1946) 74 CLR 127

Generic Health Pty Ltd v Bayer Pharma Aktiengesellschaft [2018] FCAFC 183; 267 FCR 428

Malec v JC Hutton Pty Ltd (1990) 169 CLR 638

Mineralogy Pty Ltd v Sino Iron Pty Ltd (No 16) [2017] WASC 340

Perre v Apand Pty Ltd [1999] HCA 36; 198 CLR 180

Place (Granny Smith) Pty Ltd v Thiess Contractors Pty Ltd [2003] HCA 10; 196 ALR 257

Sanda v PTTEP Australasia (Ashmore Cartier) Pty Ltd (No 3) [2017] FCA 1272

Sanda v PTTEP Australasia (Ashmore Cartier) Pty Ltd (No 6) [2019] FCA 1853

Shirt v Wyong Shire Council [1978] 1 NSWLR 631

The Commonwealth of Australia v Amann Aviation Pty Limited (1991) 174 CLR 64

The Council of Wyong Shire v Shirt (1980) 146 CLR 40

Date of hearing:

17 to 19, 24 to 28 June 2019

1, 8, 9, 11 July 2019

21 to 25, 28 to 30 October 2019

2, 3, 5, 6, 11 to 13 December 2019

Registry:

New South Wales

Division:

General Division

National Practice Area:

Other Federal Jurisdiction

Category:

Catchwords

Number of paragraphs:

1171

Counsel for the Applicant:

Mr J Sexton SC with Ms V Bosnjak and Ms Z Hillman

Solicitor for the Applicant:

Maurice Blackburn

Counsel for the Respondent:

Mr C Scerri QC with Mr J Arnott and Mr A Barraclough

Solicitor for the Respondent:

Allens

Table of Corrections

27 October 2021

In the last sentence of paragraph 865, the word “modelling” has been inserted after “Dr Hubbert’s”.

ORDERS

NSD 1245 of 2016

BETWEEN:

DANIEL ARISTABULUS SANDA

Applicant

AND:

PTTEP AUSTRALASIA (ASHMORE CARTIER) PTY LTD (ACN 004 210 164)

Respondent

JUDGE:

YATES J

DATE OF ORDER:

19 MARCH 2021

THE COURT ORDERS THAT:

1.    The proceeding be listed for the purpose of receiving further submissions on Common Questions 3 and 4 referred to in the reasons for judgment published today, and on the question of interest up to judgment in relation to the damages to be awarded to the applicant, if that question is in dispute.

Note:    Entry of orders is dealt with in Rule 39.32 of the Federal Court Rules 2011.

REASONS FOR JUDGMENT

Introduction

[1]

How the Montara oil spill occurred

[8]

The OSCP

[38]

Chemical Composition: An overview of The chemical composition and physical properties of Fresh and weathered Montara oil

[56]

The Rote/Kupang region of Indonesia

[78]

Aspects of the seaweed industry in Indonesia

[85]

the conduct of the hearing

[103]

The lay evidence

[104]

The expert evidence

[106]

The applicant’s expert evidence

[108]

The respondent’s expert evidence

[132]

The lay evidence

[144]

The applicant’s evidence

[144]

The applicant’s education and life before seaweed farming

[145]

Seaweed farming

[149]

The applicant’s business practices

[158]

The oil

[160]

After the oil

[164]

The effect of the oil in 2009 on the applicant’s life and income

[167]

Cross-examination

[168]

The evidence of other seaweed farmers, village heads and other lay observers

[171]

The volume of oil spilled

[251]

Introduction

[251]

Oil volume flow rate calculation

[255]

Professor Wereley’s analysis and calculations

[255]

Criticism of Professor Wereley’s analysis and calculations

[263]

Professor Wereley’s response

[276]

The Joint Report on Volume

[288]

Material balance calculation

[304]

Dispersants

[322]

Satellite imagery

[364]

The oil spill Trajectory Modelling Evidence

[423]

Introduction

[423]

Dr Hubbert’s modelling

[438]

Dr French-McCay’s modelling

[450]

Criticism of Dr Hubbert’s modelling

[472]

Dr Hubbert’s response

[491]

Criticism of Dr French-McCay’s modelling

[505]

The Modelling of Ocean Currents

[509]

Introduction

[509]

Ocean circulation in the Timor Sea region: the ITF

[513]

Professor Ivey’s modelling

[534]

Professor Ivey’s criticism of GCOM3D

[536]

Dr Hubbert’s response

[541]

Dr Hubbert’s criticism of SUNTANS

[542]

Dr Sprintall’s criticism of SUNTANS

[548]

The Joint Report on Currents

[549]

Methodology

[552]

Data

[584]

Model predictions

[590]

Tracking buoy data

[602]

Lagrangian Coherent Structures

[616]

GIS Mapping

[618]

Findings and conclusions on Trajectory Modelling

[622]

The weathering of Montara oil

[678]

Introduction

[678]

The Leeder Reports

[684]

Dr Stout’s analyses

[692]

Criticisms of Dr Stout’s long-term weathering study

[715]

Dr Stout’s response

[722]

Conclusions on Dr Stout’s long-term weathering study

[728]

The LEMIGAS analyses

[736]

Conclusion

[748]

The Sibert sample

[751]

Expert commentary on the lay observations of oil

[763]

Professor Ball’s evidence

[763]

Dr Fingas’ evidence

[770]

Dr Taylor’s evidence

[775]

Dr Maki’s evidence

[798]

Professor Ball’s response

[803]

Dr Fingas’ response

[811]

Dr Thorhaug’s evidence

[813]

Did Montara oil reach the Rote/Kupang region?

[814]

The applicant’s submissions

[814]

The respondent’s submissions

[822]

Conclusion

[828]

Toxicology evidence

[869]

Introduction

[869]

Adhesion

[874]

PAHs in the wax-enriched fraction

[892]

The Microtox test

[897]

The Toxic Units Model

[907]

Brown kelp as a surrogate for red seaweed

[936]

The relevance of other oil spill studies and the persistence of spilled oil

[950]

The opinions of the experts: summary

[966]

Other explanations for seaweed death and damage

[976]

Natural oil seeps

[977]

Ship traffic

[981]

Coral spawning

[984]

Ice-ice, sea temperatures, and climate change

[990]

Ocean acidification

[1003]

Did Montara oil cause or materially contribute to the loss of seaweed in the Rote/Kupang region?

[1008]

Did a duty of care exist?

[1020]

Was the duty of care breached?

[1044]

were the applicant’s Seaweed crops damaged by Montara oil?

[1051]

Damages

[1052]

Introduction

[1052]

The calculation of actual production and profit

[1060]

2008

[1062]

2009

[1081]

2010

[1089]

2011

[1093]

2012

[1100]

2013

[1121]

2014

[1129]

Estimate

[1136]

Projected production and profit

[1139]

The calculation of loss

[1159]

Answers to the common questions

[1163]

Conclusion and disposition

[1171]

SCHEDULE A

SCHEDULE B

SCHEDULE C

YATES J:

Introduction

1    This proceeding is a representative proceeding brought under Pt IVA of the Federal Court of Australia Act 1976 (Cth). It concerns alleged damage to seaweed farming activities in Indonesia. This damage is said to have occurred from an oil spill at the Montara oil field operated by the respondent, PTTEP Australasia (Ashmore Cartier) Pty Ltd.

2    In early 2009, the respondent set about suspending an oil well, referred to as the H1 Well, in the oil field. There were certain failures in this process which led, in August 2009, to a well blowout and the uncontrolled spill of hydrocarbons from the well, which remained unabated for more than 10 weeks.

3    The applicant’s case is that the respondent owed him and the Group Members a duty of care in respect of the suspension and operation of the H1 Well, and that it breached that duty. He says that oil from the blowout reached certain areas in Indonesia, including the southern coastal area of Rote, an island where he lives and carries on his occupation as a seaweed farmer. He alleges that the oil killed, and caused a drop in the production of, his seaweed crop and the seaweed crops of the Group Members.

4    The cause of action on which the applicant relies is common law negligence. He claims damages. He commenced this proceeding after the expiration of the applicable limitation period. On 15 November 2017, the Court made an order pursuant to s 44 of the Limitation Act (NT) 1981 extending the limitation period in respect of his claim: Sanda v PTTEP Australasia (Ashmore Cartier) Pty Ltd (No 3) [2017] FCA 1272. At the present time, the limitation period has not been extended in respect of any Group Member.

5    The respondent denies liability. It accepts that it was negligent in suspending and operating the H1Well, but it contends that it did not owe the alleged duty of care to the applicant or the Group Members. Further, it contends that even if a duty of care was owed and breached, the evidence before the Court does not establish that any oil spilled from the H1 Well reached the areas in Indonesia, which the applicant specified in Schedule 1 to the further amended statement of claim as areas that were reached by the oil. It also contends that, even if any of the spilled oil reached any of those areas, it would not have been in a concentration or form that would have been toxic to the seaweed crops in place at that time. Finally, it contends that the applicant’s claim of loss is not supported.

6    The applicant originally pleaded and advanced a case that dispersants applied to the oil at the time of the spill also reached Indonesian waters and killed, and caused a loss in the production of, the seaweed crops. In the course of oral closing submissions, the applicant made clear that he no longer advances that case.

7    For the reasons that follow, I am satisfied that the respondent owed a duty of care to the applicant and the Group Members, and that it breached that duty. I am satisfied that oil spilled from the H1 Well blowout reached certain areas of Indonesia (which areas are in a region conveniently described as the Rote/Kupang region), including the area where the applicant grows his seaweed crop. I am satisfied that this oil caused or materially contributed to the death and loss of his crop. I am satisfied that, although difficult to assess, and although attended with uncertainty, the applicant’s loss can be calculated, and that he is entitled to an award of damages.

How the Montara oil spill occurred

8    The Montara oil field is located within the offshore area of the Territory of Ashmore and Cartier Islands, approximately 250 km northwest of the Western Australian coast and approximately 700 km from Darwin, within Australian territorial waters in the Timor Sea. It is about 100 km from Cartier Island and 150 km from the Ashmore Reef, within an area characterised by significant oil and gas reserves known as the Bonaparte Basin.

9    In September 2003, the respondent (which at the time was known by the name Coogee Resources (Ashmore Cartier) Pty Ltd) acquired the retention lease for the Montara oil field. Between September 2003 and August 2009, it developed the field for oil production. As part of this process, it engaged Atlas Drilling (S) Pte Ltd (Atlas) in early 2009 to drill four production wells (referred to in these reasons as the H1 Well, the H2 Well, the H3 Well and the H4 Well), as well as a gas injection well. The H1 Well is the oil well with which this proceeding is concerned.

10    The procedure for drilling the H1 Well was as follows. A drilling rig (here, the West Atlas rig operated by Atlas) was moved to the position at which the well was to be constructed. A drill from the rig was used to bore a hole into the sea bed, to access the hydrocarbon reservoir from which oil was to be produced. A steel pipe casing (being lengths of steel pipe joined together, usually by screws, and often referred to as the casing string) of a slightly smaller diameter than the resulting hole was inserted into the hole. In the H1 Well, the first casing string was 13 3/8” in diameter (the 13 3/8” casing string). Cement was pumped into the lowermost joints of the 13 3/8” casing string to form a casing shoe. The cement occupied the joints, and the bottom part of the area between the hole that had been bored and the casing string (the annulus). A narrower hole was drilled through the cement in the casing shoe and further into the sea bed, and a second casing string was inserted into the hole to create a new casing string of narrower diameter. In the H1 Well, this second casing string was 9 5/8” in diameter (the 9 5/8” casing string).

11    As at 18 January 2009, the respondent intended to suspend the H1 Well. The suspension of an oil well involves a process of capping (that is, effectively “plugging”) the well to prevent the release of hydrocarbons, pending later completion of the work required for actual production of oil through the well. The respondent intended to suspend the well by using cement in the 9 5/8” casing shoe as the primary control barrier, and a shallow set cement plug from 160 m to 115 m as the secondary control barrier.

12    However, at some point between January and March 2009, the respondent determined to use a pressure-containing anti-corrosion cap (PCCC) on each of the casing strings as the secondary control barrier rather than the concrete plug. This decision was made notwithstanding the fact that the manufacturer of the PCCCs, which the respondent proposed to use, did not intend that PCCCs be used as a barrier against the uncontrolled release of hydrocarbons and did not design the PCCCs for that purpose; there was no practicably available test that could verify the internal pressure-containing capability of a PCCC; and, unlike other forms of secondary barriers (including concrete plugs), PCCCs were required to be removed prior to a casing string being tied back to a wellhead platform. “Tying back” a casing string involves adding more casing string to extend the well back up to the mezzanine deck on the wellhead platform. The fact that the PCCCs were required to be removed meant that no secondary barriers would be in place during the tying back process.

13    On 6 March 2009, the respondent applied to the Director of Energy, Department of Regional Development, Primary Industry, Fisheries and Resources of the Northern Territory (Director of Energy), who holds the responsibilities of the Designated Authority under the Offshore Petroleum Act 2006 (Cth) and the Petroleum (Submerged Lands) (Management of Well Operations) Regulations 2004 (Cth) in respect of the area within which the Montara oil field is located, for approval to suspend the H1 Well, on the basis that the planned suspension would occur in two stages. The first was to involve the cementing and pressure testing of the 9 5/8” casing string, followed by the installation of a PCCC on that casing string. The second was to involve the installation of a second PCCC on the 13 3/8” casing string. The Director of Energy gave the respondent preliminary approval for suspension of the H1 Well in response to this suspension application.

14    On 12 March 2009, the respondent made a further application to the Director of Energy for approval to suspend the H1 Well. Also on that day, the respondent issued a formal change control order to Atlas, which specified that the shallow set cement plug which had been proposed to be used as a well control barrier in the process of suspending the H1 Well was to be replaced by PCCCs on each of the casing strings.

15    On 13 March 2009, the Director of Energy granted the respondent approval to suspend the H1 Well consistently with the applications it had lodged on 6 and 12 March 2009.

16    Between 2 and 7 March 2009, the H1 Well was drilled to a depth of approximately 3,796 m, with a total vertical depth of approximately 2,654 m.

17    At this time, the foot of the 9 5/8” casing string was in the reservoir for the well, at a point that was 3 m above the point where oil and water came into contact. The 9 5/8” casing string shoe was in a horizontal position. The effect of this arrangement was that the casing string provided a potential pathway for hydrocarbons to enter the H1 Well.

18    On 7 March 2009, the respondent installed a float collar. This comprised two float valves, which were to act as one way valves to allow cement to be pumped beneath the float collar without the cement returning up the casing string, to create the cement shoe that was intended to be the primary barrier controlling the release of hydrocarbons from the H1 Well. The float collar made provision for two plugs (a bottom plug and top plug) which were intended to lock, following the pumping of cement into the 9 5/8” casing string shoe, to create a seal within that casing string. The respondent then pumped cement into the 9 5/8” casing string shoe. The cement travelled through the end of the 9 5/8” casing string and up into the annulus of that casing string. Some of the cement remained in the casing string to fill the space between the float valves. This cement formed the cement shoe. Following the pumping of the cement, approximately 9.25 barrels (bbl) of displacement fluid (consisting of inhibited seawater) were pumped into the 9 5/8” casing string for the purpose of pressure testing. The pressure in the casing string was held at 4,000 psi for approximately 10 minutes.

19    It is convenient at this point to note that when a casing string shoe is cemented, two forms of cement are usually used in concert: lead cement, which is pumped into the casing string first, followed by tail cement, which has a higher density and thickening time than the lead cement.

20    In the case of the H1 Well, the respondent’s Well Construction Standards provided that, in cementing the 9 5/8” casing string shoe, tail cement be placed within the annulus outside the casing string to a height of 100 m above the top of the hydrocarbon reservoir. However, in this case the respondent determined to place tail cement within the annulus to a height of only 69 m above the top of the hydrocarbon reservoir. To achieve this, the required volume of tail cement was 199 bbl. In addition, when cementing the shoe, the respondent incorrectly pumped only 132 bbl of tail cement, causing the cement to reach a height of only 61 m below the top of the hydrocarbon reservoir. As a result of this failure, hydrocarbons in the reservoir for the H1 Well were permitted to leach into the annulus outside the 9 5/8” casing string and compromised the integrity of the cement shoe.

21    At around 2.40 pm on 7 March 2009, the pressure in the 9 5/8” casing string was released and 16.5 bbl of fluid were returned up the casing string, comprising the 9.25 bbl of displacement fluid which had been pumped into the casing string and approximately 7.25 bbl of fluid consisting of a combination of cement and leached hydrocarbons. This return of fluid indicated that both the float valves in the 9 5/8” casing string shoe and the plugs in that shoe had failed.

22    At around 2.45 pm on 7 March 2009, the 16.5 bbl of fluid which had been returned from the 9 5/8” casing string were pumped back into that casing string. The casing string was then closed while the cement set. The effect of pumping the returned fluid back into the 9 5/8” casing string was that approximately 9.25 bbl of inhibited seawater and approximately 7.25 bbl of cement and leached hydrocarbons were forced beneath the float collar within the 9 5/8” casing string, thereby displacing cement from the 9 5/8” casing string shoe. This caused a situation known as “wet shoe”, meaning that the areas within the casing string shoe that should have consisted of cement were partly cement and partly other material, including inhibited seawater and leached hydrocarbons. The displaced cement was forced into the annulus of the 9 5/8” casing string. The top and bottom plugs in the 9 5/8” casing string shoe did not lock. The cement shoe was then subjected to pressure at 1,350 psi while the cement set.

23    Later on 7 March 2009, the respondent was provided with a report that set out the events that had occurred during the course of the attempt to install the cement shoe. Further reports detailing the process of the cement shoe installation were prepared by the Day Drilling Supervisor and provided to the respondent. No further testing or assessment of the cement shoe was undertaken by the respondent or any other person on its behalf.

24    It is clear that the respondent was informed of the process by which the cement shoe had been installed on 7 March 2009. The respondent knew, or ought to have known, that the cement shoe lacked integrity and could not be relied upon to control the release of hydrocarbons from the H1 Well. Despite this, from the period March 2009 to August 2009, the respondent relied on the cement shoe as an effective primary control barrier against the release of hydrocarbons from the H1 Well.

25    In addition to the cement shoe, the respondent’s application to suspend the H1 Well was approved, as I have said, on the basis that it put in place a secondary control barrier, being the installation of one PCCC on the 9 5/8” casing string and one PCCC on the 13 3/8” casing string.

26    Sometime in March 2009, presumably after 12 March 2009, the respondent determined not to install a PCCC on the 13 3/8” casing string. Following the installation of the cement shoe on the H1 Well as described above, the respondent removed the upper section of the 9 5/8” casing string and installed a PCCC on that casing string. That PCCC was not tested or verified in situ. The respondent also removed the upper section of the 13 3/8” casing string, but did not install a PCCC on the remaining casing string. Nevertheless, during the period March 2009 until August 2009, the respondent relied on the PCCC installed on the 9 5/8” casing string as an effective secondary control barrier against the release of hydrocarbons from the H1 Well.

27    The “overbalancing” of fluid in a casing string, in which the hydrostatic pressure of the fluid in the casing string is greater than the pressure of the hydrocarbon reservoir (with an appropriate safety margin), may be used as a control barrier against the uncontrolled release of hydrocarbons.

28    During the period from March to August 2009, the fluid used in the 9 5/8” casing string consisted of seawater, the normal pressure of which is 1.02 – 1.03 sg. The pore pressure within the hydrocarbon reservoir for the H1 Well was 1.04 sg. As a result, the H1 Well was not overbalanced and was not capable of providing a pressure-based barrier to the release of hydrocarbons from the reservoir. Further, neither the respondent nor any person on its behalf had tested or monitored the pressure of the fluid inside the 9 5/8” casing string, and the fluid inside the casing string had not been verified as being in overbalance. Nevertheless, the respondent mistakenly relied on the fluid inside the 9 5/8” casing string as an effective barrier against the release of hydrocarbons from the reservoir.

29    In sum, in suspending the H1 Well in March 2009, the respondent relied upon three control barriers to prevent the uncontrolled release of hydrocarbons from the reservoir under the well: the cement shoe; the PCCCs; and the fluid inside the 9 5/8” casing string. None of these control barriers had been tested. Each of them was deficient. One had not even been installed (the PCCC which was to have been installed on the 13 3/8” casing string).

30    On 21 April 2009, the West Atlas rig left the Montara oil field.

31    Around 7 July 2009, the respondent applied to the Director of Energy for approval of its drilling program in respect of the Montara oil field. Among other things, the application included a diagram which indicated that PCCCs had been installed on both the 9 5/8” casing string and the 13 3/8” casing string. The application was approved on 13 July 2009.

32    On 19 August 2009, the West Atlas rig returned to the Montara oil field to allow the respondent to tie back the casing strings for each of the five wells (the H1 Well, the H2 Well, the H3 Well, the H4 Well and the gas injection well), so as to complete the wells to the point of production.

33    At around 4.30 am on 20 August 2009, the West Atlas rig moved over the H1 Well. Upon examination by the respondent, it was discovered that the PCCC for the 13 3/8” casing string had not been installed, and as a result the inner threads of the uppermost portion of that casing string had rusted or corroded. In order to tie the corroded casing string back to the Montara wellhead platform, it was necessary for the threads on that casing string to be cleaned, which necessitated the removal of the PCCC on the 9 5/8” casing string. The removal took place at around 11.30 am on 20 August 2009. It was determined by the respondent that the PCCC should not be reinstalled. The PCCC was correspondingly not immediately re-installed, as it should have been. At this point, the only remaining control barrier against the release of hydrocarbons from the H1 Well reservoir was the cement shoe.

34    At around 5.00 pm on 20 August 2009, the West Atlas rig left the H1 Well.

35    At approximately 5.30 am on 21 August 2009, the cement shoe at the H1 Well failed and there was a release of hydrocarbons from the H1 Well, the volume of which the respondent estimated to be between 40 and 60 bbl. At around 7.23 am on 21 August 2009 there was a further, larger release of hydrocarbons from the H1 Well.

36    In response to the two releases of hydrocarbons from the H1 Well (together, the Montara oil spill), the respondent and Atlas evacuated 69 personnel from the West Atlas rig and the Montara wellhead platform.

37    The uncontrolled release of hydrocarbons from the H1 Well flowed for a period in excess of 10 weeks from August 2009 until around 3 November 2009. The volume of oil released into the environment from the wellhead is a contested question about which a large body of evidence was adduced. I will return to the question of volume later in these reasons.

The OSCP

38    The development of the Montara oil field required approval under the Environment Protection and Biodiversity Conservation Act 1999 (Cth) (the EPBC Act). This approval was given on 3 September 2003. It was a condition of the approval that the respondent submit an oil spill contingency plan detailing the strategy that the respondent had in place to mitigate the environmental effects of any hydrocarbon spills.

39    On 5 June 2009, the Assistant Secretary of the Environmental Assessment Branch of the Department of the Environment and Water Resources approved an oil spill contingency plan submitted by the respondent on 19 May 2009 (the OSCP). The OSCP was a revision (Version 5, dated 1 April 2009) of earlier plans that the respondent had prepared.

40    At the time, the Australian regulatory framework did not prescribe the contents of oil spill contingency plans. The Petroleum (Submerged Lands) (Management of the Environment) Regulations 1999 (Cth) simply required the maintenance of an up-to-date emergency response manual that included an oil spill contingency plan: reg 14(8).

41    The respondent adduced evidence through Dr Elliott Taylor, an expert in (amongst other things) oil spill contingency planning and response, that the OSCP was, as at August 2009, reasonable, functional and comprehensive, and both met and exceeded planning requirements specified in Australia at the time. Dr Taylor also said that the OSCP was aligned with generally accepted oil field practices, with best international practices for offshore oil spill contingency plans in place at the time, and with Australia’s National Plan to Combat Pollution of the Sea by Oil and Other Noxious Substances (the National Plan). The National Plan has operated since 1973 and is managed by the Australian Maritime Safety Authority (AMSA). It is the national integrated government and industry consultative framework regarding marine pollution preparedness and the response to the threat posed to the marine environment by oil and chemical spills.

42    The respondent relied on the OSCP to support its case that it did not owe a duty of care to the applicant and Group Members. I will discuss that case in a later section of these reasons. For present purposes, I draw attention to the fact that the OSCP provided oil trajectory information. This information included oil trajectory modelling.

43    A fundamental aspect of oil spill contingency planning is the assessment of the risks posed by various uncontrolled spill scenarios. Oil spill response planners use a hazard assessment to identify potential spill sources and the volumes of oil related to each source for a particular operation. A worst-case spill scenario is typically used to assess the potential area of influence of a major spill through oil spill trajectory modelling. In practice, this modelling typically assumes little or no intervention to contain, collect or treat the spilled oil, other than eventually stopping the spill at its source. Put another way, the modelling assumes that the spilled oil is subjected to natural environmental processes only. Dr Taylor’s evidence was that no oil spill contingency plan is expected to identify all potential spill scenarios or outcomes. Rather, the plan is intended to ensure that mechanisms for a scalable response are in place.

44    The OSCP was prepared in the context of the National Plan, which classifies oil spills according to a three-tiered system. As described by Dr Taylor, Tier 1 is for spills of less than 10 m3. Typically, this might be a spill in the course of ship transfer or bunkering at a jetty or mooring. Tier 2 is for spills of 10 to 1000 m3. Typically, this might involve shipping incidents in ports, pipeline failures or nearshore exploration or production. Tier 3 is for spills greater than 1000 m3. This is regarded as a major incident, typically involving tankers or vessels with large bunker oil volumes. Such incidents might also include, for example, collisions or vessel loss, or well blowout.

45    The National Plan itself is not directed to specific spill sources or volumes for tiered response planning purposes. In short, any spill over 1000 m3 would be considered a Tier 3 incident. Indeed, in respect of designed spill size, the National Plan provides (Section 1, para 1.6):

1.6    The National Plan is established to respond to oil spills of any size in Australian waters. For planning and operational reasons and based on the experience of spills in Australia and international criteria, a designed spill size of 21,000 tonnes [approximately 24,000m3] exists. This has been determined by National Plan stakeholders taking into account current ship type and equipment holdings and is endorsed by the Australian Transport Council … as the appropriate level for which to plan equipment and other resource requirements. Additionally, arrangement are in place to augment this capacity from overseas equipment stockpiles should any incident exceed Australia’s resource capability.

46    A report dated 4 April 2003, which was prepared by URS Australia Pty Ltd to provide preliminary information in relation to the drilling of the H1, H2 and H3 Wells, as required under the EPBC Act (the URS Report), states (at 6.4.2.4):

6.4.2.4 Well control

With current technology, the risk of a well blowout is considered low. There are elaborate monitoring systems to detect potential blowouts and such events can occur only if all of the monitoring systems fail and if the casing, wellhead or blow-out preventers (BOPs) fail catastrophically. The occurrence of such circumstances has been greatly reduced by improved back-up systems. The risk is further reduced when knowledge of the underlying stratigraphy and formation pressures is available as a result of previous drilling nearby. Such knowledge is available to Newfield through the drilling results of previous wells in the vicinity of the Licence Area and this knowledge has been used in designing the drilling programme.

No shallow gas has been encountered in previous drilling.

Loss of well control could potentially result in substantial release of hydrocarbons to the environment. However, modern techniques have reduced the possibility of a blow-out to a minimal level and a blow-out has never occurred in all of the wells drilled off the Western Australian coast. A blow-out can occur only in the extremely unlikely event that all systems fail and warning signs are ignored. The probability of a blow-out is minimised by:

    testing the BOP before starting the operation and regularly during the operation;

    pressure testing of casing strings;

    continuous monitoring for abnormal pressure during drilling; and

    providing mandatory training for the drill crew in safety procedures.

Should a blow-out occur, the volume spilled will depend on the permeability of the producing formation, the thickness of the encountered producing interval, the viscosity of the oil, the number and type of obstructions in the well hole, and the time taken to regain control and seal off the well bore. Drilling of directional ‘interception’ or relief wells to stop the flow can be undertaken, but this is considered the last resort as this operation can take several weeks to complete.

Data collected by the WA MPR on offshore exploratory and production drilling in Western Australia show that no significant oil spills have been associated with a total of over 400 offshore wells drilled to date. No major oil spill from offshore drilling operations has been known to occur in Australia.

In almost 30 years of operation, the oil and gas industry in Australia has drilled over 1,500 exploration and development wells and produced over 3,500 million barrels (556,500 ML) of oil. During this same period, the total amount of oil spilled to the marine environment from all offshore oil exploration and production activities has been estimated to be less than 1000 barrels (159,000 L), with the majority of these spills occurring during production activities (Volkman et al. 1994).

Six blow-outs have occurred in Australia, of which three occured during exploration drilling. All six were gas blow-outs and none resulted in an oil spill. There have been no blow-outs in Australia since 1984, which is evidence of the technological and procedural improvements that have occurred over the last two decades.

These statistics led the Independent Scientific Review of the Environmental Effects of the Australian Oil Industry (Swan et al. 1994) to conclude: “there is minimal oil spill threat caused by Australian explorers”.

47    Dr Taylor relied on the URS Report to understand the sources of information used in developing the OSCP. He accepted the results and information presented in the report as being “professionally complete and correct”.

48    As is clear from the above quote, the URS Report proceeded on the basis that the risk of a well blowout would be “low”. Indeed, in a later part of the report, URS concluded that such an event would be “rare”. On the other hand, URS concluded that spills from the transfer of produced crude oil from a floating production, storage and offloading (FPSO) vessel would be the main source of spills in oil production operations.

49    Proceeding on this basis, the OSCP posited the maximum realistic spill event (i.e., the worst-case scenario) to be the total loss of crude oil from one wing tank of the Montara Venture FPSO, representing 15,000 m3 of Montara crude oil spilled over a period of 12 hours. The OSCP included trajectory modelling which investigated such a spill over seven days. In his evidence, Dr Taylor pointed out that this assumed spill volume was much larger than the spill records from blowouts registered in Australia over the preceding decades.

50    The results of the modelling illustrated the probability that spills may be transported to different locations around the well. At para 2.3.4, the OSCP stated:

The results of the surface slick modelling indicated that spills of oil from Montara are unlikely to impact on the nearest shorelines (Hibernia Reef, Ashmore Reef and Cartier Islands). The shorelines of Australia, Timor and the Indonesian Islands were all predicted to be at no risk whatsoever.

During winter the overall tendency for oil spills is to move in a south-westerly direction driven by a combination of prevailing winds and the tides. The ebb and flood of the tide through this area is in a north-south direction whilst during winter the dominant prevailing winds vary from northeast to east. The combination of these two forces, together with the tendency for surface currents to bend to the left of the wind direction as a result of coriolis forcing, produces this result.

During summer the tendency is in the opposite direction, to the north-northeast. Again this result is due to the direction of the ebb tide (approximately north) and the prevailing southwest and westerly winds. The steering of surface currents to the left of the wind is also a factor. During this period (and possibly the transition months) the wind and current forcing resulted in a predominant movement of oil slicks to the north, towards the chain of seamounts to the north of Montara. Investigation of the behaviour of oil components entrained or in solution however showed that there is negligible risk of sub-surface oil impacting on these seamounts, which are at least 10 m below the surface and the closest some 30 km away.

51    Although not professing to have personal experience with the model used, Dr Taylor said that the model’s approach, and the data sets for wind and currents, and the oil properties and weathering characteristics, used in the model, were well-defined and consistent with best practice in 2009. Later, after referring to AMSA’s technical guidelines for preparing contingency plans for marine and coastal facilities (published in January 2015), Dr Taylor said that the modelling was “consistent with best practices today” for oil trajectory, weathering and mass balance projections. He said that the OSCP’s prediction that the shorelines of Australia, Timor and the Indonesian Island were “at no risk whatsoever” from oil impact, was “consistent with best practice in planning at the time of the Montara oil spill”. Dr Taylor then expressed the conclusion that:

67    … a reasonable oil field operator would not have expected or foreseen an oil spill incident with the potential to harm residents of [Nusa Tenggara Timur] given the characteristics of the oil in the production field and analysis of oil weathering and trajectories forecast for the assumed reasonable worst-case spill incident at the time.

52    I point out, for later reference, that the modelling on which the OSCP was based was carried out by Global Environment Modelling Systems Pty Ltd (GEMS) using GCOM3D, a three-dimensional hydrodynamic model which was used to model the ocean currents, and the GEMS spill model called OILTRAK3D. The modelling was undertaken by Dr Graeme Hubbert, who was called by the applicant to give evidence on trajectory modelling and ocean currents.

53    It is convenient at this stage to also refer to modelling carried out by the respondent in 2011 and revisited in 2013. It looked at a 77 day period (a “loss of well control” spill) of 84,966 m3 (534,380 bbl) with a variable flow rate peaking at 3,802 m3 per day (23,912 bbl/day) down to 690 m3/day (4,340 bbl/day). The modelling was completed for three distinct seasons, defined by the unique prevailing wind and general current conditions. The modelling predicted a 90% probability of oil making shoreline contact >10 g/m2 with Rote, for all seasons. The report of this modelling described this scenario as “credible”.

54    The applicant relied on this modelling to support his case that the respondent owed him and the Group Members a duty of care. On the question of foreseeability, he submitted that the modelling showed information that was available to the respondent in 2009, had it taken steps to access that information at that time. The applicant submitted that the modelling undertaken for the OSCP in 2009 simply looked at the outcome of the loss of oil from a vessel wing tank. However, this was a risk which could only have arisen at some time in the future, when the H1 Well was in production. According to the applicant, the real risk, at the relevant time, was of a well blowout, given the “egregious incompetence” with which the respondent purported to temporarily seal the well.

55    According to the applicant, the OSCP in 2009 simply “failed to grapple” with the risks attached to the work the respondent was in fact undertaking. In cross-examination, Dr Taylor accepted that his opinion that a reasonable oil field operator would not have expected or foreseen an oil spill incident with potential harm to residents of NTT was based on the history of operations in the area, not on the particular facts leading to the blowout of the H1 Well.

Chemical Composition: An overview of The chemical composition and physical properties of Fresh and weathered Montara oil

56    No-one knows the chemical composition of fresh (meaning, not weathered) crude oil taken from the H1 Well (Montara-1 oil). No samples of Montara-1 oil were studied or were available to be studied prior to the spill. After the spill, the H1 Well was plugged and abandoned, making it impossible to obtain any sample. Similarly, no-one knows the physical properties of Montara-1 oil. However, two other oils from the Montara field were available for study—fresh oil from the H2 Well (Montara-2 oil) and fresh oil from the H3 Well (Montara-3 oil). Montara-3 oil was collected in 2002 and analysed by Intertek and Leeder Consulting. Montara-2 oil was collected in 2017, for the purposes of this case, and analysed by Dr Scott Stout, who was called by the respondent to give expert evidence. The experts on this topic—Dr Stout, and Professor Ball and Dr Fingas (who were called by the applicant)agreed that Montara-2 oil and Montara-3 oil can be taken as suitable surrogates for Montara-1 oil.

57    There is no dispute about the chemical composition of Montara-2 oil or, in relevant respects, its physical properties, which are summarised in the following tables:

Summary of the Chemical Composition of the fresh Montara-H2 oil

Chemical Composition

Value

Units

Bulk Composition

Saturates

76

wt%

Aromatics

21

wt%

Resins (NSO)

2.4

wt%

Asphaltenes

0.30

wt%

Detailed Composition

Total SHC1

250,000

μg/g

Total PIANO2

133,000

μg/g

Total PAHs3

43,600

μg/g

Total Petroleum Hydrocarbons (TPH)

839,000

μg/g

1Saturated Hydrocarbons (n-alkanes and targeted isoprenoids, C9-C40)

2Volatiles (paraffins, isoparaffins, aromatics, naphthenes, and olefins)

3Polycyclic aromatic hydrocarbons; total of 50 PAH analytes

Summary of the Physical Properties of the fresh Montara-H2 oil

Physical Property

Value

Units

Specific Gravity (15.6ºC)

0.8502

unitless

API Gravity (15.6ºC)

34.9

º

Density (15.6ºC)

0.8494

g/mL

Wax Content

13.7

wt%

Interfacial Tension

70.58

mN/m

Pour Point

24

ºC

58    The following table provides a comparison between the two surrogates—Montara-2 oil and Montara-3 oil. Although differences exist between the values for the properties listed in the table, the relevant experts agreed that, overall, these oils are generally comparable to each other. Further, based on the apparent continuity, structure and character of the Montara field’s oil reservoir, the relevant experts agreed that there is no geologic basis to expect significant differences between the crude oil produced from different wells in the Montara field:

Comparison of Chemical and Physical Properties of Surrogate Montara crude oils

Description

Montara-H2

(2017)

(Dr Stout)

Montara-3

(2002)

(Leeder Consulting/

Intertek)

Units

Pour Point

24

27

ºC

Wax Content

13.7

11.3

wt%

API Gravity

34.9

34.8

o

Density@ 15.6 oC

0.8494

*ND

g/ml

Density@ 15 oC

*ND

0.8506

g/ml

Specific Gravity @ 15.6 oC

0.8502

0.8510

unitless

Total BTEX

28,200

32,900

μg g-1

Total PAHs

43,600

*ND

μg g-1

Total Saturates

76.1

58.1

wt%

Total Aromatics

21.3

26.9

wt%

Total Resins

2.4

*ND

wt%

Total Asphaltenes

0.30

0.98

wt%

*ND—not determined

59    In light of the above discussion, it is convenient to refer to Montara-1 oil, Montara-2 oil and Montara-3 oil as, simply, Montara oil unless it is necessary to distinguish between the three oils.

60    The chemical composition and physical properties of crude oil can be affected by weathering. The processes involved include evaporation, aerosolization, dissolution, biodegradation, photochemical oxidation (also called photo-oxidation), and wax-agglomeration and separation.

61    Evaporation is the volatilization of oil components into the atmosphere. Aerosolization is a specific type of evaporation caused by injection of the oil into the air prior to it reaching the sea surface.

62    Biodegradation is the breakdown of oil components by microorganisms in the environment. Photochemical oxidation or photo-oxidation is the breakdown of oil components due to chemical reactions caused by exposure to sunlight.

63    Wax agglomeration and separation is the precipitation of waxy components in the oil to form wax-rich aggregates and their subsequent separation from the liquid oil. This is an atypical, but not unprecedented, weathering process. It affects “high wax” oils, such as Montara oil.

64    Emulsification is another weathering process in which oil and water become mixed to form emulsions. The experts agreed that this process, albeit common, was unlikely to have affected the spilled Montara-1 oil because of its low asphaltene and resin content. The experts agreed that reports at the time of the spill of “emulsified slicks”, and “emulsions” or “possible emulsions”, were not true emulsions but were, more likely, wax-enriched oils formed by the wax-agglomeration process.

65    These weathering processes occur mostly concurrently and would have had a collective (not individual) effect on the spilled Montara-1 oil.

66    The experts on this topic were asked to consider, in conclave, the effect of the weathering processes on the visual appearance, wax content, pour point, viscosity, smell, toxicity and adhesiveness of this oil. Their observations and conclusions were based, in part, on field-collected samples of weathered Montara-1 oil analysed by Leeder Consulting at the time of the spill.

67    As to visual appearance, the experts agreed (based on photographs and sample descriptions given at the time of the spill) that, as the oil weathered, it generally became lighter in colour (brown to orange to yellow to khaki) and formed white waxy particles. As waxy aggregates formed and became increasingly abundant, the oil may have appeared to be more viscous, which is a possible explanation for the field descriptions of the floating oil, made at the time of the spill, as “emulsified slicks” or “emulsions”.

68    The experts agreed that the overall wax content of the spilled oil increased through a combination of the conventional weathering processes and the wax agglomeration and separation process referred to above. Analysis of 13 field-collected samples taken at the time of the spill showed that the wax content of the weathered oil ranged from 13% to 79%.

69    The experts agreed that the pour point of the spilled oil (the lowest temperature that oil will flow when it is cooled) increased through a combination of the conventional weathering processes and the wax agglomeration and separation process. Analysis of the 13 field-collected samples showed that the pour point of the weathered oil ranged from 30°C to 51°C. This is an increase in the pour point of Montara-2 oil and Montara-3 oil. The experts agreed that the higher pour points of the field-collected samples indicate that, at night-time temperatures, most of the weathered spilled oil would have been solid and that highly-weathered oil and wax-rich aggregates with elevated pour points would have remained as solids at daytime temperatures.

70    There was some disagreement between the experts as to whether the data on viscosity obtained from 11 field-collected samples taken at the time of the spill were reliable. It is not necessary to engage with that debate because, despite that uncertainty, the experts agreed that it was their expectation that the viscosity of the spilled oil would have increased through a combination of the conventional weathering processes and the wax agglomeration and separation process.

71    The experts were sceptical that the intensity or nature of the spilled oil’s smell could be reliably described as having changed due to weathering. Certainly, there was no data available to them to evaluate this qualitative property, which they considered to be highly subjective to the individual describing the smell and, therefore, an unreliable means to assess the weathering of oil.

72    The evidence does not disclose that there was any investigation undertaken of the toxicity of the Montara oil at the time of the spill. However, the experts agreed that the concentrations of compounds that are typically associated with aquatic toxicity—the monoaromatic hydrocarbons benzene, toluene, ethylbenzene and xylene (BTEX), polycyclic aromatic hydrocarbons (PAHs) and total aromatic hydrocarbons—were measured in multiple field-collected samples at the time of the spill. None of the samples contained detectable BTEX. All samples showed reduced concentrations of PAHs and total aromatic hydrocarbons with increasing % weight (mass) loss (a proxy for weathering). They concluded that it was likely that the toxicity of the spilled oil decreased through a combination of the conventional weathering processes and the wax agglomeration and separation process. Notwithstanding this agreement, there was substantial debate about the significance of this decreased toxicity, particularly in relation to seaweed grown in the Rote/Kupang region of Indonesia in 2009 and subsequent years. I will deal with that topic in later sections of these reasons.

73    The relevant experts disagreed on whether the adhesiveness of the spilled Montara oil (here, its ability to adhere to biological material) would increase with weathering. Professor Ball and Dr Fingas contended that adhesiveness would increase with weathering. Dr Stout contended that there was no reliable or relevant data that addressed this topic. Once again, I will deal with that topic, but only to the extent that it is necessary to do so, in a later section of these reasons.

74    Based on qualitative observations in respect of 64 field-collected samples at the time of the spill, the experts agreed that the spilled Montara oil experienced varying degrees of weathering or wax-enrichment. Evaporation was clearly the most important weathering process that initially affected the oil after its release. Water-washing, biodegradation and photo-oxidation further caused a progressive loss of non-volatile aromatics (PAHs). Weathering and concurrent wax agglomeration and separation formed increasingly wax-rich residues that contained long-chain n-alkanes, but little else. Biodegradation did affect floating oils, but probably only in sheens (not slicks).

75    The relevant experts also agreed that quantitative observations of field-collected samples showed that BTEX was rapidly and completely lost from the spilled oil that was sampled. The % weight (mass) loss (once again, a proxy for weathering) showed losses ranging from 4% to 92%, with the highest loss being to the wax-rich residues (88% to 92%). Total aromatic hydrocarbons (>C7 to C35) and total PAHs were substantially reduced in the floating oils, such that the wax-rich residues contained 1.6% of total aromatic hydrocarbons and no detectable total PAHs (i.e., >50 mg/kg-1 or 50 ppm). The total aromatic hydrocarbons that persisted in the most highly-weathered wax-rich residue that was studied were exclusively comprised of larger aromatic hydrocarbons in the C16 to C35 (mostly C21 to C35) carbon range.

76    The significance of these observations will have greater meaning when I deal in more detail with the respective cases that were advanced on the topic of the toxicity of oil in relation to seaweed. I simply note, for present purposes, that BTEX and the PAHs are the chemicals commonly associated with aquatic toxicity.

77    It is convenient to record at this juncture that a number of observations made at the time of the spill—including by seaweed farmers and other observers in the Rote/Kupang region in late 2009—concerned the presence of foam. The relevant experts agreed that observations of foam in the sea does not indicate the presence of oil or an oil dispersant, but does not preclude their presence. Dr Stout pointed out that four foam samples collected from the Ashmore Reef area during the spill, which were analysed by Leeder Consulting, contained either predominantly or exclusively chemicals derived from naturally-occurring biological material(s). Two samples contained some hydrocarbons that indicated the presence of small but varying amounts of highly-weathered oil or wax.

The Rote/Kupang region of Indonesia

78    Nusa Tenggara Timur (NTT) is one of 34 provinces of Indonesia. It is located in the Coral Triangle region of South East Asia, north of Australia. It comprises 21 regencies (or districts) and the regency-level city of Kupang. Two of the regencies, known as the Regency of Kupang and the Regency of Rote Ndao, are the focus of this proceeding. For convenience, I will refer to them as comprising the Rote/Kupang region.

79    The Regency of Kupang is located in the western-most region of West Timor on Timor Island. It includes an island just off the coast of West Timor called Semau.

80    The Regency of Rote Ndao comprises a main island (Rote) located to the south-west of Kupang, and a number of adjacent, smaller islands.

81    The Rote/Kupang region is located approximately 500 km north-west of the Australian coast, and approximately 240 km north-west of the Montara oil field. Schedule A to these reasons reproduces part of a hydrographic chart which includes this region. Rote and West Timor are located between (approximately) latitude 11˚0’0”S and 9˚0’0”S and longitude 122˚0’0”E and 125˚0’0”E. The Montara oil field is located between (approximately) latitude 12˚0’0”S and 13˚0’0”S and longitude 124˚0’0”E and 125˚0’0”E. The coast of Western Australia is visible in the south-east corner of the chart.

82    As in other areas of Indonesia, the inhabitants of the Rote/Kupang region are subject to several levels of government. The national Indonesian government is based some distance away in Jakarta, and administers the various provincial governments, including that of NTT. Each regency in NTT, known in Bahasa Indonesia as a “kabupaten”, is headed by an elected regent known as a “bupati”. Each regency contains a number of sub-districts, known as “kecamatan”. These, in turn, contain a number of villages, known as “desa”. Villages can also be divided into sub-villages, known as “dusun”.

83    Rote and its two main adjacent islands Rote Ndao and Rote Nuse comprise about 60 villages. The Regency of Kupang comprises about 21 villages. The island of Semau comprises about 14 villages, and Kupang Barat, on mainland Timor, comprises about seven villages.

84    As I have noted, in his further amended statement of claim the applicant claims that oil spilled from the H1 Well reached 81 villages located in the Rote/Kupang region. A map plotting the location of the villages, which were identified by the lay witnesses, who gave oral evidence, as being the location of their places of residence, is reproduced in Schedule B to these reasons. In the course of the hearing, this map was given the identifier SAN.941.001.0191.

Aspects of the seaweed industry in Indonesia

85    Seaweeds, also known as macroalgae, are multi-cellular photosynthetic organisms. They range from microscopic in size to tens of metres in length. While they are not technically plants, they perform the same ecological role in coastal marine systems as plants do in terrestrial systems. They are classified into four major taxonomic groups characterised by their typical colours, which are red, brown, green and blue-green algae. This proceeding concerns, principally, several species of red algae.

86    The metabolic processes of a seaweed are conducted through the surface of its entire body (thallus). Gas exchanges at the thallus enable seaweeds to generate energy through photosynthesis and conduct cellular respiration and metabolism. Seaweeds also absorb essential nutrients through the thallus. Reproduction in red algae also typically occurs by way of the thallus, which at certain phases during the life of the seaweed will produce microscopic gametes and spores. Once formed, the spores in particular are capable of growing into new seaweeds without the need for fertilisation.

87    Three types of seaweed are cultivated in the Rote/Kupang region, each of which are species of red algae. Specifically, the three species, which are collectively referred to as the eucheumatoid seaweeds, are Kappaphycus alvarezii, commercially referred to as cottonii; Kappaphycus striatum, commercially referred to as sakol; and Eucheuma denticulatum, commercially referred to as spinosum or espinosum. Cottonii and sakol are the two species which are predominantly grown in the region and represent almost all of the seaweed produced there, with very little spinosum grown by comparison. Even though classified as red algae (or red seaweed), cottonii and sakol can, in fact, exhibit various colours.

88    Natural stocks of both Kappaphycus and Eucheuma seaweeds occur throughout the Indo-Pacific region, between approximately 20° north and south of the equator. Kappaphycus tends to grow in the wild as solitary plants scattered widely through sea grass beds. For this reason, they were difficult to harvest for mass production until commercial farming of vegetative cultivars was developed.

89    Commercial farming of eucheumatoid seaweeds is mostly undertaken between 10° north and south of the equator, which contains the coastal areas of winter sea-temperature isoclines between 21°C and 24°C. These are the optimal temperatures for growth. The primary centres for commercial production are located in the Philippines and Indonesia, which fall within this geographic area.

90    Commercial tropical farming of cottonii and sakol commenced in the Philippines in 1974. Farming of espinosum commenced around the same time, but production volumes were only around 20% of the production volumes of the two Kappaphycus seaweeds. The Philippines enjoyed a monopoly on production until 1986, when Indonesia commenced commercial farming.

91    By 2006, Indonesia was the world’s leading producer of eucheumatoid seaweeds. The rapid growth in the domestic seaweed industry was due to a range of factors, including: the area is typhoon free; the seasonality and incidence of disease are minimal; the area is stable legally; farmers have clear tenure rights over farm sites; infrastructure and shipping facilities are adequate; and business essentials are available.

92    Many Indonesian coastal regions, including the Rote/Kupang region, rate well on these features. Generally, they are good for seaweed cultivation all year round and enjoy a competitive advantage over the northerly regions of the Philippines, which suffer from periodic typhoons, and the southerly regions of the Philippines, which face recurring armed insurrections that inhibit the conduct of seaweed businesses.

93    Over the past decade, Indonesia has emerged as the global “alpha” source of tropical seaweeds, meaning that it is the world’s dominant source of raw, dried seaweeds (RDS) and can conceivably supply the entire global RDS demand (around half of which is generated by processors in China). By way of example, following Typhoon Haiyan (also known as Super Typhoon Yolanda) in November 2013, spinosum production in the Philippines was virtually wiped out, causing a global shortage which was filled by Indonesian producers within several months. The most recent production data was collected in 2013. It indicates that Indonesia produced 61% of global seaweed production, which is around 300,000 wet tonnes, worth approximately US$40 million, per month. In the course of giving his evidence about the seaweed industry in Indonesia, Dr Iain Neish, who was called by the applicant, estimated that this production would have risen to well over 70% by 2019, on the basis that the industry in Indonesia continues to grow and the industry in the Philippines continues to decline.

94    The NTT province, including the Rote/Kupang region, is viewed by the industry as a region with underdeveloped potential. Dr Neish said that, as at 2018, it was not considered as a reliable, year-round seaweed source, but the region has contributed to building Indonesia’s position as an alpha tropical seaweed supplier. He said that the seaweed industry in the Rote/Kupang region had developed to the point of widespread successful farming by 2009, but this was followed by a sudden crop failure throughout the region. There had been persistent efforts to re-establish farming, which eventually recovered over a number of years.

95    The agronomic process of seaweed farming in the Rote/Kupang region is remarkably simple. It essentially involves attaching a fragment of seaweed to a line and suspending it in the water to grow. A seaweed farm will comprise several of these long lines, made from ropes, strings or strappings, which may be “hung” in the sea using a variety of configurations. Generally, empty plastic water bottles are attached at intervals to act as floats.

96    As I have mentioned above, seaweeds are able to reproduce following the production of spores from the thallus, without a fertilisation step. This feature of seaweeds enables seaweed farmers to “seed” new crops periodically using seaweed fragments from the previous crop. The seeding process usually results in three-to-five-fold growth in around six weeks, which is the usual length of the seaweed growing cycle. Dr Neish accepted in cross-examination that it was a possible “untested hypothesis” that this approach to propagation of seaweed could create a lack of genetic variation in the crop over time. However, he disagreed that this would cause the extinction of a particular variety of seaweed in a given area.

97    After the six-week growing cycle is complete, the seaweeds are harvested and dried. The most common drying technique in the Kupang/Rote region is the use of drying platforms known as “para-para”. These are constructed from bamboo strips, which make a platform frame which is covered in fine netting and on which the seaweed is laid for two to three days to dry. The dried seaweed is then sacked or baled. Dr Neish deposed that this was an excellent drying technique. It meant that seaweeds from the Rote/Kupang region are generally clean and well-dried. For this reason, RDS from the region tended to fetch prices on the high side of the Indonesian price range”. RDS is then usually sold to carrageenan processors to be made into carrageenan for industrial use.

98    Carrageenan (a hydrocolloid) is used as a thickening and emulsifying agent, primarily as a food ingredient. Its principal use is in meat packing, in which it is injected with brine into ham and other meats to keep them moist. It is also used in dairy products, for example to suspend cocoa in chocolate milk and to prevent ice crystal formation and impart a creamy texture in ice creams, and in jelly desserts. It is also used in pet food. The carrageenan derived from cottonii and sakol is called kappa carrageenan. The carrageenan derived from spinosum is called iota carrageenan.

99    Carrageenan production occurs predominantly in China, but there are also processors domestically in Indonesia and the Philippines, as well as in Europe.

100    RDS is sold for approximately US$1 to US$2 per kg. Carrageenan is sold for approximately US$10 to US$15 per kg. Indonesian export volumes of RDS (comprising 90% cottonii and 10% spinosum) grew from around 30,000 tons (worth around US$20 million) to over 100,000 tons (worth around US$110 million) between 2000 and 2008.

101    Qualitative research undertaken by Dr Neish suggests that the seaweed farming industry has provided local Indonesian residents with a major addition to their income. Seaweed is a cash crop for farmers. It can be undertaken at minimal cost. There is a ready market. Dr Neish estimates that an average seaweed farmer in the Rote/Kupang region is able to produce around 500 kg of dry Kappaphycus seaweed each month, which is generally sold for between US$4,000 and US$8,000. Seaweed farmers generally report that the income per unit effort they gain from seaweed farming is several multiples greater than income available from other sources. Indeed, few other economic choices are available. Other livelihood options in the Rote/Kupang region have tended to remain static or have declined since the development of seaweed farming. Seaweed farming thus provides an extremely important livelihood for the villagers in these areas. Dr Neish estimated that more than half the households in the region rely exclusively on seaweed farming to earn an income.

102    Dr Neish expressed the opinion that unprecedented high cottonii and sakol prices in 2018 were attracting seaweed farmers in the Rote/Kupang region to “have another go” at seaweed farming, despite their difficulties during and after the crop failure events of 2009.

the conduct of the hearing

103    The hearing of this proceeding was conducted in two broad phases. The first phase involved the taking of lay evidence from seaweed farmers in the Rote/Kupang region and other lay observers, including from that region. The second phase involved the taking of extensive expert evidence from experts across a broad range of disciplines.

The lay evidence

104    The applicant read the following affidavits of deponents who were cross-examined:

    Daniel Sanda, made on 18 August 2018;

    Silwanus Aplugi, made on 20 August 2018;

    Gustaf Lay, made on 23 August 2018;

    Gabriel Mboeik, made on 25 July 2018;

    Axel Pierre Chalvet, made on 21 August 2017;

    Adrian Sibert, made on 30 August 2018;

    Nikodemus Ndun, made on 12 October 2017;

    Lot Martinus Heu, made on 12 October 2017;

    Semin Polin, made on 15 September 2016;

    Yohan Lima, made on 13 October 2016;

    Dominggus Liman, made on 17 October 2016;

    Abner Yopi Pallo, made on 16 September 2016;

    Zadrak Patolla-Ballo, made on 26 September 2016;

    Petrus Ndolu, made on 3 April 2017;

    Abdul Rasyid Aitio, made on 23 March 2017;

    Mica Erwin Johanis Penna, made on 23 March 2017;

    Taftinus Taek, made on 22 March 2017;

    Semuel Messakh, made on 9 February 2017; and

    Yardin Adoni Lari Aplugi, made on 5 April 2017.

105    The applicant read the following affidavits of deponents who were not required for cross-examination:

    John Guiney, made on 29 September 2017;

    John Gregory Rogers, made on 30 September 2017;

    Simon Mustoe, made on 10 August 2018;

    Ghislaine Llewellyn, made on 30 August 2018;

    Ghislaine Llewellyn, made on 28 March 2019;

    James Watson, made on 30 August 2018;

    Matt Smith, made on 15 March 2019;

    Bartolo La Macchia, made on 5 March 2019;

    Antony La Macchia, made on 15 March 2019;

    Lorens Hendrik, made on 26 September 2016;

    Daud Nenokeba, made on 26 September 2016;

    Watson Sodi Mbuik, made on 17 February 2017;

    Jermias Manafe, made on 20 March 2017;

    Melkianus Mola, made on 20 March 2017;

    Marselinus Mesah, made on 3 April 2017;

    Johan Mooy, made on 2 April 2017;

    Anton Matasina, made on 8 March 2017;

    Ogus Tananggau, made on 23 March 2017;

    Resa Rehans Fatu, made on 5 April 2017;

    Thomas Dethan, made on 5 April 2017;

    Nathan Kearnes, made on 3 May 2019;

    Nathan Kearnes, made on 26 November 2019;

    Lewis Hamilton, made on 6 May 2019; and

    Lewis Hamilton, made on 18 November 2019.

The expert evidence

106    The expert evidence presented in this proceeding was extensive. It was, by and large, organised according to a number of topics, most of which are reflected in the structure of these reasons. The topics on which expert evidence was called were Satellite Imagery, Dispersants, Currents, Trajectory Modelling, Chemical Composition of Oil, Toxicology, Volume, Observations of Oil, Oil Spill Contingency Planning and the Seaweed Industry in Indonesia.

107    With the exception of Observations, Contingency Planning and the Seaweed Industry in Indonesia, expert conclaves were held in respect of each of these topics, and each resulted in a joint expert report prepared by the participating witnesses. The experts who participated in each conclave gave their oral evidence concurrently. Professor Steinberg did not participate in the Toxicology conclave. He gave his evidence and was cross-examined in the traditional manner. Dr Neish was the only expert witness who gave evidence on the Seaweed Industry in Indonesia. Dr Taylor was the only expert witness who gave evidence on Contingency Planning. Although there was no conclave on the topic of Observations, evidence was given concurrently on that topic by Professor Ball, Dr Fingas, Dr Taylor and Dr Maki.

The applicant’s expert evidence

108    The applicant called expert evidence from the following witnesses.

109    Professor Andrew Ball. Professor Ball is a Distinguished Professor who holds a PhD in microbiology and has taught and researched in environmental microbiology for 33 years. His research focusses on the interaction between pollutants in the environment and the natural microbial community; in particular, the ability of microorganisms to biodegrade petroleum hydrocarbons. Professor Ball presented five reports dealing with the topics of Chemical Composition, Toxicology and Observations, and participated in the Chemical Composition and Toxicology conclaves.

110    Dr Mervin Fingas. Dr Fingas is a scientist who holds a PhD in environmental sciences, Masters degrees in chemistry and business and has published over 950 papers, over 150 of which relate to oil spill properties and behaviour, over 100 of which relate to oil analysis, over 80 of which relate to dispersants, over 70 of which relate to oil fingerprinting and many which relate to oil or chemical toxicity. He has worked in oil spills for over 45 years, including the Deepwater Horizon spill in the Gulf of Mexico, has established a laboratory at Environment Canada to study and develop measurement techniques for oil spill behaviour, and has served on two US National Academy of Sciences committees relating to oil properties and behaviour. Dr Fingas presented six reports, one of which was revised, dealing with the topics of Chemical Composition, Dispersants, Toxicology and Observations, and participated in the Chemical Composition, Toxicology and Dispersants conclaves.

111    The respondent criticised Dr Fingas’ evidence. It noted that Dr Fingas had given evidence on “a host of topics”. It submitted that, in many respects, his evidence was “unsatisfactory, and should not be accepted on any contested issue”. The respondent appeared to advance two principal reasons for making this submission.

112    The first concerns Dr Fingas’ evidence in relation to analyses carried out by LEMIGAS, an Indonesian governmental oil and gas research organisation. The respondent’s criticism appears to be based on no more than the fact that Dr Fingas disagreed with the respondent’s own witness, Dr Stout, on what the LEMIGAS analyses revealed. In coming to his view about those analyses, Dr Fingas applied a regression analysis (discussed below) and argued that the CEN 15522 – 2 Protocol used by Dr Stout was “relatively new”—a proposition with which the respondent disagrees.

113    The second reason concerns Dr Fingas’ evidence in relation to dispersants. In giving that evidence, Dr Fingas disagreed with Dr Coehlo, who was called by the respondent, as to the interpretation of certain entries in AMSA logs concerning the effectiveness of dispersants that had been applied to the spilled oil. The authors of the entries were not called to give evidence.

114    Dr Fingas interpreted the entries as recording the percentage of oil targetted with dispersant (i.e., the percentage of oil “hit” with the dispersant). Dr Coehlo interpreted the entries as recording the percentage of oil removed from the sea surface by the dispersant.

115    Dr Fingas repeated his interpretation in oral evidence. He later developed this by saying that the percentage referred to in the entries was the percentage of oil that the operators targeted, which they felt would be dispersed by the dispersant.

116    When cross-examining counsel suggested to Dr Fingas that this was a fanciful reading of the relevant entries, he disagreed. He explained his interpretation as follows:

I’m sorry. I disagree. Because – simply because the length of time that it would take for a dispersant to actually work and for the oil to disappear from sight and which you could say was actually dispersed is too long for them to lay around in the vessel without going on to the next slick.

117    When it was put to Dr Fingas that he did not honestly believe the interpretation he had given and that, by this answer, he was attempting to make the evidence fit with his views about dispersant effectiveness, he said:

That is incorrect, because I have talked to operators in the past and this is how they’re taught. They’re taught to recognise the signs after dispersant has been applied that it may disperse or will not disperse. And so that is the percentage and very rough percentage that they will report.

118    In closing submissions, the respondent submitted that Dr Fingas had either given dishonest answers on this topic or was so biased in his views about dispersant effectiveness that he was unable to read the log entries objectively and rationally.

119    I do not accept that submission. I do not think that Dr Fingas gave his evidence on this topic, or on any other topic, dishonestly. He explained his interpretation of the log entries. I do not think that his explanation was fanciful, although his interpretation of the log entries is not one that I would adopt. I think that Dr Coehlo’s interpretation is to be preferred. However, Dr Fingas is not to be criticised for expressing a different view to Dr Coehlo on the interpretation of an operational document of which neither he nor Dr Coehlo was the author; nor is he to be criticised for expressing a different view to Dr Stout in relation to what the LEMIGRAS analyses reveal. Indeed, a feature of this case has been the remarkable number of disagreements between experts on the many issues that were canvassed across the broad range of topics considered in the evidence. I do not accept that, on the topics he addressed, Dr Fingas’ evidence was unsatisfactory. I reject the respondent’s broad submission that Dr Fingas’ evidence should not be accepted on any contested issue.

120    Dr Erich Gundlach. Dr Gundlach is a coastal geologist who has over 40 years’ experience related to oil spill assessments and the application of imagery and aerial photographs to determine spill location and shoreline impacts, and works extensively with oil spill models. His experience includes the Metula spill in the Strait of Magellan, the Amoco Cadiz spill in France, the Ixtoc 1 spill in the Gulf of Mexico, the Exxon Valdez spill in Alaska, the Gulf War spills in Kuwait and Saudi Arabia and the Deepwater Horizon spill in the Gulf of Mexico. Dr Gundlach presented three reports dealing with the topics of Satellite Imagery and Trajectory Modelling, and participated in the conclaves which took place on both of those topics.

121    Dr Graeme Hubbert. Dr Hubbert is a physical oceanographer who holds a PhD in physics and has worked in oceanography since 1981, during which time he has spent 17 years in government research institutes, including the Bureau of Meteorology (BoM), where he developed the first Australian 3D ocean model for environmental studies. In 1993, he established a consulting company called Global Environmental Modelling and Monitoring Systems Pty Ltd (GEMMS, previously referred to by the acronym GEMS), which has worked with the US Navy and, for the past 20 years, AMSA to develop ocean modelling systems applied mainly to search and rescue operations and environmental impact studies. Dr Hubbert presented two reports dealing with the topics of Trajectory Modelling and Currents, and participated in the conclaves which took place on both of those topics.

122    Dr John Luick. Dr Luick is a physical oceanographer who holds a PhD in that field and works as a consultant through Austides Consulting, which specialises in marine environmental consulting and marine software development, which he established and operates. He also holds appointments as an Honorary Senior Lecturer at Flinders University, a Visiting Scientist with the South Australian Research and Development Institute, and an Expert Adviser at Tridel Engineering (Dubai). He has over 30 years’ experience in oceanographic research and consulting. Dr Luick presented one report dealing with the topics of Trajectory Modelling and Currents, and participated in the conclaves which took place on both of those topics.

123    Dr Iain Charles Neish. Dr Neish is a marine biologist and businessman who holds a PhD in zoology and has worked with seaweeds and seaweed farmers in aquaculture systems since 1965. Dr Neish has extensive experience in seaweed value chains and the development of seaweed aquaculture agronomy systems on every continent except Antarctica. Over the past 41 years, he has been involved with the seaweed industry in South East Asia, and has been particularly involved in that industry in Indonesia since 1986, during which time Dr Neish played a role in industry development for the carrageenan industry and other seaweed industry diversification and development ventures. Since 2008, Dr Neish has also participated in surveys and value chain analyses that have included engagement with hundreds of active seaweed farmers in Indonesia. Dr Neish is currently undertaking seaweed industry development ventures as a Research and Development Advisor to PT Sumber Tanaman Samudra, a seaweed farming company, and as a Director of PT Sea Six Energy Indonesia, a seaweed processing company. A more comprehensive summary of Dr Neish’s qualifications may be found in Sanda v PTTEP Australasia (Ashmore Cartier) Pty Ltd (No 6) [2019] FCA 1853 (at [4] – [9]), which dealt with various objections which were made to his expert report. Dr Neish presented one report dealing with the topic of the Seaweed Industry in Indonesia, and did not participate in any conclave.

124    Dr Janet Sprintall. Dr Sprintall is an observational physical oceanographer who holds a PhD in that field and has researched large-scale ocean circulation and inter-basin exchange at the Scripps Institute of Oceanography since 1993. She has a particular interest in the physical oceanography of the marginal seas in the Western Pacific Ocean, including the Indonesian, Philippine and Solomon archipelagos, and has spent the past 20 years researching the Indonesian Throughflow current (ITF) which runs between the Pacific Ocean and Indian Ocean. Dr Sprintall presented one report dealing with the topics of Trajectory Modelling and Currents, and participated in the conclaves which took place on both of those topics.

125    The respondent criticised Dr Sprintall’s evidence as it related to the reliability of the modelling evidence I discuss in later sections of these reasons. Dr Sprintall said that she had only limited confidence in the two models discussed. In the course of propounding the reliability of the modelling it advanced (SIMAP/SUNTANS), the respondent submitted that Dr Sprintall appears to have adopted (perhaps subconsciously) the role of an advocate for the applicant’s case. The respondent submitted that Dr Sprintall’s presentation in the concurrent evidence session on Trajectory Modelling was “more in the nature of a submission than an independent opinion based on her expertise”. This was because, in the respondent’s submission, Dr Sprintall took it upon herself to express a view about the likelihood of Montara oil reaching NTT based on her assessment of the reliability of the applicant’s lay evidence.

126    I do not accept that criticism of Dr Sprintall’s evidence. Dr Sprintall’s view was that all models have errors and uncertainties. She also noted that there was relatively poor agreement between certain buoy trajectories and the SIMAP/SUNTANS model trajectories advanced by the respondent. In the course of expressing that view, Dr Sprintall said:

So probably the best verification as to the reliability of the trajectories of the oil and/or the dispersants comes from the observations of oil and sheen in the Timor Sea evident in the AMSA daily maps and the surveillance flight reports, as well as the multiple firsthand eyewitness accounts of the presence of oil along the coast of the regencies of Rote and Kupang in Indonesia. That is the only way that the oil could have been observed in the Timor Sea is because the ocean currents had carried it there.

127    I do not accept that, in giving that evidence, Dr Sprintall was acting as an advocate for the applicant or advocating the reliability of the lay witness accounts of oil sightings. As an observational physical oceanographer, Dr Sprintall was doing no more than pointing to data sources that she thought might be more reliable than the modelling on which the parties were relying, her assumption being (without expressing a view either way) that the eyewitness accounts of oil sightings were themselves reliable. I do not accept that Dr Sprintall was purporting to express a personal view about the reliability of the lay evidence or in any way intending to usurp the role of the Court in fact-finding.

128    Professor Peter Steinberg. Professor Steinberg is a professor of biology at UNSW and Director and CEO of the Sydney Institute of Marine Science, who holds a PhD in biology and whose expertise is in the fields of seaweed biology and ecology and the coastal ecology of systems across the world. Professor Steinberg is one of the more senior seaweed ecologists and coastal ecologists in Australia. He has over 30 years’ experience in marine biology and ecology, has authored nearly 200 refereed papers or book chapters concerning the ecology, biology, chemistry or biotechnology of seaweeds and/or aspects of coastal ecology; is the inventor named in nine patents; and is considered one of the founders of the field of marine chemical ecology, particularly as it relates to seaweeds. Professor Steinberg presented two reports dealing with the topic of Toxicology. He did not participate in the conclave on that topic or give concurrent evidence.

129    Dr Anitra Thorhaug. Dr Thorhaug is a marine botanist and marine ecologist who holds a PhD and Masters degrees in marine biology and chemical oceanography, and has postdoctoral research experience in algal physiology and the membrane biophysics of algae. She is the president of an environmental foundation called the Greater Caribbean Energy and Environment Foundation. Dr Thorhaug has over 50 years’ experience in marine pollution work. Her relevant experience for tropic benthic oil spill work includes assessing the impacts of the 1991 Kuwait spill on the Arabian Gulf for the United Nations International Oceanographic Commission, acting as an adviser to British Petroleum in the National Environmental Resource Damage Assessment process of the Deepwater Horizon spill, exploring nearshore Caribbean oil spill methods, and creating a protocol for the Caribbean region. She has received a large number of awards for her work, including several Lifetime Achievement Awards and the 1982 United Nations Environmental Program Gold Medal for “Decade of Distinguished Research in Tropical Pollution and Ecology”. Dr Thorhaug presented one report dealing with the topic of Toxicology, and participated in the conclave which took place on that topic.

130    Professor Brian Towler. Professor Towler is a professor of chemical engineering and Chair of Petroleum Engineering in the Centre for Coal Seam Gas in the School of Chemical Engineering at the University of Queensland, and holds a PhD in that field. He has been conducting research into aspects of oil and gas production operations for over 30 years, including as the head of the department of Chemical and Petroleum Engineering at the University of Wyoming. He has also worked in the gas industry, including bringing the Mereenie oil field in Queensland into production. He has been an expert witness in several oil spill cases, including giving evidence regarding the causes of the blowout which occurred in the Deepwater Horizon spill. Professor Towler presented three reports dealing with the topic of Volume, and participated in the conclave which took place on that topic.

131    Professor Steven Wereley. Professor Wereley is a professor of mechanical engineering at Purdue University and holds a PhD in that field. His professional expertise is in quantitative image analysis, which he describes as “quantitative evaluation of images for the purpose of extracting flow information”, and he is the author of a leading text in that field, Particle Image Velocimetry: A Practical Guide. He has calculated the volume of the Deepwater Horizon spill, the Royal Dutch Shell pipeline weld failure in Nigeria in 2008, and the Tesoro spill in North Dakota, United States of America in 2013. Professor Wereley presented two reports dealing with the topic of Volume, and participated in the conclave which took place on that topic.

The respondent’s expert evidence

132    The respondent also called expert evidence from the following witnesses.

133    Dr Martin Blunt. Dr Blunt is the Shell Professor of Reservoir Engineering at Imperial College London and a fellow of the Royal Academy of Engineering. He holds a PhD in physics and has written two textbooks on reservoir engineering, one of which describes the material balance methodology he used in the context of this proceeding to calculate the volume of oil released during the Montara oil spill, and has taught on the subject of fluid flow principles for over 20 years. He has also published over 200 papers and won a number of awards for his teaching and research in this field. Dr Blunt was also involved in calculating the volume of oil released in the Deepwater Horizon spill. Dr Blunt presented one report dealing with the topic of Volume, and participated in the conclave which took place on that topic.

134    Dr Gina Coelho. Dr Coelho is an oil spill response scientist with Sponson Group Inc. and a member of the permanent oversight panel for the annual Clean Gulf conference and the Oil Spill Recovery Institute (Alaska-based) Science and Technology Committee. She holds a PhD focussed on dispersant use and response policy and has over 25 years’ experience working as a dispersant subject matter expert in environmental research, consulting, program management, group facilitation, and regulatory compliance. This experience includes co-facilitating approximately 20 Ecological Risk Assessments for the US Coast Guard for spill response planning, establishing three dispersed oil testing facilities in the United States of America, Brazil and New Zealand, and supporting responses to over 100 oil spills worldwide, including assisting British Petroleum in subsea dispersant injection testing and operational monitoring in the Deepwater Horizon spill. Dr Coelho also lead authored the BP Oil Spill Dispersant Use Manual and co-authored the IPIECA Subsea Dispersant Injection Good Practices Guide, and is a member of a panel of authors which published a book on deep water oil spills and future response technologies. Dr Coelho presented two reports dealing with the topic of Dispersants, and participated in the conclave which took place on that topic.

135    Dr Deborah French-McCay. Dr French-McCay is an environmental consultant oceanographer who holds a PhD in oceanography and has approximately 35 years experience in oil spill modelling. She specialises in model development and the application of models to various oil spills, for planning response and risk assessments and to support natural resource damage assessment. Models which she has developed have been applied worldwide for response planning, risk analyses, and spill assessments. Dr French-McCay has worked as an expert for the US National Oceanic and Atmospheric Administration (NOAA), with which she most recently performed modelling of the Deepwater Horizon spill as lead of the Offshore Water Column Technical Working Group, which evaluated the trajectory and fate of the oil and impacts to marine fish and invertebrates. Dr French-McCay is also the principal investigator and primary author of more than 100 technical reports and papers, some of which document the development, algorithms and assumptions of the oil spill model which she utilised to provide her evidence in the present case. Dr French-McCay presented four reports dealing with the topic of Trajectory Modelling and participated in the conclave which took place on that topic.

136    The applicant objected to certain passages in Dr French-McCay’s report dated 8 December 2019, which was provided in response to analyses undertaken by the applicant comparing Dr French-McCay’s trajectory modelling with recorded observations of Montara oil taken from other data (for example, AMSA observations). The applicant’s analyses were provided in an affidavit made by Nathan Kearnes, which I discuss in a later section of these reasons. Because Dr French-McCay’s 8 December 2019 report and Mr Kearnes’ affidavit were provided late in the hearing, the parties were content for me to deal with the objections to the report in these reasons. Given the nature of the objections, I am satisfied that this was an appropriate course to adopt.

137    The applicant’s objections are to paragraphs 3 to 5 (including Table 1) and to parts of paragraphs 6, 15 and 16 of the report. The single objection is that these paragraphs or parts of paragraphs are not responsive to Mr Kearnes’ affidavit. The respondent’s response is that these paragraphs or parts of paragraphs should be admitted as providing necessary context and background to the opinions expressed by Dr French-McCay. I accept the respondent’s submissions. I consider these paragraphs and parts of paragraphs to be responsive to Mr Kearnes’ affidavit. I will, therefore, admit this evidence.

138    Dr Oscar Garcia-Pineda. Dr Garcia-Pineda is a geoscientist with more than 15 years’ experience in the management of projects related to aerial and satellite remote sensing. He is the director of Water Mapping LLC, which provides remote sensing of oil spill services, and an adjunct scientist at the Florida State University Center for Ocean-Atmospheric Prediction Studies. Dr Garcia-Pineda’s experience includes work with the National Aeronautics and Space Administration (NASA), to develop and study the capability of sensors to detect surface ocean features, and with NOAA, on a number of spills. With these organisations, Dr Garcia-Pineda developed an image processing algorithm for mapping oil spills from synthetic aperture radar imagery, which has been adopted as the operational tool for the semi-automatic detection of oil spills in the United States of America. He has also authored and co-authored more than 20 articles in this field and presented at numerous conferences. Dr Garcia-Pineda presented one report dealing with the topic of Satellite Imagery and participated in the conclave which took place on that topic.

139    Professor Gregory Ivey. Professor Ivey is a physical oceanographer at the University of Western Australia who holds a PhD on ocean mixing and ocean dynamics. He has worked in ocean processes, ocean circulation and ocean mixing, including numerical ocean circulation modelling, for 39 years. His specific areas of expertise include tide and wind-drive flows in the coastal ocean, the impact of tropical cyclones on the ocean, internal tidal flows and small scale turbulent mixing. He has published nearly 200 academically refereed publications describing his research on these processes. He has worked extensively on physical oceanographic processes in the waters of the Australian North Shelf (extending from Ningaloo in Western Australia into the Timor Sea), and in those regions has conducted observational work using fixed moorings and ship-based observations of ocean currents and turbulent ocean mixing. Professor Ivey presented one report dealing with the topics of Currents and Trajectory Modelling, and participated in the conclaves which took place on both of those topics.

140    Dr Alan Maki. Dr Maki is a toxicology and water quality biologist who holds a PhD in fisheries and wildlife management, and has worked in environmental research and management, including in various aspects of the effects of oil on the environment, for over 45 years. He has published over 200 papers and books on numerous aspects of aquatic toxicology and chemistry and served as Science Advisor to the US Environmental Protection Agency for 12 years. Dr Maki was also Chief Environmental Scientist at Exxon Mobil, where he worked for over 35 years, including as Chief Scientist for the science and environmental impact studies completed in relation to the Exxon Valdez spill in 1989, and Chief Scientist for British Petroleum during the Deepwater Horizon spill. He is the former president of the Society of Environmental Toxicology and Chemistry. Dr Maki presented one report dealing with the topic of Toxicology, and participated in the conclave which took place on that topic.

141    Dr Scott Stout. Dr Stout is an organic geochemist who holds a PhD in geology and has over 30 years’ experience in geochemistry and the petroleum industry. He has expertise in the chemical composition of natural/shale gas, crude oil, coal, manufactured gas plant, gasoline, diesel and other fuel-derived sources of hydrocarbons in terrestrial and aquatic environments, and has authored or co-authored over 160 publications, including three textbooks concerning environmental forensic aspects of oil spills. Dr Stout has worked in both an exploration and production capacity for the oil industry (including in Indonesia) and in an environmental management capacity. His company has, for the past 15 years, provided petroleum chemical analysis of oil spill events to understand the effect of oil on the environment and chemical changes experienced by oil after it is released into the environment. Dr Stout’s recent experience includes working for NOAA in response to the Deepwater Horizon spill, during which his laboratory analysed over 34,000 samples and provided approximately 18 reports to the US government to assist it in settling its case with BP plc. Dr Stout presented four reports dealing with the topics of Dispersants and Chemical Composition, and participated in the conclaves which took place on both of those topics.

142    Dr Elliott Taylor. Dr Taylor is an environmental consultant who holds a PhD in oceanography and has worked in oil spill response, specialising in planning, preparedness and shoreline response, for over 30 years. His experience includes responding to the Exxon Valdez spill, the Deepwater Horizon spill and a range of other oil spills in marine environments and in inland waters and rivers. He has extensive experience in spill preparedness, including developing over 100 oil spill contingency plans, training programs and preparedness evaluations for industry and government. He has worked with the International Maritime Organisation to assist various countries to develop their spill preparedness capabilities, and developed a range of best practice guides, manuals and tools to assist in this process. Dr Taylor presented two reports dealing with Contingency Planning and Observations. He did not participate in any conclaves.

143    Dr Michael Zaldivar. Dr Zaldivar is a flow assurance engineer who holds a PhD in chemical engineering and has worked in the oil and gas industry for 17 years. He is the co-founder of a flow assurance company called evoleap, through which he provides expertise in multi-phase flow in pipes (being the concurrent flow of oil, gas and water in pipes). Dr Zaldivar has participated in over 50 oil and gas projects around the world, including as a multiphase flow expert in litigation concerning the Deepwater Horizon spill. Dr Zaldivar presented one report dealing with Volume, and participated in the conclave which took place on that topic.

The lay evidence

The applicant’s evidence

144    The applicant is a seaweed farmer. He lives with his family in Oenggaut, a village located near the south west extent of rote. Oenggaut is divided into five sub-villages. The applicant has lived in one of these, Tunggaoen Timur, his whole life. This sub-village and another have a combined population of 327 residents. The applicant said that his village was a traditional one that centres on family, church and community.

The applicant’s education and life before seaweed farming

145    The applicant made an affidavit on 18 August 2018 in which he deposed to the nature of his seaweed farming business, his observations of the arrival of (what he described as) oil in the waters off Inggurae Beach, and the effect of that event on his business. A redacted version of this affidavit is in evidence—redacted because, like a number of other witnesses, the applicant gave oral evidence of his observations and on other matters in contest in the proceeding.

146    The applicant does not speak English, but he made the affidavit in English with the assistance of two interpreters. His native languages are Bahasa Indonesia and the Delha dialect native in the west of Rote. The applicant deposed that he reads, writes and understands Bahasa Indonesia well. He deposed that the affidavit was carefully read aloud to him by one of the interpreters before he signed it, and that he understood it well. A version of the affidavit translated into Bahasa Indonesia and dated 18 August 2018 was admitted into evidence. The extent to which the applicant was involved in the preparation of the affidavit was the subject of cross-examination, which I address below.

147    The applicant attended primary school for five years from the age of ten. When he left school in 1973, he helped his mother and step-father with family crops. He lived at home with them until he was married under local custom in 1982 to Viktoria Sanda Bessie. He and his wife moved into a house he built. They have five children.

148    After the applicant married, and before he started seaweed farming, he earned income from extracting products from palm trees to make sugar, and from fishing. His estimated average annual income before he started seaweed farming was less than 2,000,000 IDR, which was just enough to survive and provide for his family.

Seaweed farming

149    In 2000, the applicant was introduced to seaweed farming by the then bupati of Rote. Officials from the Rote Department of Marine Affairs and Fisheries provided the applicant and others with seaweed seed (small pieces of fresh new branches from a clump of seaweed) of the cottonii hijau and cottonii merah varieties, and ropes on which to grow them. The applicant has only ever grown cottonii hijau seaweed.

150    The applicant first began seaweed farming as part of a collective. From when he started in 2000, his returns were regular enough that he stopped other work. The applicant does not remember how much money he made in 2000-2001, but his evidence is that it was several times more than he had earned in the years before. He was able to improve his home, pay for one child’s education, and plan to educate another.

151    In 2002, with the encouragement of the government, the applicant and the other nine members of his collective agreed to split up their seaweed equally and start new, individual plots using the seaweed as seed. The government provided rope as a grant, and the applicant procured other materials he required.

152    The applicant selected a new area for his plots after carefully observing the water off Inggurae Beach, which was close to his home. The first he selected is identified as Plot 3. He did not buy this plot or any other since. The process for selecting a plot was done by informal discussion and agreement with other local villagers who had an interest in seaweed farming, including the heads of his sub-village and another neighbouring it. The applicant says his ownership of Plot 3 and his other plots has always been recognised by other villagers and that he has never heard of any ownership disputes between others.

153    The framework on which he grew his seaweed consisted of wood posts driven into the sea floor, between which ropes of about 1.5 m in length were tied. The seaweed, and plastic bottles to keep it afloat, were tied to the ropes.

154    The applicant said that he harvested seaweed year-round, although most harvesting took place in the nine months of the dry season. The seaweed usually took about 35 days to mature from seed. It was harvested from the ropes and then dried in the sun. The dried seaweed was then stored in large bags typically weighing about 70-80 kg.

155    The applicant said that approximately once a month a collector representing a seaweed buyer bought the seaweed. The price was determined by the buyer, who also weighed the seaweed once the applicant had agreed to sell it.

156    During the wet season, the seaweed continued to grow but the harvest and drying stages were made somewhat more difficult by wet weather.

157    The applicant’s recollection was that:

(a)    From 2002-2005, the crops always performed well.

(b)    2006 was a good, consistent season without any unusual weather events.

(c)    2007 was similar to 2006, but with more favourable conditions in the wet season, meaning that he and his wife could access the plots more often during the wet season than in 2006. He said that the seaweed crop grew well.

(d)    2008 was the best year for growing seaweed. February was calm and good harvests could be made from early February. The conditions during the wet season at both the start and end of 2008 meant harvests could be done regularly throughout the rainy season. The applicant said that, in 2008, a long rope could produce five baskets of wet seaweed.

(e)    In 2009, the harvest started late because of wet and windy conditions in February. The applicant said that harvesting ended when the oil arrived: see below.

(f)    In 2010, there was no harvest at the beginning of the dry season because his crops had died in 2009. Harvesting only started when there was enough seaweed on the ropes. During examination in chief, the applicant said that towards the end of 2010, he was growing seaweed on 17 ropes, but that none had developed sufficiently to harvest. In 2010, he only filled part of one basket of wet seaweed per day.

(g)    In 2011 and 2012, harvesting could not be commenced at the beginning of the dry season because the crops were still recovering from the damage of 2009. In 2011, the rain was comparatively heavy in the wet season but that the dry season was much the same as in previous years. In 2012, he had about 160 ropes of seaweed growing and the weather was similar to 2011. His wife, Viktoria, looked after it mostly. At this time, on harvesting, a long rope produced only about two baskets of wet seaweed, compared to five in 2008, as the seaweed was not as healthy and grew poorly.

(h)    In 2013 and 2014, the seasonal conditions were good and consistent. In 2013, he grew seaweed on 200 ropes. On harvesting, a long rope yielded about two to three baskets of wet seaweed. In 2014, he grew seaweed on 227 ropes. Once again, on harvesting, a long rope again yielded about two to three baskets of wet seaweed.

(i)    In 2015, the seasonal conditions were good and consistent until stormy weather around Christmas time. He still had 227 ropes of seaweed and the yield of a long rope, on harvesting, remained at about two to three baskets of wet seaweed.

(j)    In 2016 and 2017, February was stormy and windy and harvesting started a little later than in 2015.

The applicant’s business practices

158    The applicant deposed that he had never used the Internet or owned a computer. He said that he makes phone calls with his mobile phone but does not know how to take photos with it. He said that he has not seen maps like those exhibited to his affidavit until they were shown to him by his lawyers in the preparation of the affidavit. He deposed that he has never kept written records of his seaweed business and does not know of any farmer in Oenggaut who does.

159    The applicant said that his income and expenses for his seaweed business have always been paid in cash. From about 2004 to 2007 and from 2018 until the time of the hearing, he kept a bank account in which he would occasionally deposit extra profits. However, he said that he does not know how to make money transfers and relies on his half-brother to occasionally make transfers for him. The applicant said that he does not budget for the future; it is his custom to live day by day.

The oil

160    The applicant said that in September 2009 he saw oil appear in the sea at his plots and that the seaweed died shortly after. Although unsure of the exact date, he said that this happened between the middle to the end of that month.

161    He said that the oil took the form of yellow-grey blocks and that the sea was otherwise dark. The blocks were about the size of a golf ball. When he touched the blocks, his hands felt smooth. As for the seaweed, it looked the same for two or three days but then turned white. Aside from the blocks in the water, the applicant also said that he saw rainbow colours in the water when the sun rose on the first and second day that the oil was there. The applicant also said that he saw many dead fish of various kinds at this time.

162    The applicant said that after the first two days he saw more dead fish. He also smelled the scent of oil. He saw yellowish blocks attached to his ropes and seaweed crops. Upon entering the water, his skin felt smooth with oil.

163    He observed other farmers’ plots. They were similarly affected. He observed the same conditions at a different site about 1 km from his home. In the following days, the applicant checked on his own plots and found that the water had not cleared. The seaweed had started to turn white. As more days passed and the water cleared, he found that his seaweed was gone.

After the oil

164    The applicant said that, in about March 2010, he bought some more seed and tied it to his ropes. The plants died after a week.

165    Later in 2010, the applicant was supplied with seed by the Department as a grant. He was told it was called sakol. It was different to the cottonii seaweed that he had grown before the oil arrived. The applicant said that he has only grown sakol since; cottonii has not been available. As I have already recorded, he was unable to grow seaweed that could be harvested for sale in 2010.

166    The applicant said that in the low-yield years, especially 2010-2011, the sakol seaweed he grew was soft and mushy”, and resulted in a lighter-weight dry seaweed compared to the cottonii.

The effect of the oil in 2009 on the applicant’s life and income

167    The applicant said that the oil that arrived in 2009 killed his seaweed crop and that his business has never fully recovered. Prior to 2009, he says he could provide for his family comfortably and give money to his Church. He said that, today, the seaweed is growing again but does not provide as much income for him as it did prior to 2009. He said that he has taken on labouring work as well as managing the seaweed plots with his wife.

Cross-examination

168    The applicant was taken to his affidavit in cross-examination. He was challenged on the extent to which he had prepared to give his evidence, and the extent to which he had been involved in the preparation of his own affidavit given his limited understanding of English and limited literacy in Bahasa Indonesia. He said that he had not read the Bahasa version of his affidavit before swearing the English version, but that the Bahasa one had been read to him by an interpreter.

169    In closing submissions, the respondent said the applicant’s evidence should be treated with considerable caution, noting that he gave evidence of events up to 12 years prior with a level of detail at times that was surprising given the passage of so much time. Further, the respondent submitted that there were inconsistencies in the applicant’s evidence that he was unable to explain, which suggested a lack of candour.

170    Despite some inconsistencies in his evidence, I accept that the applicant was an honest witness. I accept the evidence he gave concerning the observations he made in September 2009 when oil appeared in the sea at his plots. I deal with the respondent’s other criticisms of the applicant’s evidence when considering the calculation of the applicant’s damages.

The evidence of other seaweed farmers, village heads and other lay observers

171    A large body of evidence was called from other seaweed farmers and the heads of villages in the Rote/Kupang region concerning the observations they made in late 2009, particularly in the September/October period, of the presence of (what they described as) oil on beaches and in the surrounding waters, and the impact of that oil on the seaweed crops growing in the area at that time. In many cases, this evidence was given orally through interpreters and subjected to cross-examination.

172    As to be expected, the evidence differed from witness to witness, no doubt reflecting each witness’ attempt to recall events that had occurred many years beforehand. There were, however, noticeably recurring observations across the evidence, such as the observations of waxy clumps of material, variously coloured, that was slippery or oily to the touch and which irritated the skin, and the presence of rainbow-coloured sheen on the water. Some witnesses spoke of the smell of kerosene or diesel oil or other fuel smells.

173    The substance of this evidence is summarised in Schedule C to these reasons. A number of the witnesses referred to maps (some maps better than others) showing the locations of their villages or seaweed farms. Where possible, I have included these maps in Schedule C. Schedule C also summarises the evidence of lay observers who were not seaweed farmers or heads of villages.

174    The respondent challenged the reliability of this evidence, submitting that the Court should exercise considerable caution before accepting it. The respondent pointed out, correctly, that, in their affidavits and oral evidence, the witnesses were relying on their memories of events which occurred many years ago. The respondent submitted that it is highly improbable that anyone would could recall with precision the characteristics of a substance observed almost ten years ago. The respondent pointed to the lack of photographs, notes or other contemporaneous records to support these recollections.

175    The respondent also contended that, in the case of the seaweed farmers, there was a risk that the potential of receiving a financial benefit from this litigation “impacted the construction of the impression that they presented”. It is not entirely clear to me what the respondent means by that expression. I assume that the respondent means that, subconsciously, the witnesses presented their recollections in a way that enhanced their prospects of receiving a financial gain.

176    The respondent devoted a section of its written submissions to outlining the respects in which it contended that the testimony of some of the witnesses was unsatisfactory, or should be treated as unreliable, either generally or in certain respects, or should be treated with caution.

177    One recurring theme in the respondent’s submissions was that the memories of some seaweed farmers had been contaminated, to a very large extent, by “consensus” versions of the facts they had discussed many times. To explain, the process of information gathering for the purposes of this proceeding involved meetings attended by seaweed farmers from particular locales. These meetings were called “sign up” meetings—meaning that the seaweed farmers were “signed up” to the class action (the definition of Group Members in the further amended statement of claim includes the requirement that they have signed a particular funding agreement before the commencement of the proceeding). At the “sign up” meetings, a number of topics were discussed, including the nature of the proposed class action and certain facts about the Montara oil spill. Information was collected and entered into standard forms—called Form V2 and Form V3. Form V2 was a summary of seaweed production for the locale for certain years. It contained collective information concerning the nature of the seaweed crops and the quality of seaweed grown in those years. It also provided information on average sales prices for dried seaweed in the relevant locale. Form V3 was a listing of seaweed farmers and the quantity of their individual dried seaweed production in 2008.

178    Affidavits, annexing some of these forms, are in evidence. Some of the witnesses were cross-examined on the information contained in them. These forms provide the genesis for the contention that, through the instrumentality of these meetings, the seaweed farmers reached consensus views. In cross-examination, this was extended to consensus views of the observations that had been made of the arrival and the appearance of oil in the coastal regions of Rote/Kupang, and in the seaweed farms. This led the respondent to submit that the Court could have no confidence that the descriptions of what the witnesses saw, as recounted in their affidavits or oral evidence, actually reflected their recollections as opposed to impressions drawn from, or at least strongly influenced by, many discussions which must have taken place.

179    The caution expressed through this submission is entirely appropriate. Generally speaking, the witnesses who were cross-examined in this way accepted the possibility, when it was put to them, that their recounting of what they observed in the water in late 2009 was, or might have been, affected by discussions they had had over the years or accorded with a consensus view. This acknowledgement sounds to their credit. My impression is that all the witnesses spoke frankly on this topic, although at times some had difficulty in following the line of questioning put to them. What does strike me is that the descriptions given by the witnesses of what they observed are not uniform, but vary in matters of detail. This will be apparent from the summaries I have provided in Schedule C. These descriptions are sufficiently different, in each case, to lead me to conclude that, when giving their evidence, the witnesses were relying on, and seeking to express as best they could, their personal observations, albeit that these observations might have accorded with a view that the witness regarded as also shared by others and was thus, in that sense, a consensus view. Some witnesses advanced support for their observations by reference to the fact that others had made a particular observation. In such cases, I do not think that the witness was resiling from his evidence that he, personally, had also made that observation.

180    Some witnesses gave an account in oral evidence of their observations that varied from the account given in their affidavit evidence. However, when this happened, the variation was, generally, to supply greater detail of what they had observed, which had not been included in the affidavit. Once again, when challenged, the witnesses adhered to the truthfulness of their oral accounts.

181    I now turn to consider the respondent’s criticisms of some particular witnesses.

182    The respondent criticised the evidence given by Mr Nikodemus Ndun. Mr Ndun is a shopkeeper and seaweed trader in Nemberala village. Amongst other things, Mr Ndun gave evidence of seaweed prices in various years.

183    The respondent submitted that Mr Ndun was “not an impressive witness” and that his evidence should be treated with “great caution”. This submission was based on Mr Ndun’s evidence of the number of truckloads of seaweed he sold in the wet and dry seasons, and of seaweed prices in 2008. I make no findings in that regard. It has not been necessary for me to rely on Mr Ndun’s evidence of these matters. The respondent did not challenge the reliability of Mr Ndun’s evidence of seaweed prices in other years.

184    The respondent submitted that the evidence given by Mr Silwanus Aplugi was so unsatisfactory that it should not be accepted as a whole. Mr Aplugi is a teacher and a seaweed farmer. At the time he gave his evidence, he was also the head of Anarae village.

185    According to the respondent, Mr Aplugi had knowingly given a false estimate of his seaweed production in 2008, when signing a Form V3. The figure given in that form for Mr Aplugi’s production was 10,000 kg. In oral evidence, Mr Aplugi said that his production was, in fact, between 15,000 to 17,000 kg, but he only put in 10,000 kg for the purposes of the form. Earlier in his cross-examination, Mr Aplugi said that his production for 2008 was 11,000 kg.

186    Mr Aplugi did not accept that the entry he had made in the Form V3 was false. He said that he thought that the Form V3 was only for “data collecting” purposes and that, for that reason, it was “okay” for him to “give a false number”.

187    The respondent also pointed to the fact that Mr Aplugi had attended a “sign up” meeting and had been told that oil from the Montara oil spill had killed the farmers’ crops in 2009.

188    Based on the differences in the figures given by Mr Aplugi for his seaweed production in 2008, his acknowledgement that the figure he gave for the Form V3 was “false”, and the fact that Mr Aplugi had attended the “sign up” meeting and been told of the Montara oil spill, the respondent submitted that the Court could have “no confidence” in his evidence.

189    I do not accept that submission. Mr Aplugi was cross-examined at length, but he was not challenged on his account of his observations about the death of his seaweed crop in mid to late September 2009. His account included his observation of “bubbles of oil” in the form of “candles and liquid” that were “yellowish and chocolate”. He observed that his seaweed became soft. There were dead birds and fish, and coral had become detached. He experienced itchiness in his arms and legs after he had entered the water and came into contact with these substances.

190    It is not necessary for me to resolve the differences concerning Mr Aplugi’s seaweed production in 2008. It is enough for me to say that, whatever criticisms might be made of his evidence on that topic, I am not persuaded that the evidence of his observations of oil in mid to late 2009 was false or cannot be relied upon, particularly when the respondent did not challenge those observations in any way. The fact that some years later Mr Aplugi was told about the Montara oil spill does not lead me to a different view.

191    The respondent submitted that the evidence given by Mr Axel Chalvet should not be accepted. Mr Chalvet is a French national who lives near the village of Boa, on the south-western coastline of Rote. He said that, in September 2009, he observed a large amount of “waxy, white greasy substance” floating all over the ocean. He said that it looked like “a very large river”, approximately “a couple of hundred metres” wide. He saw it moving east to west with the wind. At the beach at Boa, Mr Chalvet noticed that this material was “everywhere, accumulating in little whirlpools all over the beach on the sand, making clumps”. He noticed that this substance came and went over “a couple of weeks”. When he went fishing about 10 km to the south of Ndana Island (which is just off the coastline of Rote, near Boa), Mr Chalvet noticed a lot of grease and wax floating around. He said that it made his boat “really dirty”. He had to clean it “more than once”.

192    Mr Chalvet also observed “quite a bit of waxy substance” at Kite Beach—so named because it is a location for kite-surfing. It is an eastern facing beach on the southern coastline of Rote, also near Boa. He saw this waxy substance accumulating on the beach. Mr Chalvet also observed “pools of wax” at Oenggaut Beach, although there was more of this in the water than on the beach because the beach is west-facing. He observed the loss of seaweed crops at this time.

193    The respondent submitted that Mr Chalvet’s evidence should not be accepted because, it said, this evidence varied significantly from “the contemporaneous records of his observations”. The contemporaneous record was an email that had been broadcast by Mr Chalvet’s mother on 19 October 2009 referring to “stinky and oily pollution” reaching Rote’s shores. Mr Chalvet’s mother wrote:

One can see the white and yellow poisoning foam coming toward the beaches instead of dolphins and whales as usual.

194    Mr Chalvet accepted that this was a description he had given to his mother at the time, although in one part of his evidence he referred to this description as his mother’s, not his. In cross-examination it was put to him that this was, in fact, the best description of what he had seen. It was also put to him that what he had observed were algal blooms in the water. Mr Chalvet disagreed. He affirmed what he had seen, saying:

... what I saw was waxy and greasy. White and yellow foam is not waxy and greasy, plankton bloom is not greasy and waxy and not stinky.

195    Later, the following exchange took place:

And what I want to put to you is that this email and the description in the paragraph starting “one can see” represent a more reliable record of your observations in 2009 than what you can recall now?---That’s up to you to make that decision. You know, I don’t think so. I know exactly what I saw and I remember it very well, sir.

196    Although Mr Chalvet appears to be the provenance of the information in his mother’s email, it is not entirely clear to me that the use of “foam” was, in fact, of Mr Chalvet’s choosing. But even if Mr Chalvet did use the word “foam” when speaking to his mother, I do not see this word as encapsulating the entirety of Mr Chalvet’s observations at the time, remembering that his mother’s email also referred to “stinky and oily” pollution reaching Rote’s shores.

197    Mr Chalvet was an impressive witness. He gave his evidence confidently and calmly. He did not strike me as someone who would give the account he did, unless he was certain of what he had seen. I accept his evidence. I do not accept that his mother’s email provides a more reliable record of his actual observations at the time.

198    The respondent also submitted that the substances that Mr Chalvet saw were the same as the substances sampled by a Mr Sibert and supplied to Leeder Consulting for analysis. I discuss the Sibert sample in greater detail below. As I there explain, the integrity of that sample is seriously in question. No sound factual findings can be made about whether it contained or did not contain Montara oil at the point of its collection in late September 2009.

199    The respondent submitted that Mr Gabriel Mboeik’s evidence was unreliable and should be treated with considerable caution. Mr Mboeik is a seaweed farmer from Oelua village. He gave evidence that, in September 2009, the sea where his seaweed was grown was full of colours, and that the ropes on which his crop was grown were yellowish in colour with the seaweed chocolate in colour. He said that the smell of the seaweed was like “solid oil” and that it was soft to touch and made his skin feel itchy. He said that he saw dead fish in the water and, where there were trees, blocks of oil were attached to them. He said that, in the following days, his seaweed died and was washed away.

200    The respondent’s criticism of Mr Mboeik’s evidence was based on his denial in cross-examination that, before September 2009, any part of his seaweed had turned white or gone limp or soft, or had broken off his ropes and washed away.

201    Mr Mboeik later accepted that some of his seaweed had, in fact, broken off and washed away in the windy conditions in January and February 2009, as he had recounted in his affidavit. He said that this was “the season of waves” and that every year the seaweed could be broken off for this reason. Mr Mboeik explained that when he had given his initial denial in cross-examination he was intending to refer to the fact that he had never had a problem with seaweed breaking off because of oil. The respondent submitted that Mr Mboeik’s initial denial in oral evidence was false and therefore demonstrated his unreliability as a witness.

202    Next, the respondent relied on Mr Mboeik’s denial in cross-examination that, between 2006 and September 2009, he had any issue with seaweed breaking off his ropes and washing away. In his affidavit he had said that around October and November 2007 some of his seaweed had broken off in windy conditions and washed away. When challenged in cross-examination, Mr Mboeik accepted that this had happened. Once again, the respondent submitted that Mr Mboeik’s initial denial in cross-examination about seaweed breaking off and washing away between 2006 and September 2009 was an indicator of the unreliability of his evidence.

203    In his cross-examination, Mr Mboeik also said that in October and November 2007 the tips of some of his seaweed had become white. He was also picked up on this, but he explained that he was still able to sell his seaweed, “so it was equivalent to no problem”.

204    It is tolerably clear that when Mr Mboeik was giving his answers in cross-examination which the respondent said were “false” (and which Mr Mboeik denied were false), his focus was on the major problem he experienced in September 2009, which was that his seaweed had died and washed away. He was not considering what might be described as relatively insignificant day-to-day operational losses in running his seaweed farm which did not impact on him selling his seaweed. When Mr Mboeik’s oral evidence is considered in context, including with his affidavit evidence, I do not consider it to be unreliable. I accept Mr Mboeik’s account of what he saw in the water near his crops in September 2009.

205    The respondent submitted that the evidence given by Mr Gustaf Lay should be treated as unreliable. Mr Lay is a seaweed farmer and buyer from Tablolong village. His evidence was that early one morning in late September 2009 he observed that the water where his seaweed farm was located had changed colour. It was shiny and looked like a rainbow. He saw blocks that were coloured like chocolate and blocks that were yellowish and greyish. These blocks were the size of his fist. They resembled the texture of a candle and were oil. After touching them his skin felt itchy. He saw dead fish and other dead marine life in the water. His seaweed became soft. It did not recover and he was unable to harvest it.

206    The respondent submitted that Mr Lay’s evidence was unreliable because it was internally inconsistent and not supported by a file note taken by the applicant’s lawyers on 28 October 2014. The file note recorded a meeting at Tablolong which Mr Lay attended. In relation to Mr Lay, the note refers to him expressing his thanks for the visit and for “caring for life as farmers”. The file note seems to mention the Montara oil spill, and then proceeds with a number of dot points, including one which states: “Since spill from oil to now”. It is not apparent what that sentence was intended to convey.

207    In a later part of the file note, the question is posed: “When was first problem in 2009?” This is followed by:

March 2009/May2009. Isis.

I take the reference toIsis to mean so-called ice-ice disease.

208    Mr Lay explained that this was not a problem as such. It was more a situation that was anticipated. In Mr Lay’s experience, seaweed crops were prone to ice-ice disease at this time of the year. He said that March was the month for seaweed planting and that if the crops were not controlled—particularly should the ropes begin to sink—ice-ice disease could occur in April, when the season changes. He said, however, that he anticipates the problem by harvesting. If white spots (he said “dots”) begin to develop on the stems of the seaweed, Mr Lay said that he immediately picks it. Mr Lay accepted that ice-ice could develop on the tips of the seaweed if exposed to the sun. He also said that he had been told by other farmers that ice-ice disease could develop if the water is too warm. He said, however, that he had not experienced this problem himself. He said that, in his experience, the temperature of the seawater at Tablolong “has always been normal”.

209    The respondent submitted that Mr Lay’s evidence about there not having been a problem with ice-ice disease in March/May 2009 was false because the file note had, in fact, made a reference to it. I do not accept that submission. The file note makes a reference to ice-ice disease but is uninformative on this topic. The note that was made is not inconsistent with the explanation given by Mr Lay that ice-ice disease at this time of the year was a problem to be managed and that, to the extent that it was a problem, it was, in substance, an operational one that was posed each year, not just in 2009. The fact that the file note refers to it in response to a directed question for 2009 does not necessarily mean that it was a problem of any particular significance for that year. I accept Mr Lay’s explanation. He said that ice-ice was not a severe problem in April 2009 and not a problem for him in March or May 2009 or, so far as he was aware, for other seaweed farmers in Tablolong.

210    Next, the respondent submitted that it was significant that the file note did not record Mr Lay observing oil in the seawater around Tablolong in 2009. This is not entirely correct. The file note refers to the oil spill, indicating that this was part of the conversation in which Mr Lay participated or at least the context in which the conversation occurred. In other words, the meeting proceeded on the basis that the seaweed had, in fact, been damaged by the oil. It is true that the file note does not record the particular observations of oil which Mr Lay gave in evidence, but there is nothing in the file note to suggest that Mr Lay was, at this time, asked to give a detailed account of what he had seen in the water at Tablolong in late September 2009. I do not accept that Mr Lay would necessarily have volunteered such a description without being asked to give one. Therefore, I do not attach much significance to the absence of any such description in the file note.

211    The file note does record that the “white colour” (presumably a reference to the seaweed turning white) was first seen in “Sept/Oct 2009”. Mr Lay said that he informed Mr Phelps (the author of the note) about this. In cross-examination, Mr Lay said that the tips of his seaweed went white and the seaweed went “limp”, which he said “started after the oil spill” and “destroyed all the seaweeds”. He distinguished this from ice-ice disease which, in his view, was not a disease that attacked the tips of seaweed, but, firstly, the stems of the seaweed and then the whole plant.

212    The file note also records:

What do they know about oil spill?

In the beginning they did not know. Thought it was disease.

Any knowledge of why oil spill occurred? They did not know.

Don’t know of Commission of Inquiry

213    In cross-examination Mr Lay said that, in the beginning, he did not know where the oil had come from. He initially denied telling Mr Phelps that, in the beginning, he thought that the seaweed had been affected by a disease, but later accepted that it was possible that he had said that, because “it has been quite a long time”. Nevertheless, Mr Lay denied that his seaweed crop had been affected by a disease. He said:

Once again I have to say I know the disease of seaweeds. I know the disease of ice-ice and it was not an ice-ice disease. I would term it a disaster. ...

214    I accept that it is possible Mr Lay might have told Mr Phelps that in the beginning—meaning, when the tips of the seaweed started to turn white and the seaweed became limp in late September 2009—he harboured the thought that his seaweed might have been affected by a disease, albeit not ice-ice. However, I note that Mr Lay was not the only participant from Tablolong at this meeting. The other participants included Mr Jackarius (the head of Tablolong village in 2014), Mr Zakarius Doroh (presently the head of Tablolong village), and Mr Mester Eryon Bessie (the secretary of the village). It is possible that one of the other participants conveyed the initial thought of disease to Mr Phelps, who recorded it. I infer that if, at any time, Mr Lay did harbour that possibility, the thought dissipated quickly. Mr Lay said:

... on the first day I inspected that that situation had changed, a lot of oil there. On the second day I still had hope that everything could recover. On the third day the same. On the fourth day it was a Sunday, we could not work. So that the plan was for the following day I would pick up the harvest. I took a boat for the purpose of harvesting, but the crops were all destroyed.

215    Mr Lay was directly challenged on the fact that he had seen oil near his seaweed in September or October 2009. He rejected the assertion that he had not seen oil.

216    I accept Mr Lay as an honest witness whose evidence is generally reliable. I accept his account of what he observed in the water in and around his seaweed crop in late September 2009.

217    Notwithstanding the respondent’s various criticisms of it, I consider the lay evidence of the observations that were made to be reliable. On the whole of that evidence, I am left in no doubt that, at the time, all witnesses (seaweed farmers, village heads, and other lay witnesses) witnessed a single, strikingly unusual, and unique event in the Rote/Kupang region, which coincided with the quick and dramatic loss of local seaweed crops. I am satisfied that this event was so striking that it is likely that it was fixed in their minds, notwithstanding that it was an event that, not unnaturally, was the topic of conversation between them, perhaps on many occasions, in the following years.

218    The respondent submitted that many of the lay witnesses’ observations were inconsistent with the substances they observed being oil from the H1 Well blowout. The respondent sought to support this submission by the expert evidence that was given with respect to the weathering of Montara oil and the expert evidence that specifically commented on the lay witnesses’ observations. I will discuss these strands of evidence in a later section of these reasons. However, it is convenient to record now that, in closing submissions, after contending that weathered Montara oil would not have looked or smelled as the witnesses had stated, the respondent advanced the following propositions: (a) the expert evidence shows that the substances observed by the lay witnesses cannot have been Montara oil; (b) therefore, some other event or phenomenon caused those substances to reach Rote/Kupang in large volumes, if the lay witnesses’ evidence is reliable; (c) the state of weathered Montara oil was “otherwise indicative of the unreliability of the lay witness testimony”.

219    It is not clear to me how these propositions stand together. The first two propositions proceed on the basis that the lay witnesses did not observe Montara oil but the widespread arrival of another substance or other substances due to some other event or phenomenon. The third proposition appears to be that the lay witnesses’ observations were unreliable because the witnesses did not reliably describe weathered Montara oil. However, the presence of Montara oil in these locations is the very proposition that the respondent denies.

220    The respondent also contended that a “striking feature” of the applicant’s case is that oil from the H1 Well blowout reached not only the southern coast of Rote, but also the northern and western coasts of Rote, and Kupang. The respondent pointed to the fact that none of the modelling predicted those outcomes. I will deal with this submission in a later section of these reasons.

221    In closing submissions, the applicant drew attention to the evidence of some of the lay observers who were not seaweed farmers or village heads. Even though the gist of their evidence is included in Schedule C, I will now recount their evidence in a little more detail.

222    Matthew Smith was an aerial observer who was deployed by the Australian Marine Oil Spill Centre (AMOSC) to assist AMSA during the oil spill response effort. Mr Smith swore an affidavit in which he deposed to his observations of the oil during that period. The affidavit included photographs, mud maps and Surveillance Flight Reports related to observations of the spill.

223    Mr Smith undertook daily aerial observations for AMOSC in September and October 2009 from a Dornier search and rescue aeroplane. On each sortie he undertook various activities, including directing vessels which were part of the oil spill response containment and recovery operations to significant patches of oil; performing surveillance on nearby reef systems; and identifying the extremity of the oil and sheen.

224    Over the reefs near the blowout, Mr Smith observed sheen, but no thick, heavy oil. Identifying the extremity of the oil was a difficult task as the oil and sheen did not form a clear unbroken line on the sea surface. He said it was also difficult to identify the edge of the oil and sheen as there was a vast area of the Timor Sea to cover and conditions were variable. He was confident of his observations on some days and not on others, depending on the conditions. He also had no way of knowing how long the oil and sheen he observed had been on the sea surface.

225    Mr Smith said that the pilots of the Dornier aircraft had some flexibility regarding flight paths in Australian airspace but were not permitted to enter Indonesian airspace without prior approval. He did not say whether that approval was granted for any sorties he undertook.

226    To identify the edge of the oil and sheen, the plane tracked to its last known location before following a band of thicker oil as identified by Mr Smith. He said he could not exclude the possibility that oil and sheen had travelled beyond his line of sight and that he may have occasionally lost track of its edges. In closing submissions, the applicant submitted that it was therefore possible that the oil and sheen he observed extended further towards Rote and Kupang than was evident in his mud maps, Surveillance Flight Reports and other contemporaneous AMSA documents.

227    The applicant drew my attention to a number of sightings recorded in the AMSA Surveillance Flight Reports in September and October 2009 of oil or features consistent with oil north of Australia’s Exclusive Economic Zone at distances between 50 and 83 km from Rote. The applicant argued that it was plainly possible for that oil to have travelled north and reached the coastlines of Rote and Kupang.

228    Mr Smith was not cross-examined. However, in closing submissions, the respondent pointed to Mr Smith’s observations about the difficulty of identifying oil from the air in support of the contention that numerous phenomena can be mistaken for oil from above.

229    In a later section of these reasons I also refer to how oil spilled on the ocean surface can aggregate into filaments called Lagrangian Coherent Structures (colloquially, “tiger tails”). These structures imply that oil can travel via conduits, thereby potentially impeding its visual detection by, for example, aerial surveillance.

230    Dr Ghislaine Llewellyn, a marine program leader at the World Wide Fund for Nature Australia (the WWF), led an expedition by the WWF to the Timor Sea to document the consequences of the Montara oil spill on marine wildlife between 24 and 29 September 2009.

231    On behalf of the WWF, Dr Llewellyn commissioned Simon Mustoe, Director of Applied Ecology Solutions, to prepare an independent biodiversity survey of the area likely to be affected by the blowout. Mr Mustoe is an ecologist and made his own affidavit, referred to below.

232    The WWF chartered a boat for the expedition, leaving Darwin on 24 September 2009. Joining Dr Llewellyn and Mr Mustoe on the expedition was Kara Burns, a freelance photographer contracted by the WWF to take photographs; Deborah Glasgow, an expert in marine mammal observation and Chris Sanderson, an expert in bird observation and an employee of Applied Ecology Solutions. Lindsay Moller, a journalist and photographer from The Australian newspaper also joined, as well as a small crew.

233    Dr Llewellyn says Mr Mustoe compiled the data and observations from the expedition into a report titled “Biodiversity Survey of the Montara Field Oil Leak” dated 22 October 2009 (the WWF Report). The observations outlined in the WWF Report are recorded below. Dr Llewellyn says the locations identified and observations recorded in the WWF Report accord with her own recollections.

234    In particular, Dr Llewellyn says that on 27 September 2009, she smelled a foul, strong chemical smell and felt a slight burning sensation in the back of her throat. She assumed this was caused by the gases emitting from the wellhead platform and the boat changed course to evade the smell. Towards the end of the same day, Dr Llewellyn observed patches and windrows of oil on the surface of the water. The character of the oil was variable but she says they were in the boat for hours and there was a heavy blanket of oil on the surface of the water as far as the eye could see. The smell changed from earlier in the day and was more akin to the smell of a petrol station forecourt.

235    As Dr Llewellyn had not expected to see slicks or large amounts of oil, she had not made preparations to collect samples of oil. She therefore improvised with available materials and devised a system for the collection of samples. The coordinates where samples were collected was recorded by Mr Mustoe in the WWF Report. Samples were taken between 26 and 29 September 2009.

236    Dr Llewellyn’s second affidavit, sworn on 28 March 2019, included further photographs taken by Ms Burns as well as Mr Moller.

237    Simon Mustoe affirmed an affidavit on 10 August 2018 which included a copy of the WWF Report. The WWF Report sets out the locations travelled to and observations made during the field survey. The survey was carried out predominantly in an area to the northeast of the H1 Well, substantially within Australia’s Exclusive Economic Zone. The route is shown in Figure 6 of the WWF Report. Mr Mustoe deposed that he observed Ms Llewellyn collecting water and oil samples referred to in the WWF Report during the field survey.

238    On 25 September 2009, the WWF Report records that the observers noted some white specks in the water that they thought might have been broken up cuttlefish. However, the WWF Report notes that after the observers saw wax particles the following day it is possible, with hindsight, that the white specks were wax particles.

239    On the morning of 26 September 2009, the survey team crossed a patch of thick white snowflake-like material, which appeared to be a flocculating waxy compound that they presumed was residue from the oil spill. There was an obvious surface sheen layer associated with the wax particles. The survey team recorded surface sheen, and wax particle density and size, systematically throughout the day. In the morning, they mostly passed through areas of patchy light sheen with small wax particles of varying densities. At about midday, they crossed a dense waxy slick, and then a heavy algal bloom with some wax particles within it. In the afternoon they modified their course to head just north of the Jabiru drilling platform.

240    A record of the surface oil observed was kept from 26 to 29 September 2009. The expedition recorded the extent, weight, size and density of the oil. The results of this survey were summarised in Figure 24 and section 11.3.6 of the WWF Report. At times little or no surface oil was evident but at other times both surface sheen and heavier patches of oil were observed.

241    Notably, the WWF Report records that, on the afternoon of 27 September 2009, at about 40 nm from the H1 Well, the survey team observed a very heavy patch of surface oil, covered in a moderate to thick yellowish-brown layer. They observed rainbow patterns on the water and distinct trails of oil behind particles of a yellowish wax. Ripples on the water revealed a blackish streaked tinge to the waves. The area smelled strongly of turpentine.

242    On 28 September 2009, the survey began about 25 nm due east of the H1Well and headed northwest. Light oil sheen was evident in this area but there were not any particularly heavy patches of weathering oil. However the survey team observed long and broad slicks of surface sheen containing waxy particles of varying size and density. At dusk, they encountered the very thick area of oil slick observed the day before.

243    The survey team took photographs recording the behaviour of the oil observed on the sea surface. These are shown in Figure 22 to the WWF Report. The WWF Report observed that surface oil could be identified by extensive patches of continuous glassy water; particles of white waxy residue of varying sizes and densities; smell; or in moderately heavy patches, a clearly visible oil layer on waves or in the wake of the vessel. With regard to the particles of white waxy residue, the WWF Report observed that the larger of these could be seen to leave an oil trail on the surface.

244    With regard to the extent of the slick, the WWF Report concluded that there was extensive patchy surface sheen throughout most of the surveyed area, even in waters situated over 100 nm from the source of the spill. The furthest that surface sheen was found with any certainty was about 140 nm from the Montara H1 Well. It was assumed that this oil had originated from that source. The report concluded that there were areas where particles of white waxy residue of varying size and density were floating on the surface, and one particular area (referred to above) where the surface sheen was particularly thick and accompanied by slicks of yellowish wax particles.

245    Professor James Watson was commissioned by the Commonwealth Department of the Environment, Water, Heritage and the Arts (DEWHA) to lead an expedition to the Timor Sea to undertake a rapid survey of cetaceans, birds and marine reptiles (megafauna) in the Montara oil spill region. The purpose was to identify megafauna in the region and address the impact of the Montara oil spill on them.

246    Professor Watson engaged two colleagues from the University of Queensland to assist with the rapid survey, which was undertaken between 25 September 2009 and 4 October 2009. Professor Watson’s observations and findings were presented in a report titled, “A rapid assessment of the impacts of the Montara oil leak on birds, cetaceans and marine reptiles”, dated 23 October 2009 (Professor Watson’s Report).

247    The survey team conducted five days of transects at sea, incorporating 279 10 minute strip transects covering a distance of 668.5 nm and a total survey area of 99,040 ha. The area covered was directly north and northwest of the H1 Well extending to the Ashmore Reef as recorded at Figure 1 of Professor Watson’s Report.

248    In these surveys, a total of 124 10 minute strip transects were in waters visibly affected by oil, representing 44% of the total number of strip transects made. The oil was more prominent in transects directly north of the H1 Well, as shown in Figure 4 of Professor Watson’s Report.

249    Professor Watson’s Report does not describe the appearance of the oil in great detail, but referred to the variable coverage and thickness of oil on the water in different areas. Figure 7 of the report shows an example of a thick layer of oil on the surface of the water. It is a yellowish-brown colour and appears to be textured, with an area of sheen connected to it.

250    The report found there was a significant risk that a change in conditions could push the slick towards the breeding islands and reefs to the north, west and south of the H1 Well, or into deeper waters to the west and north of the Ashmore reef.

The volume of oil spilled

Introduction

251    There was a considerable body of evidence directed to estimating the volume of oil discharged from the H1 Well over the 75 day period of the oil spill.

252    On 21 August 2009, the respondent informed AMSA that the volume of oil being spilled may have been between 200 to 400 bbl/day. The respondent has not provided evidence in this proceeding to support that rate of release, but it did, however, provide the rate of 400 bbl/day as an assumption to be used by Dr French-McCay in her trajectory modelling: see below. Ultimately, the volume of oil released was of no consequence to Dr French-McCay’s modelling because, in her opinion, the trajectory of the released oil would not change; only its concentration would change at the locations which, on her modelling, the oil reached. The position was otherwise with Dr Hubbert’s modelling. Dr Hubbert disagreed with the notion that the volume of released oil did not affect its trajectory over time. He said that notion failed a “common sense” test.

253    As I explain below, in the trajectory models used by Dr French-McCay and Dr Hubbert, oil is represented as collections of particles or “spillets”. In these models, the particles are released intermittently to represent the continuous flow of oil into the ocean as it is affected by winds, ocean currents and turbulence. Dr Hubbert noted that Dr French-McCay’s model limited the number of particles that could be released—specifically, no more than five particles each 30 minutes. An increase in oil volume did not result in more particles being released. Rather, in Dr French-McCay’s model, a greater volume of oil was assigned to each particle. Therefore, the five particles ended up going to the same locations regardless of the volume of oil released. Dr Hubbert said that the model he used responded to an increase in the volume of oil by releasing a proportionally greater number of spillets every two minutes.

254    The applicant’s case is that a far greater volume of oil than 200 to 400 bbl/day was released during the spill. He sought to prove this in two ways. The first way was through Professor Wereley’s analysis and calculation of volume flow, based on photographic observations of oil discharging from a horizontal drain pipe on the wellhead platform during the course of the spill. The second way was through calculations performed by Professor Towler based on a “material balance” assessment of the H1 Well.

Oil volume flow rate calculation

Professor Wereley’s analysis and calculations

255    The blowout of the H1 Well caused liquids and gases (flux) from the Montara reservoir to flow out of the top of the wellhead. The flux hit the bottom of the drilling floor. Some portion of the flux continued upwards into the drilling mezzanine while another portion of it dropped back onto a helipad. The flux in the drilling mezzanine followed a path over to the West Atlas rig and then to the sea via a vertical drain pipe. The flux on the helipad followed a path through the wellhead platform to the sea via a horizontal drain pipe. Photographs taken at the time show other avenues for flux entering the sea. However, for his analysis, Professor Wereley relied on his observation of oil discharging from the horizontal drain pipe. It will be appreciated, therefore, that Professor Wereley’s analysis and calculations were somewhat conservative at the outset in that they did not purport to quantify all the oil that entered the sea as a result of the spill.

256    The helipad drain system was a long, circuitous pipe network that led from several drains on the helipad deck to the horizontal drain pipe that discharged to the sea. The network was comprised nearly entirely of straight runs of 200 mm (8”) nominal size pipe connected by many 90° elbows, and a few valves. The horizontal drain pipe was the bottommost pipe in the network. It was approximately 13 m long and terminated with a 45° (from the horizontal) angled section.

257    The flux can be considered as comprising three components: condensate (i.e. oil), gaseous hydrocarbons, and water. Professor Wereley’s evidence was that the gaseous hydrocarbons would have been carried away by the air. He said that the flow would have contained no water, having regard to the temperature at which, and force with which, water would have exited the wellhead. As Professor Wereley explained in his first report:

5.5    Because of the vigour with which the flux out of the well impinged on the drilling floor, the liquid was broken up into a fine spray comprised of many droplets. During this time, the heat of the liquid, already near the boiling point, combined with the agitation of the droplets provided ideal conditions for those droplets to evaporate. To the extent that the water, which flowed from the reservoir, was not already steam, the conditions upon its exit from the well head were such that it would have evaporated.

258    Professor Wereley’s examination of the available photographs of the wellhead platform and the West Atlas rig taken at the time of the spill confirmed (for him) that the flow, at least from the horizontal drain pipe, contained materially no water.

259    There are several ways of calculating the amount of oil that spills into the sea from an uncontrolled well. One of these is by visual observations taken from photographs to calculate the velocity of the liquid. The calculation of velocity is then converted to a volume flow rate. This method of calculating the velocity of a liquid from its trajectory is an accepted approach in fluid mechanics. As Professor Wereley explained:

6.1(e)    A single photograph of oil spilling from a broken pipeline can be used to estimate the amount of oil being spilled by that pipeline. For example, oil flowing from inside a pipeline out through a weld failure in the pipe arches high into the air creating a curved path through the air due to gravity’s effect pulling it back to the earth. From the path of the oil recorded in a single image, not a video, it is possible to calculate the speed of oil as it emerged from the pipe. When the speed is combined with the size of the weld failure, the amount of oil being spilled can be calculated.

260    This is the method that Professor Wereley used. He described the principle for calculating oil flow out of the horizontal drain pipe as follows (omitting footnotes):

8.2    The oil flows out of the mouth of the horizontal drain pipe. After it leaves the horizontal drain pipe, two forces act on it. The first of these is gravity which pulls the oil downward at 32.174 ft/s2. The second of these is wind which pushes the falling oil in the downwind direction.

8.3    The curve that the oil follows as it bends toward the sea below can be described mathematically. That description can then be used to extract the oil speed as it exits the pipe, as well as the wind direction.

8.4    Qualitatively the oil’s path can be illustrated by imagining the upper and lower limits of the oil speed as shown in Figure 8. If the oil drips out of the end of the horizontal drain pipe it will fall straight downwards and hit the sea directly underneath the drain pipe. If the oil is moving very fast (as out of a fire hose), it will travel straight along the direction of the end of the pipe (45 degrees downward from the horizontal) and follow a nearly straight line, ultimately hitting the sea. The actual oil speed in this oil spill is somewhere between these two extremes. Once the oil speed is known, the oil volume flow rate can be calculated by multiplying by the area of the pipe that is filled with the oil.

                

    Figure 8. The diagram shows the two limits on the path that oil can take as it falls downward to the sea.

261    To perform his initial calculations, Professor Wereley used two photographs taken on 21 August 2009 (designated as PTT.617.003.9947 and PTT.617.003.9939); one photograph taken on 22 August 2009 (designated as PTT.620.003.8765); one photograph taken on 26 August 2009 (designated as PTT.600.026.5696); one photograph taken on 9 September 2009 (designated as PTT.620.003.8819); and one photograph taken on 3 November 2009 (designated as PTT.620.004.0916). Thus, the oil flow from the horizontal drain pipe was measured on five days. Eight separate measurements were made. Professor Wereley summarised the results of his analysis and calculations in a table (Table 2):

Table 2. Summary of measurements

Date

Image Name

Oil Flow Rate (Bbl/day)

21 August 2009

PTT.617.003.9947

1250

21 August 2009

PTT.617.003.9947

362

21 August 2009

PTT.617.003.9939

1582

21 August 2009

PTT.617.003.9939

668

22 August 2009

PTT.620.003.8765

417

26 August 2009

PTT.600.026.5696

1170

9 September 2009

PTT.620.003.8819

715

3 November 2009

PTT.620.004.0916

464

262    Professor Wereley calculated the cumulative volume of oil spilled as follows:

17.2    Given the quantitative measurements and the qualitative observations, it seems clear that the oil continued flowing throughout the course of the oil spill. Based on the materials which I have been asked to consider, I am unable to observe any quantitative trend regarding the oil volume flow rate. For example, the volume flow rate of the oil exiting the horizontal drain pipe measured on 9 September 2009, was both considerably higher and lower than measurements for volume flow rates in August 2009. Consequently, it is my opinion that the best way to estimate the oil flow rate over the course of the spill is to average the first seven numbers in Table 2. These seven measurements are representative of the oil volume flow rate before the well head platform caught fire. I have excluded the measurement for 3 November 2009 as it is only representative of the short period of time that the well head platform was on fire. It is clear that less oil was spilled into the ocean when the well head platform was on fire as much of the oil burned and did not drain into the sea. The best estimate of the oil volume flow rate between 21 August 2009 and 31 October 2009 is:

Qavg = (1250+362+1582+668+417+1170+715)/7 = 881 bbl/day

17.3    The total volume of oil spilled is then the sum of the average daily oil volume flow rate times the number of days before the well head platform caught fire and the oil volume flow rate after the well head platform caught fire times the number of days before the well was killed:

Total volume = 72 days x 881 bbl/day + 3 days x 464 bbl/day = 64,824 bbl

(Emphasis in original.)

Criticism of Professor Wereley’s analysis and calculations

263    The respondent called Dr Zaldivar to prepare a report responding to Professor Wereley’s first report. In his report, Dr Zaldivar said that there were many errors, assumptions and simplifications in the methodology that Professor Wereley had used to calculate flow rate. He identified four matters which he saw as “key deficiencies” which, in his opinion, rendered Professor Wereley’s conclusions “completely unreliable and invalid”.

264    First, Dr Zaldivar said that Professor Wereley’s treatment of the effect of wind on the oil draining from the horizontal drain pipe was incorrect because Professor Wereley did not correctly account for the horizontal and vertical components of wind drag on fluid velocity. Dr Zaldivar argued that, when the correct equations are taken into consideration, it is not possible to calculate a flow rate using Professor Wereley’s approach. This is because there were too many unknown variables. Dr Zaldivar argued that, because the formulation used by Professor Wereley was incorrect, all the calculated flow rate results should be “considered unreliable and invalid”. Dr Zaldivar also pointed out that Professor Wereley had adopted inconsistent assumptions: Professor Wereley had argued that wind friction could be ignored but, in his calculations, he had added a horizontal component of wind speed to the fluid velocity where the fluid exits the horizontal drain pipe.

265    Secondly, Dr Zaldivar identified a unit conversion error which, he said, caused Professor Wereley’s calculations to be “fundamentally flawed”. Dr Zaldivar contended that, when this error was corrected, the results of Professor Wereley’s calculations were “completely out of the range of reality” and “nonsensical”. Dr Zaldivar argued that this underscored (what he saw as) the “fundamental unreliability” of Professor Wereley’s method.

266    Thirdly, Dr Zaldivar said that Professor Wereley had incorrectly assumed that the flow exited the horizontal drain pipe at a 45° angle (the exit angle of the pipe bend at the distal end of the drain pipe). In Dr Zaldivar’s opinion, the length of the pipe bend was not sufficiently long to cause the entire flow to deviate along the bend, especially at higher velocities. Dr Zaldivar replicated Professor Wereley’s calculations but changed the exit angle to 30°. This showed that Professor Wereley’s fluid velocity was nearly 86% greater for an exit angle that was only 50% greater than the angle which Dr Zaldivar used. According to Dr Zaldivar, this demonstrated the sensitivity of Professor Wereley’s calculations to the exit angle selected.

267    Fourthly, Dr Zaldivar said that Professor Wereley’s fraction filled calculation (i.e., how much of the pipe was filled with oil) was incorrect due to several factors, namely:

(a)    Professor Wereley had treated the fraction filled as 100% in cases where, in Dr Zaldivar’s view, there was insufficient photographic information available to make that determination (Dr Zaldivar said that Professor Wereley’s method of estimating the apparent fraction filled was “just a guess”);

(b)    the photographs that Professor Wereley had used showed only the “apparent fraction filled” (meaning that, as fluid exiting the drain pipe does not have enough time to bend completely along the pipe bend, the apparent fraction filled as seen at the exit of the pipe bend will always be higher than the actual fraction filled in the horizontal portion of the pipe: in other words, there is an appearance that the pipe is fully filled when, in fact, it is not); and

(c)    Professor Wereley had assumed that there was no water and no entrained gas in the liquid phase; any water would reduce the calculated volume of the oil that was discharged and entrained gas flowing in or with the oil would reduce the oil flow rate.

268    Dr Zaldivar advanced a number of other criticisms. He criticised Professor Wereley’s use of three of the photographs to carry out his calculations. In essence, Dr Zaldivar argued that the photographs were unsuitable, in various ways (including their low resolution), for the way in which Professor Wereley used them. With respect to two of the other photographs, Dr Zaldivar criticised the way in which Professor Wereley used them.

269    Dr Zaldivar criticised Professor Wereley’s analysis and calculations. In particular, he criticised the way in which Professor Wereley arrived at his calculation of cumulative volume. Dr Zaldivar argued that there was a clear trend of decreasing flow rate over time. He said that Professor Wereley’s estimated cumulative volume of oil discharged (64,824 bbl) was “between three and four times too high just based on the way [Professor Wereley] performs the cumulative calculation”.

270    Dr Zaldivar also argued that Professor Wereley had used an incorrect measurement for the internal diameter of the horizontal drain pipe (Professor Wereley used its nominal measurement of 8” whereas Dr Zaldivar said the internal diameter of the pipe was 7.813”). Dr Zaldivar said that Professor Wereley’s incorrect pipe diameter accounted for an over-estimate of the calculated flow rates by 4.6%.

271    Dr Zaldivar also argued that Professor Wereley’s analysis of the photographs had not accounted for perspective effects due to the wind velocity.

272    Further, Dr Zaldivar argued that Professor Wereley had failed to determine the uncertainty of the estimates he had used in performing his calculations. Dr Zaldivar said that if Professor Wereley had performed a scientific analysis of the uncertainty in this approach, he would have found that some of the inputs he used greatly affected the flow rates calculated using his methodology.

273    Dr Zaldivar also introduced the possible confounding effect of the activation of the fire systems on the platform and drilling rig at the time of the spill. He argued that if the fire systems were active, then water from those systems would have mixed with oil in the horizontal drain. If so, Professor Wereley would have overstated the observed flow from the horizontal drain because that flow would have included a potentially large amount of seawater.

274    It is convenient at this point to record that there is no direct evidence before me that the fire systems had, in fact, been activated at the time the photographs were taken. It is only Dr Zaldivar’s speculation that such systems might have been activated. He sought to give life to this speculation by referring to one photograph showing a difference in the colour of the flow from the vertical drain pipe on the West Atlas rig (a lighter colour) compared to the flow from the horizontal drain pipe on the wellhead platform (a darker colour). I am not prepared to draw any conclusion from this colour difference. It is very indirect evidence and, on the other evidence before me, the colour difference could be attributable to any one or more of a number of different factors. I am not satisfied on the evidence that the fire suppression systems had been activated in the photographs on which Professor Wereley relied. I would add that, had the fire suppression systems been activated, then there seems to have been no reason why the respondent could not have called direct evidence of that very matter.

275    Dr Zaldivar expressed his overall conclusions as follows:

5.1 Dr. Wereley used a method of estimating fluid velocity by applying a mathematical formulation that describes the trajectory of the fluid exiting the drain pipe and calculating a fluid velocity that predicted a trajectory that matches the observed trajectory in photographic evidence. In the process, he overly simplified or made unjustified assumptions about several factors:

1)    He assumed that the exit angle would always be 45° downward to the horizontal, which is incorrect based on the short length of pipe bend at the end of the horizontal drain pipe.

2)     He assumed that the apparent fraction filled in the photographs is the fraction filled in the horizontal pipe without regard for the fact that fluid at the edge of the horizontal pipe had to travel a mere 11.2 inches before it struck the top surface of the pipe bend, which would result in an apparent fraction filled of 100%. He also ignores the perspective effects on apparent fraction filled.

3)     He simultaneously assumed that air offers both zero resistance by ignoring the effect of air in calculating the trajectory, and infinite resistance by adding the horizontal component of wind velocity to the fluid exit velocity immediately upon its exit. He also assumed that wind velocity had no vertical component, and that the off-plane component of wind velocity had no apparent effect on the observed trajectory.

4)     He assumed that there is no water in the fluid exiting the horizontal drain pipe while providing no scientific basis for his assumption.

5)     He assumed that there is no gas entrainment in the liquid, while multiphase correlations show that at high liquid velocities, gas would certainly be entrained as dispersed bubbles in the liquid phase. He makes no effort to address this on a scientific basis.

6)     His calculation of cumulative volume discharged heavily weights the flow rates during the first five days of the spill.

7)     He made several mistakes in his report and did not consistently report all the calculated parameters, e.g., wind velocity is only reported in the first calculation and completely ignored in the rest of the calculations.

8)     He made no effort to extrapolate and compare the trajectory, which is calculated based on very few points near the exit of the drain pipe, to the longer trajectories visible in the photographs.

5.2     Beyond these mistakes and deficiencies in the report, the most important mistake is his conversion factor from ft3/s to bbl/d. Fixing this mistake puts his results in a range of flow rates that are wildly inconsistent, by an order of magnitude, with the reservoir simulations that were used in support of the well kill effort. The fact that the well kill effort was successfully performed lends credence to the reservoir simulations, based on which the oil flow potential of the well was estimated to be around 1500 bbl/d.

5.3     Based on all these factors, my opinion is that Dr. Wereley’s report cannot be relied upon to obtain a scientifically sound estimate of the total volume discharged during the incident. Even in the event one decides to accept his methodology to calculate flowrate, I found several omissions and mistakes that resulted in an overstatement of the calculated flowrates.

Professor Wereley’s response

276    Professor Wereley took on board Dr Zaldivar’s criticisms. He prepared a second report.

277    Professor Wereley accepted that there was inconsistency in assuming that the friction of the oil with the surrounding air was unimportant while at the same time assuming that the oil jet assumed the speed of the wind surrounding it. Professor Wereley accepted that the inconsistency needed to be resolved. He held to the assumption that the oil jet is not slowed considerably by the air because, at the exit of the horizontal drain pipe, the air only had a short time to act on the jet. However, Professor Wereley removed from his calculations the assumption that the oil immediately matched the speed of the air around it. In other words, he proceeded on the basis that the oil stream does not respond to the air drag near the horizontal drainpipe exit. As Professor Wereley put it, this approach is equivalent to saying that the momentum of the oil jet is large compared to the drag acting on it near the pipe exit. Professor Wereley said that this was a conservative assumption because neglecting the effect of drag on the jet (however small) would lead to a smaller jet speed in his calculations.

278    Professor Wereley accepted that he had made a unit conversion error in his calculations.

279    Professor Wereley accepted that the flow rate is dependent on the angle of the oil jet as it leaves the horizontal drain pipe. He therefore modified his jet flow calculations by extracting the exit angle from a “best fit” model of the jet motion, rather than assuming it from the geometry of the horizontal drain pipe.

280    With respect to the calculation of the fraction filled, Professor Wereley rejected the criticism that his methodology was “just a guess”. He said that his analysis of the photographs was based on commonly used methodologies in the open channel flow field (the study of flows in partially filled conduits, pipes and the like). As to Dr Zaldivar’s criticism that there was insufficient spatial resolution to see the jet height, Professor Wereley used a method to determine the fraction filled which involved deriving size ratios of the height of the jet exiting the horizontal drain pipe in images with better resolution, where the jet can be seen exiting the pipe and flowing at a more or less steady flow rate as it drops to the sea, and applying those ratios to the affected, lower resolution images.

281    Professor Wereley maintained his view that there was an insignificant amount of water in the discharge. Nevertheless, to compensate for uncertainties, he calculated what the flow rate would be if the flow contained 10% water. The figure of 10% appears to be based on a statement in the respondent’s Dynamic Kill Simulations and Evaluations report (relied on by the respondent to calculate the parameters for the operation to “kill” the well in order to stop the spill) that said that the condensate contained “less than 10% water cut”.

282    Professor Wereley agreed that air could be mixed with the oil in the drain system in such a way that the two would not separate as quickly as he had originally assumed. However, he considered that Dr Zaldivar had over-predicted the proportion of gas that might be entrained. Professor Wereley accepted that the flow out of the horizontal drain pipe is pulsatile. It was sometimes fast, and sometimes slow, which supported the view that the oil was mixed with air. As I have noted, the drain system is comprised of numerous straight sections of pipe with many 90° elbows and other fittings along its path. Professor Wereley explained that at each of these elbows or other fittings, the turbulent flow in the drain system thoroughly mixes the air carried along by the oil flow with the oil itself. Many small bubbles would be generated by repeated agitation, creating a mixture that would not be quickly separated. As a result, Professor Wereley modified his jet flow calculation by assuming that the liquid flowing out of the horizontal drain pipe is carrying air bubbles mixed in it.

283    With respect to Dr Zaldivar’s contention that, in calculating cumulative volume, Professor Wereley should have adopted a model in which oil flows slow with time, Professor Wereley observed that, while that may be the normal behaviour of an oil well, the H1 Well was not performing normally. He suggested, for example, that it was possible that the hole in the cement shoe would have eroded over time such as to cause an increase in flow. He said that, in order to make a fact and observation-based measurement, he had only used the data that he, himself, could measure. He did not use an outside model of the expected reservoir behaviour. However, as each instantaneous spill flow rate measurement was representative of the spill on a different date, and those dates were arbitrarily spaced through the duration of the spill, Professor Wereley re-calculated the spill volume using a method common in engineering called the trapezoidal rule, rather than relying on a simple average of all measurements he made. As Professor Wereley explained it, by applying the trapezoidal rule, instantaneous flow measurements (bbl/day) are turned into a total spill volume (bbl).

284    Professor Wereley recognised that there was uncertainty in each of the main components of his measurements, including: the scale factor; the speed of the flow calculation; the fraction filled calculation; the volume fraction of gas carried by the flow; and the fraction of the condensate that is water. He reasoned that the uncertainties in the scale factor and flow speed calculations could be ignored in the presence of the much larger uncertainties in other quantities.

285    In relation to the possible activation of the fire systems, Professor Wereley noted that one of the assumptions with which he had been provided was that all electricity to the West Atlas rig was turned off by 9.00 pm on Friday, 21 August 2009. If so, any fire suppression systems that may have been operating would have ceased by that time and any fire suppression fluids would have ceased running off the rig at that time or shortly thereafter. Two flow measurements that Professor Wereley calculated were based on photographs taken on 21 August 2009 during the day. He acknowledged that, consequently, it is possible that some part of the calculated flow was water from the fire suppression system on that day. However, as I have said, I am not satisfied on the evidence that the fire suppression systems had been activated in the photographs on which Professor Wereley relied.

286    Professor Wereley then re-computed his flow rate calculations and presented his revised results in his second report as follows:

15.1    My calculations have been revised incorporating the following factors that were discussed in detail above:

(a)    The correct inner pipe diameter of 7.813 inches.

(b)    The drag of the surrounding air on the jet just after it leaves the HDP is neglected.

(c)    The units are properly converted from ft3/sec to bbl/day.

(d)    The downward angle of the oil jet emerging from the HDP is extracted from the jet trajectory rather than imposing a 45 degree angle.

(e)    Fraction filled is treated differently in cases where the spatial resolution of the photo is too low to directly and accurately judge the fraction filled (i.e. PTT.600.026.5696) or cases where there is a fast and slow flow in the same photo and only one of those is emerging from the HDP exit in the photo (PTT.617.003.9947 and PTT.617.003.9939).

(f)    The model accounting for the air mixed in and carried along with the oil flow as bubbles using Gomez’s (2000) equation.

(g)    The calculation of total spill volume has been made by integrating the measurements made using the trapezoidal rule.

15.2     In order to calculate the total amount of oil spilled into the environment, the instantaneous oil spill rates (i.e. bbl/day) must be integrated into a single number. In order to bound the maximum and minimum possible spill sizes, the uncertainties in the fraction filled, volume fraction of gases, and water content have been considered as shown in Table 2 below.

Table 2. Oil Spill Flow Rates

Date

Image Name

Day #

Exp Oil Flow Rate (bbl/day)

Max Oil Flow Rate (bbl/day)

Min Oil Flow Rate (bbl/day)

21 Aug

PPT.617.003.9947

1

19100

21000

13300

0

0

0

21 Aug

PPT.617.003.9939

1

18900

20800

13200

9400

10300

6500

22 Aug

PPT.620.003.8765

2

8300

9100

5800

26 Aug

PPT.600.026.5696

6

10500

11600

7300

9 Sept

PPT.620.003.8819

20

12300

13500

8600

3 Nov

PPT.620.004.0916

75

3600

4000

2500

287    After applying the trapezoidal rule, Professor Wereley calculated that the total oil spill volume over the period of the spill was between 452,000 and 714,000 bbl. This is approximately 6,000 to 9,000 bbl/day.

The Joint Report on Volume

288    The presentation of Professor Wereley’s revised calculations did not end the disagreements between the two experts on this question or bring them any closer. In fact, the area of disagreement was widened by some of the accommodations that Professor Wereley had made to meet Dr Zaldivar’s criticisms. For example, a new area of disagreement was the way in which Professor Wereley had analysed and digitised the trajectory of the oil jet from the horizontal drain pipe in his second report.

289    The various disagreements carried on into a Joint Report prepared by the four experts on spill volume (the Joint Report on Volume), and into the concurrent evidence session. Dr Zaldivar maintained that, notwithstanding the revisions which Professor Wereley made to accommodate his (Dr Zaldivar’s) criticisms, there were many errors, assumptions and simplifications in Professor Wereley’s methodology that rendered his conclusions “completely unreliable and invalid”. He also argued that the uncertainties associated with the inputs that Professor Wereley used, and their impact on the uncertainty of the calculated flow rate, had been “grossly understated” by Professor Wereley and provided “false confidence in the validity of the total volume estimate”.

290    No further light will be thrown on the differences between Professor Wereley and Dr Zaldivar by further summarising that debate in these reasons. The following matters are, however, important.

291    First, Dr Zaldivar’s instructions from the respondent were to prepare a report responding to Professor Wereley’s first report. Dr Zaldivar steadfastly interpreted these instructions as meaning that he was not to provide an estimate of spill volume and that he was confined to critiquing Professor Wereley’s methodology and calculations. Indeed, in the Joint Report on Volume he specifically identified this as his role. It was a role he performed with some vigour.

292    Secondly, Professor Wereley’s task was to find the best method possible, in the circumstances of the available evidence (which, admittedly, was limited), to calculate spill volume. As Professor Wereley himself explained, there is no “ground truth” in his calculations. Rather, they stand as one analysis through which the Court can gain an appreciation of the likely size of the spill (in terms of volume and rate of discharge) from objective evidence rather than unsupported assertion. Three experts provided that assistance in terms of using their expertise to arrive at reasoned calculations. Dr Zaldivar did not. His role was a more limited and directed one. Even then, the main thrust of his criticisms was to raise the prospect of error, without demonstrating that there was error, or error that would lead to an appreciably different outcome in actual result. Thus, his contentions that Professor Wereley’s calculations were “unreliable” or “invalid” reside in, and do not transcend, the realm of argument and debate.

293    The notable exceptions to this were Dr Zaldivar’s identification of Professor Wereley’s conversion error and Professor Wereley’s inconsistent inclusion of a wind drag factor in the equations he used. However, when these errors were pointed out, Professor Wereley quite properly accepted them as such, and made corrections to accommodate for them. Dr Zaldivar did perform a calculation—based on what he regarded to be Professor Wereley’s flawed methodology—in which he integrated the area under the curve to obtain a cumulative discharge volume. By this calculation, Dr Zaldivar arrived at a total volume spill of 16,178 to 21,181 bbl. However, in keeping with the limitations he placed on the instructions he had been given, Dr Zaldivar did not undertake his own analysis of the volume of oil spilled, or provide his own calculations based on that analysis. Thus, Dr Zaldivar did not present an alternative estimate of the amount of the oil spilled which could be exposed to evaluation and comment by his fellow experts—in particular, which could be set against Professor Wereley’s analysis.

294    In this regard, Dr Zaldivar did present several alternative analysis options in the Joint Report on Volume. But, having done that, he once again retreated to the position that he had received no instructions to proceed with them. He said that he would “undertake any of these options … upon receiving written instruction from the Court and an agreement that resolves how I will be compensated for such efforts”.

295    I do not regard it to be the role of the Court in an adversarial trial system to give such instructions or provide advice to parties as to the evidence they should adduce. In the end, Dr Zaldivar’s discussion of these options provided no meaningful assistance to the Court. The applicant submitted, with considerable justification, that Dr Zaldivar’s contribution to the topic of the volume of oil spilled could best be described as that of a sniper firing shots at Professor Wereley’s work: see the analogy described by Martin J in Mineralogy Pty Ltd v Sino Iron Pty Ltd (No 16) [2017] WASC 340 at [741] – [743].

296    Thirdly, the thrust of Dr Zaldivar’s various criticisms of Professor Wereley’s work seems to be that it was carried out incompetently, both methodologically and as a matter of computation or calculation. If that is what Dr Zaldivar intended to convey, then that is certainly not my impression of Professor Wereley as an expert, or of his evidence. There was no challenge to Professor Wereley’s expertise.

297    Professor Towler, who calculated oil volume using material balance (see below), reviewed the reports prepared by Professor Wereley and Dr Zaldivar and expressed the opinion that the methodology that Professor Wereley had used to determine flow rate was, in fact, correct. For completeness, I record that Professor Blunt, the other expert on oil spill volume, did not comment on Professor Wereley’s work or Dr Zaldivar’s criticisms.

298    Whilst there were difficulties in carrying out a flow rate analysis on the evidence available to Professor Wereley, and whilst the possible impact of those difficulties on the estimates provided by Professor Wereley must be recognised, I do not accept that his estimates can simply be cast aside and ignored as “unreliable” or “invalid”.

299    Fourthly, Professor Towler expressed the opinion that the original estimate of spill volume made by Professor Wereley (i.e., in his first report) was the most reliable estimate using the flow rate method. He noted that, when Professor Wereley’s conversion error was corrected, the total volume of oil spilled would be in the order of 3.64 million bbl. He said that the averaging used by Professor Wereley in his first report was likely to be more accurate than applying the trapezoidal rule, because reservoir flow rates tend not to decline much in the first 75 days of production. In fact, a typical decline rate for a petroleum reservoir is about 10% per annum. This would amount to a decline of about 2% over 75 days. Therefore, according to Professor Towler, Professor Wereley’s original assumption of a constant rate for 72 days was reasonable.

300    Relatedly, Professor Towler disagreed with Dr Zaldivar’s suggested method of integrating the measured flow rates into a total spill volume, which was based on a decline rate of 85% over 75 days. Professor Towler said that this decline rate was quite unrealistic and led to a decreased estimate of the total volume spilled.

301    Professor Towler also pointed out that when Dr Zaldivar carried out his calculation of the amount of oil spilled using Professor Wereley’s methodology, he (Dr Zaldivar) committed the same conversion error for which he had criticised Professor Wereley. Professor Towler’s evidence was that if Dr Zaldivar had used the correct unit conversion factor in his calculations (i.e., the one Dr Zaldivar said that Professor Wereley should have used), his total spill volume would have been between 906,000 and 1.18 million bbl, not 16,178 to 21,181 bbl.

302    Fifthly, Professor Towler noted that the H1 Well was designed to be able to flow at 28,500 bbl/day, through a tubing string of 5.5” (external diameter). However, this tubing string had not been installed before the blowout. Consequently, the reservoir fluid was free to flow through the 9 5/8” casing. This means that the oil would have been able to flow at more than 28,500 bbl/day. On the evidence, there was also an expectation that, once completed, the well would flow at 9,000 to 10,000 bbl/day, under controlled conditions. However, during the blowout, the conditions were not controlled and, according to Professor Towler, it would be expected that the well was flowing at a rate that was not restricted by its well potential.

303    Sixthly, as I have already noted, Professor Wereley’s calculations were based only on the flow from the horizontal drain pipe and not the other sources of oil flow which would have contributed to the volume of oil spilled into the sea over the 75 day period that the well was uncontrolled.

Material balance calculation

304    Material balance is an independent method of verifying the amount of oil that was lost from the reservoir feeding the H1 Well during the blowout. To conduct the analysis, the pressure loss in the reservoir is used. Utilising the known volumes of oil and gas originally in the reservoir, the original and final pressures before and after the blowout, the properties of the reservoir fluid, and an estimate of the water that has invaded the reservoir from the aquifer during the blowout, it is possible to calculate the fluid lost from the reservoir.

305    Professor Blunt described the material balance method in the following simplified terms:

3.4     ... When the well is drilled through the reservoir, it encounters the fluids in the reservoir rock: oil, natural gas and water. These fluids are stored at very high pressures in tiny pore spaces. The rock is under enormous pressure from the weight of rock above it. The well allows a release of oil out of the reservoir to the surface. As oil, water and gas start to flow through the rock’s connected pores, the pressure of the oil in the reservoir decreases. When the pressure decreases, the remaining oil expands, pushing the oil out of the reservoir. As the pressure of the fluids within the rock pores drops, the rock is compressed down, squeezing the pore spaces, pushing out more oil. The material balance principle says that the volume of oil, gas and water that comes out of the reservoir must be equal to the combined volume expansion of oil, water and gas, and the compression of the rock pore space. To calculate the oil produced from this fluid expansion and rock compression, only the compressibility of the fluids and rock, the size of the reservoir connected to the well, and the change in pressure need to be determined. These are the three variables that will be discussed throughout the remainder of this report.

3.5     In even simpler terms: the pressure-driven change in the volume of oil, gas, water and rock pore space in the reservoir is equal to what comes out of the well. This is the central concept in reservoir engineering: to relate what was produced to changes in the reservoir. We keep track of the volume of oil produced by calculating the volume of oil displaced from fluid expansion and pore contraction.

3.6     The advantage of the material balance approach is that it does not require knowledge of changing flow rates over time, or other indirect measures of volume, such as the extent of the spill, instead computing directly the total amount of oil produced.

306    Using the material balance method, Professor Towler determined that the amount of oil spilled during the blowout was in the range of approximately 520,000 bbl to 3 million bbl, with the most likely amount to be about 1.8 million bbl (on a simple average, 24,000 bbl/day). Professor Blunt determined that the amount was between 57,000 bbl to 74,000 bbl, with a mid-range value of 69,000 bbl (on a simple average, 920 bbl/day). He nevertheless accepted in cross-examination that, within the range of uncertainty, the volume of oil released, by reference to his calculation, could have been 100,000 bbl or more (on a simple average, approximately 1,333 bbl/day).

307    There is, obviously, a significant difference between the two ranges that were calculated. The significant difference between them is the approximations that each had made.

308    To explain, the Montara oil field has a gas cap above the oil. It is also connected to a large body of underground water (aquifers). The water in the aquifers, and the gas in the cap, expanded when the pressure dropped during the blowout.

309    Professor Blunt assumed, for the purposes of his calculation, that the expansion of the water was matched by the production of water, and that the expansion of gas was matched by gas production. Therefore, on his approach, the oil produced and spilled equalled the expansion of oil.

310    Professor Towler considered Professor Blunt’s material balance model to be highly simplified. He argued that Professor Blunt had ignored several important drive mechanisms. According to Professor Towler, the most important of these was the invasion of water from the aquifer into the hydrocarbon reservoir. As water invades the reservoir, it pushes oil in front of it and drives the oil towards the producing well. Consequently, oil is preferentially produced over water. Professor Towler saw the aquifer influx as the primary drive mechanism, with gas expansion and oil expansion as the secondary and tertiary drive mechanisms, respectively. The three mechanisms were accounted for in his model, whereas Professor Blunt’s simplified model relied on oil expansion alone for estimating the volume of oil spilled.

311    Professor Towler said that there were three aquifers contributing to a strong water drive. The first is an aquifer directly connected to the Montara hydrocarbon reservoir, which feeds the H1, H2, and H3 Wells. Professor Towler opined that this aquifer was the main contributor to the aquifer influx during the blowout. It contained 6.5 billion bbl of water, had a reported permeability of 2000 milli-darcies (mD) (a very high permeability) and a thickness of 44 m. The second aquifer, referred to as the Regional Aquifer, contained 24 billion bbl of water and was also connected to the Montara hydrocarbon reservoir. The third aquifer was located in the region of a formation called the Plover Formation. It contained 400 billion bbl of water. Professor Towler estimated that approximately 3.6 million bbl of water invaded the Montara hydrocarbon reservoir during the blowout. This was a very small fraction (0.0066%) of the 6.5 billion bbl of the water in the first aquifer, and an even smaller fraction of the water in the other two aquifers.

312    Professor Blunt considered Professor Towler’s estimate of water influx to be excessive. In his view, the water flow would have been impeded by faults in the reservoir and aquifer, and that these impediments had to be accounted for in the calculation. In assessing the degree of impediment, Professor Blunt relied on aquifer properties analysed in a study called the Montara Aquifer Study (2003) (the Montara Aquifer Study). Professor Blunt estimated the aquifer influx to be 57,000 bbl of water. This was based on the assumed permeability used in the Montara Aquifer Study of the connection between the Plover Formation and the Montara hydrocarbon reservoir, recorded as 33 mD. He argued that if the aquifer influx had been of the order estimated by Professor Towler, the spill would have been comprised almost entirely of water.

313    Whilst not disputing that account had to be taken of the tortuosity between the large aquifer in the Plover Formation and the oil and gas accumulation in the Montara hydrocarbon reservoir, Professor Towler argued that, in using the figure of 33 mD, Professor Blunt had used the wrong value for permeability because, during the blowout, the aquifer that was invading the reservoir was the 6.5 billion bbl in the immediate vicinity. In other words, there was no basis to assume that the permeability of this aquifer was the same as the permeability of the more distant aquifer in the region of the Plover Formation. Therefore, in his analysis, Professor Towler used an aquifer permeability of 1000 mD, even though the Montara Aquifer Study reported the permeability of the first aquifer as 2000 mD. In his oral evidence, Professor Towler said that he selected the figure of 1000 mD in order to be conservative. He also assumed that, in relation to this aquifer, the faults had minimal effect.

314    A further matter pointed out by Professor Towler was that, even though Professor Blunt estimated the aquifer influx to be only 57,000 bbl of water, this figure was not taken up by Professor Blunt; aquifer influx did not appear in his material balance equation.

315    In the Joint Report on Volume, Professor Towler and Professor Blunt agreed that determining the amount of water influx during the blowout represented a significant uncertainty. They agreed that the well bore was sufficiently conductive to allow a release of more than 1 million bbl of oil if the flow were unimpeded. As I have said, the issue on which they disagreed was how significantly the oil flow would be impeded. The evidence does not enable that determination to be made. I am left only with the separate possibilities posited by Professor Towler and Professor Blunt.

Conclusion

316    On the evidence before me, I am satisfied that the oil released over the 75 days of the spill was far greater than the suggested rate of 200 to 400 bbl/day. Of course, one will never know with certainty the volume of oil actually spilled in that period or the rate at which it was released. But taking the lowest volume estimate in the evidence adduced in this proceeding (Professor Blunt’s mid-range estimate), and assuming a relatively steady rate of release over the period (a simplifying assumption I consider to be reasonable for my purposes), it can be seen that, as a minimum, the oil would have been released at a rate of 920 bbl/day. This is significantly greater than the rate assumed in Dr French-McCay’s base case modelling (400 bbl/day).

317    The applicant submitted that once it is accepted that the rate of release was greater than 200 to 400 bbl/day, the precise volume of oil spilled need not be quantified. That might be so. However, I should record that I am satisfied that it is likely that the volume of oil spilled was far greater than Professor Blunt’s estimate and that, correspondingly, the oil was spilled at a far greater rate than implied by that estimate. Professor Blunt and Professor Towler acknowledged that there is considerable uncertainty involved in the volume of the aquifer influx that would drive the release of oil. This uncertainty is reflected in the two estimates made on the basis of a material balance calculation. I note, however, their agreement that the well bore was sufficiently conductive to allow a release of more than 1 million bbl of oil (on a simple average, approximately 13,333 bbl/day over the period in question) if the flow were unimpeded.

318    Between Professor Blunt’s and Professor Towler’s respective estimates is the estimate made by Professor Wereley in his second report (6,000 to 9,000 bbl/day). The estimate made in Professor Wereley’s first report significantly exceeds this rate (once his conversion error is corrected).

319    I accept that there are difficulties (and hence uncertainties) in making a flow rate calculation on the basis of the available evidence. But even taking his lower estimate, it should be recognised that Professor Wereley’s analysis was directed to a single source of discharge (the horizontal discharge pipe on the wellhead platform), and not the other sources of discharge at the H1 Well.

320    Taking all the evidence on spill volume into account, I am satisfied on the balance of probabilities that, over the period in question, oil was being discharged at an uncontrolled rate in excess of the ranges considered in Dr Hubbert’s modelling (in other words, in excess of 2,500 bbl/day).

321    It is appropriate to recall that the modelling carried out by the respondent in 2011 and revisited in 2013 included a loss of well control spill over 77 days, discharging 84,966 m3 or 534,380 bbl of Montara oil at rates varying between 3,802 m3/day or 23,912 bbl/day down to 690 m3/day or 4,341 bbl/day. This scenario was regarded as credible.

Dispersants

322    Dispersants were applied to the oil spilled from the H1 Well in the period 23 August to 1 November 2009.

323    The term “dispersant” is used to refer to a variety of chemical spill-treating agents that promote the formation of small droplets of oil which then “disperse” throughout the top layer of the water column.

324    Dispersants are mixtures of two components: surfactants and solvents. Surfactants (surface-active agents) are the active ingredients of dispersants which are mixed with a solvent. The solvent has two functions: it reduces the viscosity of the surfactant, which enables it to be sprayed, and it promotes the penetration of the surfactant into the oil slick. The surfactants in dispersants are also used in many common household products, including soaps, skin creams, baby bath, cosmetics, shampoos, mouthwash, intestinal medications, and even food.

325    Some portion of released oil will disperse into the water column whether or not chemical dispersants are used. Natural dispersion of floating oil is a process facilitated by wave action that breaks the oil into small droplets. It is affected by the properties of the oil and the amount of wave energy at the sea surface. In general, oils with lower viscosity are more amenable to natural dispersion than those with higher viscosity. Likewise, higher wave energy produces more natural dispersion.

326    Dispersants work because surfactant molecules have two distinct and linked parts: one is lipophilic (attracted to oil) which orients itself into the oil; the other is hydrophilic (attracted to water) which orients itself into the water. The surfactant molecules align themselves at the oil/water interface and reduce interfacial tension. Interfacial tension causes oil molecules to stick to each other in the form of a slick. High interfacial tension keeps the slick intact on the surface. The surfactant molecules interfere with this tension by reducing forces between the oil molecules, which enhances the breakup of an oil slick. When mixing energy is applied (for example, through wind, waves, and currents), the dispersant-treated oil slick will break up into many tiny microdroplets that are less than 100 µm in diameter. The microdroplets generally tend to stay suspended in the water column, while larger droplets are more likely to float back to the surface and re-coalesce into the slick.

327    One motivation for using dispersants is to treat an oil slick in the hope that the surface slick does not contact a shoreline. A second motivation is to reduce the impact of the oil on birds and mammals on the water surface. A third motivation is to promote the biodegradation of oil in the water column.

328    There is some dispute in the evidence as to whether biodegradation is enhanced. Dr Fingas said that the effect of dispersants on biodegradation is still a matter of dispute and that some current dispersant formulations can, in fact, inhibit biodegradation. He said that no enhancement of biodegradation is clearly shown in any recent studies.

329    Dr Coelho strenuously disputed Dr Fingas’ statements in this regard, noting that Dr Fingas had cited no references in support. She said that Dr Fingas’ statements were contrary to dozens of peer-reviewed publications on dispersed oil degradation, which she referred to and summarised in her first report. Dr Coelho said that the use of dispersants provides naturally-occurring, oil-degrading bacteria greater access to the oil by creating a diluted mixture of small oil microdroplets. The microdroplets are inherently more biodegradable than large, un-treated oil droplets because a greater surface area-to-volume ratio is available for microbial attack.

330    There was also disagreement between Dr Fingas and Dr Coelho about the effectiveness of dispersants both generally and in relation to their application to the oil spilled from the H1 Well.

331    Dr Fingas said that the main cause of dispersant ineffectiveness is high oil viscosity, which results in little dispersant mixing of the oil. He said that when dispersant droplets hit viscous oils they may run off into the water column rather than mix with the oil.

332    Dr Fingas said that the second cause of ineffectiveness is the lack of stability of the oil droplets in the water column over time. He referred to the phenomenon of resurfacing oil, which is the result of two separate processes: destabilisation of the oil-in-water emulsion and desorption of surfactant from the oil-water interface. The latter results in less surfactants in the oil droplets, and thus more coalescence of the droplets. The more coalescence that occurs, the larger the droplets will be, and the faster they will rise to the surface, reforming a slick.

333    In response, Dr Coelho referred to certain literature dealing with the effectiveness of dispersants on viscous oils. She also referred to multiple field studies on dispersant effectiveness which, she said, have repeatedly documented the fact that once a dispersant is applied, the dispersed oil microdroplets mix into the underlying several metres of water column. She said these microdroplets are diluted horizontally and vertically into the ocean so that the individual microdroplets cannot collide with each other. If they cannot collide, they cannot re-coalesce. In this connection, Dr Coelho drew a distinction between closed and open systems. The ocean is an open system in which the dispersed oil droplets dilute such that the tendency for them to coalesce is minimised. Dr Coelho said that this is a critical fact that inhibits suspended, neutrally-buoyant microdroplets from reforming into larger, positively-buoyant drops that rise to the surface and again become a surface slick.

334    With regard to the application of dispersants applied to the oil spilled from the H1 Well, Dr Fingas referred to data obtained from reports prepared by Leeder Consulting of the concentrations of oil under the oil plumes treated with dispersants at the time of the spill. These were so-called “grab samples” from which Leeder Consulting measured total petroleum hydrocarbons (TPHs) before and after the application of dispersants. The samples were taken at various depths over varying time periods. Based on these data, which showed low concentrations of TPHs in the water column after the application of the dispersants, Dr Fingas opined that the dispersants had low or no effectiveness on the oil that was treated (in oral evidence he said that the “treated oil had <10% effectiveness”). As I understand Dr Fingas’ point, the fact that concentrations of TPHs were detected showed this lack of efficacy. According to Dr Fingas, one of the principal reasons for this was the high wax content of the Montara oil. He said that the data implied that the dispersants entered the water column rapidly and were transported with surface currents.

335    Dr Fingas also analysed the data of sea surface temperatures at the time of the spill and compared these temperatures to the pour point of fresh Montara crude (27°C). He said that if the sea surface temperature was near or close to this temperature, the dispersants would run off the oil and not be effective. At the time, the sea surface temperature near the H1 Well was between 25 and 30°C. Dr Fingas said that this implied that the oil was sometimes near the point of not being dispersible because it was nearly solid and had a high viscosity.

336    Dr Fingas then considered various observations made at the time of the spill as to the effectiveness of the dispersants that had been applied, together with various contemporaneous photographs and videos. The comments made by observers were to the effect that the applied dispersants were, indeed, effective to varying degrees. Dr Fingas doubted the accuracy of these comments for a number of reasons, not all of which need to be summarised here. I refer to two of them below—the colour of treated plumes and the degree and mode of agitation of the dispersant-treated oil.

337    In response, Dr Coelho criticised the Leeder Consulting data relating to oil concentrations under the treated plumes on which Dr Fingas relied. She said that the “results [did] not reveal much about dispersant effectiveness”. This was because of the time that had elapsed between the application of the dispersant and the sampling that had been done. Dr Coelho said that the sampling should have occurred at shorter time intervals because dispersed oil concentrations dilute to very low concentrations in a short amount of time and the dispersed oil does not remain directly under the slick, but is transported away and quickly diluted. As I understand Dr Coelho’s point, the Leeder Consulting data was obtained belatedly and could not show the oil that was, in fact, dispersed by the applied dispersants and carried away.

338    With regard to Dr Fingas’ analysis involving the comparison of the pour point of crude Montara oil and sea surface temperatures at the time, Dr Coelho said that Dr Fingas’ analysis had failed to account for “the dark oil absorbing heat once the sun shines on the slick”. She expressed the view that if the oil was returning to a less viscous state due to daytime warming and sun exposure, it would “again become amenable to dispersant application”.

339    With regard to Dr Fingas’ comments on the contemporaneous observations of dispersant effectiveness, Dr Coelho expressed differing views. Once again, not all of Dr Coelho’s comments need to be summarised here.

340    As foreshadowed, I will refer to two matters in contention between Dr Fingas and Dr Coelho: the colour of treated plumes as an indicator of dispersant effectiveness, and the degree and mode of agitation that would promote the dispersion of the treated oil.

341    As to the first matter, Dr Fingas remarked on the absence of “effective coffee-coloured plumes” (which would show that the treated oil was being dispersed) and the presence of “white” plumes (which would indicate that dispersant had run off the oil and not functioned to disperse it) in the material he examined, which informed his judgment about the lack of accuracy of the contemporaneous observations of dispersant effectiveness.

342    Dr Coelho said that coffee-coloured plumes are not the only indication of dispersant effectiveness and that, although the appearance of “white” plumes might be from dispersant only, that appearance could also arise from poor viewing conditions. She said that current guidelines suggest that colour is not necessarily a good indicator of oil amenability to dispersants. Dr Coelho gave this evidence:

3.4.2      the initial dissipation of dispersant-treated oil does not happen at all locations in the treated slick at the same time. Individual plumes of dispersed oil may be seen at various locations shortly after (but often not immediately after) the passage of a wave through the area. The color then fades as the dispersed oil concentration in water decreases as dilution proceeds. The color change of the dispersed oil plume is important to note, but its appearance can take minutes to many hours to form, depending on conditions. During a US EPA study conducted during 2010, Hemmer et al. (2011) described visual observations of chemically dispersed oil solutions prepared using different dispersant products. Their visual observations ranged from “cloudy pearlescent white” to “very cloudy brown” to “very dark brown”. The NOAA Dispersant Application Observer Job Aid mentions that no color change may be observed when effective dispersion is taking place (NOAA, 2007).

3.4.3     Naturally dispersed oil, or oil treated ineffectively with dispersant, contains a much higher proportion of larger oil droplets that are only temporarily dispersed. A plume of dispersed oil may be briefly observed in the water after a wave has passed through, but the plume will be dark-brown or black with individual oil droplets evident upon close inspection. Spraying dispersant into water that contains no oil typically results in a white cloud of dispersant in the water.

3.4.4     Of course, adequate viewing conditions must exist before valid observations can be made. As with other visual observation techniques used in oil spill response, it is essential that the observer has the appropriate knowledge and training to make valid observations. Dispersed oil plumes may not be visible in poor light conditions, such as those present during semi-darkness or when grey, low clouds prevail. Bright sunlight and clear water create the best viewing conditions. If the water is turbid and contains high levels of suspended sediment, it might not be possible to distinguish the color of the dispersed oil plume from the background water color. Some dispersed oil plumes may be hidden under areas of surface oil if the prevailing currents cause the dispersed oil to drift under untreated oil areas. The presence of a colored plume is one possible indicator that the dispersant is working. However, its absence is not a definitive sign that the dispersant is not working. In any event, visual observations cannot be translated into a percentage of oil dispersed estimate.

343    As to the second matter (agitation), Dr Fingas remarked that, on many days on which the dispersant was applied, the sea was “very calm” and “too calm for dispersion”. Observers noted that agitation by the vessels applying the dispersant (for example, boat prop wash) promoted dispersion. Dr Fingas thought that this was “probably misleading” because such agitation was a temporary phenomenon which would “disappear in about 10 minutes”.

344    Dr Coelho said that, while it was once believed that waves greater than 1 m were required for proper mixing of the dispersant into the oil, this was no longer the case. She said that current guidelines now suggest that a wind speed of approximately 7 knots (a light to gentle breeze) with a wave height of 0.2 m or greater, is sufficient for mixing the dispersant into the oil. She also said that it is now advised that dispersants can be sprayed onto floating oil in flat, calm conditions, with dispersion occurring when appropriate sea conditions arise. Further, Dr Coelho’s reviews of AMSA daily reports indicated to her that the operators applying the dispersants clearly recognised when additional agitation from the vessel was necessary to achieve dispersion and that they provided the energy necessary to achieve the dispersion they observed at the time.

345    In contrast to Dr Fingas’ opinion that the application of dispersants to the Montara oil had “low or no effectiveness”, Dr Coelho expressed the opinion that overall dispersant effectiveness on the fresh and weathered oil (in the locations where it was used) “could easily have achieved 70+% effectiveness”. Dr Fingas criticised Dr Coelho’s opinion as “an unsupported statement” which was at odds with an analysis he conducted of field trials reported in the literature, which showed that the average effectiveness of dispersant application on oil was 16%. In expressing her opinion, Dr Coelho acknowledged, fairly, that it is not possible to accurately quantify the proportion of the total amount of spilled oil that has been chemically dispersed at any particular time.

346    In a subsequent report, to meet Dr Fingas’ criticism, Dr Coelho elucidated the basis for her opinion about dispersant effectiveness. She said that she reviewed AMSA’s operational data (visual observations, fluorometric data and laboratory results) as well as AMSA’s operational logs, which enabled her to conclude that dispersant operations were well-coordinated and effectively targeted fresh (unemulsified) oil. According to Dr Coelho, these activities resulted in the rapid formation of dispersed oil plumes. Generally, some external agitation was required to enhance the dispersion, but at other times the dispersion process was reported as only requiring naturally-occurring wave action.

347    Dr Coelho noted that a test conducted with one dispersant (Ardrox) on 7 October 2009 reported 100% effectiveness based on a combination of fluorometric data and associated visual field observations. Other field observations (logged in Daily Operational Reports) reported visual effectiveness estimates from 50% to 100% for any given slick location, and also documented that operations ceased in cases where the oil was no longer amenable to dispersion (for example, where the oil was waxy and not affected by dispersion even after agitation).

348    Dr Coelho said that the operational logs and videos taken by AMSA reflected “a well-trained, professional, and objective on-scene team conducting the dispersant operation and associated field observations”. Based on videos taken at the time, Dr Coehlo said that, in her “strong” opinion, the “dispersant” plumes observed by Dr Fingas were in fact “dispersed oil plumes”.

349    Dr Coelho noted that the videos she reviewed did not extend to the full duration of any dispersant operation, which frequently involved both spraying and agitation. Therefore, there was no video evidence that slicks re-formed and persisted at the end of a given operation. Dr Coelho said:

Executive Summary

In the absence of extended video data, I am of the opinion that the ultimate visual effectiveness estimates of dispersant operations should be based on the logs of shipboard observers who were able to observe the entire operation on a given slick, then make final estimates on the effectiveness of the dispersant operation to remove floating slicks.

My conclusions have been informed by a broad range of experience and past scientific efforts reported in peer-reviewed publications and other technical media. My estimate of 70% dispersant effectiveness, as stated in my primary report, is consistent with my enhanced understanding of the Montara dispersant operation, provided by the dispersant operations field log report evidence and video evidence. It is reasonable to expect high dispersant effectiveness when considering that the dispersant application was made on dispersible fresh oil, by well-trained, professional crews.

(Emphasis in original.)

350    I should record that, although Dr Coelho relied on fluorometric data and laboratory results, she acknowledged their limitations. Dispersant efficacy can be measured quantitatively in a laboratory environment but the results cannot be translated directly into quantitative estimates of dispersant effectiveness at sea. What the laboratory data do provide is information such as the likely “window of opportunity” for dispersant use and the relative effectiveness of different products and the effect of the dispersant treatment rate.

351    The effectiveness of dispersants used on oil spilled at sea can be estimated by fluorometry but, as Dr Coelho explained:

3.1.2    In situ fluorometry is used only to provide a qualitative indication of a relative increase of oil in the water column. As such, fluorometry is a response monitoring tool to make real-time “go” versus “no-go” decisions for continued dispersant use; however, it is not an accurate means of characterizing the amount of oil dispersed into the water column (Tan, 2011 AMOP). Additionally, in situ fluorometry is susceptible to significant variation due to effects from fouling of the instrument, sensor drift, the local environment, and frequency of calibration (Earp et al., 2011).

352    The fluorometric field evidence on which Dr Coelho relied was AMSA Operational Monitoring Study 02 Monitoring of Oil Character Fate and Effects – Report 03 Dispersant Treated Oil Distribution. The authors of that report stated (and Dr Coelho noted) that operational monitoring is never intended to determine dispersant efficiency. Further, the authors made clear the limitations of their study: they had limited time to sample; different dispersants were used; and energy levels varied.

353    Thus it can be seen that Dr Coelho’s estimate of “70%+” effectiveness is based more squarely on her informed understanding of the qualitative observations made by others in the field at the time, to which she assigned a percentage figure to reflect her own qualitative assessment or level of satisfaction regarding dispersant effectiveness. The figure of “70%+” has no greater foundation than that. Of course, Dr Fingas’ figure of <10% effectiveness is also a qualitative assessment.

354    The gap between the different opinions expressed by Dr Fingas and Dr Coelho on the effectiveness of the dispersants applied to the oil from the H1 Well did not narrow during the course of concurrent evidence. Each held to their respective views. I am left with two very different qualitative assessments (<10% effective v 70%+ effective) which, properly understood, are really no more than impressionistic expressions of opinion.

355    Further, what does it really mean to say that the application of dispersants was <10% effective or 70%+ effective? The evidence of the observers was that sometimes the application of the dispersants was effective; sometimes it was not. When it was effective, they expressed a view about how effective the application was, expressed as a percentage figure. But the reality is that there is no way of quantifying the oil that was treated and, if it was treated, no way of quantifying the oil that was completely dispersed or dispersed to some extent. There is also no way of quantifying the oil that appeared to be dispersed but might have subsequently reformed into a slick. Thus, leaving aside their necessarily impressionistic character, the percentages expressed lack a meaningful reference.

356    For completeness, I record that Dr Stout also gave evidence on this topic. He referred to the analysis by Leeder Consulting of the grab samples on which Dr Fingas relied. It will be recalled that Dr Fingas remarked on the low TPHs measured in the water column after the application of the dispersants as demonstrating the dispersants’ lack of effectiveness in dispersing the Montara oil. Dr Stout said that the fact that there were increases in the measured TPHs after the application of the dispersants (albeit small increases) indicated that some dispersion of oil had occurred and that this, in turn, suggested that the dispersants were effective.

357    I am not persuaded by that evidence. In fact, it appears to be significantly qualified, if not contradicted, by other evidence given in the Joint Report on Dispersants by Dr Stout and Dr Coelho who said that measured TPH concentrations in grab seawater samples will never reflect theoretical TPH concentrations based on a mass balance calculation because, following the application of a dispersant, the sampling can never be done with sufficient resolution over time and space. Further, according to Dr Coelho, there was insufficient field sampling of the treated Montara oil to judge overall dispersant effectiveness. She said that it would require thousands of samples to even partially characterise the effectiveness of the dispersant operation. Dr Stout also appeared to acknowledge this during the concurrent evidence session on this topic.

358    Dr Stout also referred to his long-term weathering study (to which I will return in greater detail below) in which he mixed evaporated (21%) Montara oil with a dispersant (Slickgone NS) in the ratio 20:1 by volume. When he did this he observed that the oil/dispersant mix immediately “disappeared” into the seawater in the microcosm. Dr Stout conducted a further experiment in which he mixed fresh Montara oil with Slickgone NS in the same proportion. When the dispersant was added to the floating oil in the microcosm, he observed that it immediately entered the seawater.

359    Dr Stout said that these observations were important because they showed that fresh and weathered Montara oil were both amenable to dispersion, at least by Slickgone NS. However, in saying this, Dr Stout acknowledged that his long-term weathering study did not attempt to replicate the effect of energy or the degree of weathering on dispersant effectiveness. Further, the small volume of oil used in the long-term weathering study was equivalent to a sheen (<10µ thick), not a slick, which would represent an insufficient mass of oil to treat with dispersants.

360    Dr Stout’s studies provide limited assistance on this topic. The unanswered question remains as to how effective, in some quantifiable sense, was the application of the dispersants to disperse the oil on which it was used, remembering also that a number of different dispersants were used at different locations where slicks were observed and treated.

361    One important fact that emerges from the evidence is that only a part of the spilled oil was treated with dispersants, and only a small part at that. The point emerges most clearly from the following passage in Dr Coelho’s first report:

20 (h)    Much of the Montara oil was never treated with dispersant, based on the overall small volumes of dispersant applied, compared to the overall released oil volume. During a continuous release, unless dispersant operations can proceed 24-hour per day, it is unlikely that all surface oil can be treated. Poor weather and night conditions preclude the use of dispersants to treat all oil. It is my opinion that the presence of slicks far from the well site is most likely because they were never treated with dispersants.

362    This is a particularly telling fact. It leads me to conclude that a very large part of the oil released from the H1 Well was not treated with dispersants. And, as I have found, the volume of oil released over the 75 days of the spill was far greater than the volume suggested by the reported release of 200 to 400 bbl/day. Given the prevailing temperature and sea conditions at the time, I am not persuaded that much of the untreated oil was dispersed by natural forces, particularly as it underwent increased weathering over time. Dr Fingas and Dr Coelho agreed that the increased weathering of oil generally reduces its susceptibility to dispersion. In the case of Montara oil, they agreed that increased weathering, with wax aggregation and separation, would have produced solid wax aggregates that were not amenable to dispersion.

363    Further, the effectiveness of the dispersants on the treated oil varied. Sometimes the dispersants were observed to be effective; at other times they were observed to be ineffective or only partially effective, leaving undispersed plumes to be moved, as the untreated oil was moved, by ocean currents and winds. The evidence does not enable me to be more precise than that.

Satellite imagery

364    Dr Gundlach, who was called by the applicant, carried out a review of data and materials relevant to determining the spread of oil, to express an opinion on whether oil from the Montara oil spill reached the coastal areas of the Rote/Kupang region.

365    He used a variety of data sources for his review, including: satellite imagery taken from a number of sources including the NASA MODIS Terra and Aqua satellites (including images analysed or interpreted by others), images made available by the European Space Agency (ESA) and images from Google Earth; overflight maps, flight path information and reports on observations prepared by AMSA; wind and oceanographic current data from published analyses of satellite data (specifically, NASA Quick Scatterometer (QuikSCAT) wind records and maps of ocean current from the MODIS satellite as processed by the Australian Integrated Marine Observing System (IMOS)); and computer oil spill modelling results sourced from Asia Pacific ASA Pty Limited, known as Asia-Pacific Applied Science Associates (APASA), which had provided reports to the respondent and DEHWA at the time of the spill. He followed the location of the oil from the morning of the blowout on 21 August 2009 until 28 November 2009, a period of 100 days.

366    Dr Gundlach’s review focussed on, but was not limited to, spill locations that were (a) north of the Australia-Indonesia maritime boundary (175 km from Rote and 80 km from the Montara H1 Well), and (b) which crossed north of latitude 11° 30' (90 km from Rote and 165 km from the Montara H1 Well).

367    Dr Gundlach summarised his review method as follows:

1.3 Review Methods

I reviewed the source of the spill location information, particularly to determine if overflights (a) spotted oil on the northernmost part of their overflight path, and then if they passed within the area between the Rote/Timor shorelines and roughly 50 to 100 km offshore.

I used the wind and current data to consider, based on the last spotting from overflight or satellite data, if winds and currents could reasonably push surface oil from the point of last observation closer to and possibly impacting Rote and adjacent islands.

When reviewing satellite imagery, I particularly looked for potential oiling on the water’s surface as visually revealed by streaking and contrasts in colouration compared to surrounding waters. For shoreline oiling, I looked for black or dark colourations that are flat in appearance and not textured or lumpy as is shoreline wrack composed of seagrass or seaweed.

I looked at all data in a series of chronological steps based on the quantity of available data and the potential for impacts to Rote Ndao and Kupang. In each case, I include the key data utilised for my analysis to enable further review by the reader and to indicate the quality of the data (or lack thereof).

Natural sheens and streaks are present in the world’s oceans, including being evident in historical imagery outside the spill–related 2009 time frame within the Montara area. For this reason, I use a combination of information to support my conclusions, particularly oil spill modelling results coupled with streaks/sheens/or heavier slicks shown on satellite imagery. Aerial observations made from surveillance overflights, coupled with modelling and imagery are also used in my analysis, but unfortunately these overflights never went closer than approximately 30 km to the coast and commonly stayed 50 to 70 km offshore.

368    Dr Gundlach summarised the results of his review, as follows:

1.4 Results

My analysis of available data reveals that impacts first occurred to Rote on 8 September and continued over a period extending until 17 October. After 17 October, data indicate that oiling remained closer to the Montara wellhead and did not reach Rote. However, APASA spill modelling shows that a large amount of oil remained floating in the area (maximum aerial coverage of 112,000 km² in late November), some of which might have come ashore during this period in Rote and Kupang.

369    In his principal report, Dr Gundlach presented a more specific analysis of the likely passage of the spilled oil by reference to five time periods: 30 August to 6 September 2009; 7 to 11 September 2009; 12 to 18 September 2009; 19 to 30 September 2009; and 1 to 21 October 2009.

370    Before summarising this evidence, it is necessary for me to say something about two matters which must be understood in order to gain an appreciation of the import of Dr Gundlach’s review. The first concerns the notion of hindcasting in the context of computer-based oil spill trajectory modelling, a topic with which I deal more extensively in a later section of these reasons. The second concerns the reliability of remote sensing techniques. Both matters were covered by Dr Gundlach in his report.

371    Forecasts provided by computer modelling are used to predict the likely position of the oil to aid response operations. A hindcast (or backtrack) is an after-the-fact simulation which uses data on known oil locations to enable the computer model to make better forecasts of the later positions of the oil. For example, knowing where oil has been observed enables the currents and winds data in the model to be manipulated to cause the model to mimic the observed oil locations. The hindcast is, in effect, used as a stepping stone to a new and further forecast. The capacity to augment and calibrate the relevant data in hindcasting means that hindcast modelling is preferred to forecast modelling. Nonetheless, Dr Gundlach remarked:

4.3.3     Oil Spill Model Forecasting versus Hindcasting

The test of an oil spill model is not how well it hindcasts after knowing where oil was found and then altering computer input parameters to match, but how the forecasted results match what is then observed in the future.

372    On the question of remote sensing, Dr Gundlach acknowledged, and discussed, the limitations of, and challenges presented in making judgments on the basis of, remote sensing when tracking spilled oil from helicopters and airplanes, and when using various types of satellite information.

373    With respect to satellite imagery, he pointed out that a “confluence of circumstances” must be present to detect an oil spill in satellite imagery, including: the availability of images; the resolution of the images relied on; and the influence of cloud cover, sun angle, and sea state. He acknowledged that only a limited set of images might have the right conditions for oil detection.

374    In the context of remote sensing, Dr Gundlach also discussed the issue of false positives related to natural phenomena. He said:

4.1.1     False-Positives Related to Natural Phenomena

The open sea has many colourations due to cloud cover, sun angle and glint, as well as a host of natural phenomena that may appear similar to an oil slick. Principally among these in the offshore realm are the presence of plankton blooms, fish spawning, natural calm areas (may be due to light natural organic sheen), seaweed concentrations and cloud shadows. Sediment plumes are also seen in the Montara spill area at Timor Island. …

Observers on surveillance overflights of the Montara spill also may have had difficulty separating out natural slicks from spilled oil. Rather than confirming oil or natural slicks, such terms as yellow, white or [orange slicks were] used in later aerial surveys. …

I am well aware of natural colourations and sheens in satellite imagery, including that found in historic images in the Rote area but outside the time frame of the Montara spill. Therefore, I ask the following questions as a minimal confirmation that the observation of an anomaly (e.g. sheen, streak or difference in colouration) on an image is probably spill related:

    Can it be tied via a connected pathway to the spill source?

    Do images or overflight reports, including those from a previous or post-date observation, support oil in that area?

    Do wind and/or current data support the movement of oil to that location?

    Do available modelling forecasts or hindcasts predict oil movement to (or near to) that location?

Confirming oil in nearshore and onshore conditions can be equally difficult if there is not a direct oiled pathway from the spill site. …

When detailed shoreline images are available (e.g. Google Earth), I look for black or dark shoreline accumulations and if there are slicks associated with the shoreline banding. Oiling on a shoreline will appear flat, whereas natural wrack composed [of] seaweed and/or seagrasses is likely to be textured or lumpy. I also look to see if satellite imagery and/or oil spill modelling support the presence of oil in that area at that time.

375    I now return to Dr Gundlach’s findings.

376    According to Dr Gundlach, the first sightings of oil crossing into Indonesian waters was on a MODIS Aqua satellite image dated 26 August 2009. On 30 and 31 August 2009, this was confirmed by another satellite image and by aerial observations. APASA’s oil spill trajectory modelling and the satellite imagery showed that the movement of the oil progressed northwardly. As at 6 September 2009, the oil was positioned approximately 80 km offshore of Rote. It was further offshore of Kupang.

377    From 7 to 11 September 2009, the winds and currents data used by Dr Gundlach, and the APASA modelling, indicated to him the continued transport of the oil to the north and to the west. APASA model hindcasts, prepared on 17 September 2009, showed a continuing movement of the oil towards Rote and Kupang, with impacts occurring to Rote on 8 September 2009, and continuing with more substantial impacts to Rote and Kupang on 9 and 10 September 2009. According to Dr Gundlach, APASA’s hindcasting of the oiling at Rote was supported by a MODIS Aqua image from 8 September 2009. Two interpreted MODIS satellite images from 10 September 2009 showed a large oil plume offshore of Rote. LandSat satellite images taken on the same day showed streaking in and around the northeast part of Rote. Dr Gundlach concluded that the APASA modelling supported a finding that the streaking was related to oil slicks.

378    Dr Gundlach also had recourse to Landsat satellite images taken on 8 and 11 September 2009 and Google Earth images, which he understood to be derived therefrom. He concluded that the imagery from 8 September 2009 showed several locations that appeared to be oiling on the shoreline at Rote, in the general area indicated as oiled by APASA’s modelling and the images referred to in the preceding paragraph of these reasons. Dr Gundlach drew particular attention to certain satellite images from 8 and 11 September 2009 (reproduced in his report) of the northeast and northwest shorelines of Rote. In respect of the images of the northeast shoreline he said:

21.    I believe that the black material on the shoreline has a high likelihood of being oil and not wrack composed of seaweed or seagrasses. These black swash lines are not similar on other images prior to or after this time frame. I would also expect it to be similar to other areas having seagrass or offshore algal beds, and it is not. Lastly, it doesn’t show the thicker texture commonly involved with wrack on the shoreline. By 11 September, most but not all of the [apparent] oiling was washed away, which is common on exposed sandy shorelines. These images support the previously discussed APASA modelling output and satellite images showing oiling in this same area.

379    Dr Gundlach said that his review of marine currents, from 8 to 11 September 2009 on a 4 hour basis or less, indicated a constant throughflow to the north-northeast between Timor and Rote, supporting the modelling and the indications of oiling he observed from the satellite images.

380    Dr Gundlach summarised the likelihood of oil impacts during this period as follows:

5.2     Spill Location 7 to 11 September 2009

To recap, oil impacts to the northeast portion of Rote during the 8 to 11 September 2009 timeframe are supported by several sets of imagery, APASA hindcast modelling, and wind and current data. Impacts to Kupang are shown by APASA modelling on 9 and 10 September. Likely oiling to NE and SW Rote and Kupang is illustrated by a MODIS Aqua image and APASA spill modelling on 8, 9 and 10 September. Google Earth images show likely oiling to NE and NW Rote on 11 September.

381    With respect to the period 12 to 18 September 2009, Dr Gundlach said (after reviewing and discussing various data):

5.3     Spill Location 12 to 18 September 2009

… winds and currents continue in a mostly northerly direction towards Rote. … The APASA hindcast of spill distribution from 11 to 15 September shows contact with all shorelines in Rote and close to Kupang. The MODIS terra image from 17 September shows oil close to Rote. A slight deviation in the APASA model forecast for 17 – 18 September would likely bring oil to Rote. The islands of Sawu and Raidjua are directly in the path of APASA-predicted impacts for 17 and 18 September. Overflights did not reach Rote nor extend to the furthest northerly location of observed oiling. Flights turned back before reaching the end of oiling in the Rote direction. The closest overflight to Rote observed oiling within ~30 km on 13 September.

382    With respect to the period 19 to 30 September 2009, Dr Gundlach said (after reviewing and discussing various data):

5.4     Spill Location 19 – 30 September [2009]

… wind and current data, and imagery support the extension of likely oiling of Rote at least through 26 September. … A Landsat image, supported by APASA modelling for 26 September shows likely oiling extending along most of Rote northeast to the channel between Timor and Rote, coming close to Kupang. Oil spill modelling by APASA shows several dates when oil was predicted to be in these areas. Oil surveillance flights continue to find oil in Indonesian waters but remain far (60 km and more) offshore of Rote.

383    With respect to the period 1 to 21 October 2009, Dr Gundlach said (after reviewing and discussing various data):

5.5     Spill Location 1 to 21 October 2009

A summary of the likely oiling in the Rote–Kupang area is shown in Figure 5–78. Imagery from 3 October 2009 show likely oiling around southwest and northeast portions of Rote. APASA modelling for the same day shows oil within ~21 km to the shoreline but misses concentrations seen on imagery. Aerial surveillance flights remained 60 to 130 km offshore so were in no position to observe oiling close to Rote. Imagery from 10 October 2009 shows potential oiling around Rote, Sawu and Raidjua. APASA modelling for the same day shows oil 50 km offshore of Rote extending far westerly and south of Sawu and Raidjua. In the model output, the influence of a strong westerly current is evident. Had the westerly flow been less and/or more northerly, there would be closer conformance to the image from 10 October.

APASA model forecasting for the period of 13 to 17 October shows oil distribution coming within 20 km of Sawu and 45 km to Rote. A later spill forecast shows oiling within 17 km of Rote on 17 October. A very minor change in model input parameters would bring oil onshore at the time. APASA modelling also predicts the extent of oiling for the period 17 to 20 October to cover an east–west direction extending over 1,100 km. A satellite image from 17 October shows anomalies (colouration and streaking) extending around NW Rote to the islands of Sawu and Raidjua. The conformance of potential oiling in the satellite image coupled with the extensive oil coverage predicted by the APASA model supports that oiling likely reached these areas at this time.

After 21 October, modelling and satellite imagery indicates that oiling remained closer to the wellhead. Even though oiling was close to the wellhead, significant differences between the modelled and satellite–observed spill position of ~30 km are evident within 50 km of the wellhead. As the distance to Rote and Kupang are at least five times greater (250 km compared to 50 km), differences between modelled and actual oil position are likely to be much greater.

After 20 to 21 October, overflights and limited available satellite imagery show no sign of oil reaching the Rote area. A review of these data are provided in Appendix 2.

From 21 October until near the end of November, APASA provided spill location forecasts (Figure 5–79). The Montara well was capped on 3 November 2009, however, as indicated by APASA modelling, extensive oiling remained on surface waters and showed a substantial increase in areal coverage from approximately 30,000 km² to 212,000 km² until modelling was discontinued in late November 2009.

384    I note for later reference that some of the APASA modelling on which Dr Gundlach relied was the subject of review by Dr Hubbert who, as I later explain, was critical of this modelling because it significantly underestimated, in his opinion, the locations to which the oil had drifted.

385    As the case developed, particular focus was placed by the respondent on Dr Gundlach’s use of, and reliance on, satellite imagery. The respondent relied on a report from Dr Garcia-Pineda to challenge Dr Gundlach’s evidence insofar as it was based on conclusions to be drawn from the imagery, and in particular the shoreline imagery which Dr Gundlach said showed, with a high degree of likelihood, oil and not wrack.

386    Amongst other things, Dr Garcia-Pineda analysed private and public records of satellite imagery collected before, during and after the Montara oil spill which covered the area of the incident, and regional boundaries. The images he and his team analysed included images taken by the MODIS Aqua and Terra, Landsat 5, Landsat 7, RADARSAT-1, ENVISAT, ALIOS, and GeoEye satellites. He provided daily maps (on some days multiple maps were prepared) showing the satellite image footprints of the areas of interest, and made remarks about the interpretation of the images so far as they related to the presence or absence of surface oil. He and his team assigned levels of confidence (low, medium, high, or false positive) as to whether particular images showed the presence of oil. Dr Garcia-Pineda explained that there could be cases where a feature is accorded a mixed classification, such as low-medium or medium-high. This would be where some areas of an image comply with certain conditions, but those conditions change on another area of the image:

156    For example, there could be cases where a feature (partially or on one side) fits the characteristics of a medium confidence level, but on some other areas of the feature the illumination conditions are not so good (or there are some imaging artifacts) and the feature on that particular region has a low confidence level, therefore the confidence level would be in combination “medium-low”.

387    As to confidence levels more generally, Dr Garcia-Pineda said:

157    This confidence level is a type of interval estimate obtained from a review of qualitative aspects to understand the classification as a quantitative measure. A High confidence level would mean that the probability for a given feature to be oil is above 75%. A Medium confidence level would mean that the probability for a given feature to be oil is between 50% and up to 75%. And finally, a low confidence level means that a given feature is below 50% probable [sic] to be actually floating oil. On cases when given its particular characteristics, a feature is identified as false positive, then I assign that classification meaning that [it] is highly probable that given feature is not oil. Another situation could be when the imaging conditions are not optimal, but the feature itself shows clearer signs of attributes that could be associated with oil, then for those cases I could assign a medium to high level of confidence. Again, all of these confidence level intervals can be confirmed and proven if ancillary data would be available.

388    Like Dr Gundlach, Dr Garcia-Pineda stressed the importance of ancillary data and in situ observations when assessing satellite imagery to detect oil on the surface of water:

153    During large oil spills, the assessment of the spill’s extent on satellite imagery needs to be confirmed and complemented using in situ observations from responding operations. For example, records of flight paths, vessel tracks, aerial photography, dispersant applications, or sample collections logs can be used to confirm that features observed on satellite imagery are, in fact, from oil slicks and not signatures generated by other processes (i.e., false positives or oil look-alikes). Within reasonable space and time gaps, all available satellite and in situ information should be used to validate possible oil features on satellite imagery. By space and time gaps, I mean observations made within a distance where events could be correlated by location (typically no more than 20 km) and time (typically no more than 24 hours). This process is done by overlaying the spatial records (chronologically) with the polygons of the delineated possible oil slicks. Then, a temporal/spatial analysis is carried out to confirm that features observed correspond to oil, and not to other look-alike features.

389    However, in carrying out his analysis of satellite imagery, Dr Garcia-Pineda, unlike Dr Gundlach, based his report solely on image characteristics. As he made clear:

158    At the time of this report, I was not provided ancillary data, therefore, this report has been limited to the identification of features on satellite imagery (based on their aspect and imaging conditions). Further analysis is needed to correlate the observed features with in situ observations records. This correlation process would provide an extra level of confirmation for classification of the features. As I stated previously, when a limited number of in situ observations are available, meteorological and oceanographic records can be used to track the direction and speed of possible oil slicks in an attempt to correlate features that are separated in space and time. Records of meteorological or oceanographic conditions during the Montara oil spill event were not available for this analysis …

390    In this part of his report, Dr Garcia-Pineda was referring to the use of ancillary data and in situ observations to confirm an assessment already made from satellite imagery. However, I take his observations about the utility of ancillary data and in situ observations to apply equally to the interpretation of satellite imagery in the first place, not merely as a confirmatory tool.

391    Subject to the limitations of his analysis, based on image characteristics alone, Dr Garcia-Pineda identified 19 independent cases of possible oil features inside a buffer extending 50 km from the shorelines of Rote and Timor. These features were observed on eight different days (10, 11, 26, and 27 September, and 5, 13, 20, and 29 October 2009). The closest feature was detected 10 km from the Timor shoreline on 5 October 2009, with a medium level of confidence.

392    It is appropriate that I also draw attention to a collection of images that were obtained from various satellites on 26 September 2009. Dr Garcia-Pineda and his team assessed some of these images as showing, with a medium level of confidence, surface oil in the range of 36 km to 43 km from the shoreline of Rote. He and his team assessed other images taken on the same day as showing, with a low level of confidence, surface oil in the range of 37 km to 44 km from the shoreline of Rote. The different confidence levels expressed with respect to these images appear to be a function of the individual quality of the images for assessment purposes. Further, Dr Garcia-Pineda and his team assessed images obtained from various satellites on 13 October 2009 as showing, with a medium level of confidence, surface oil in the range of 32 km to 36 km from Timor.

393    Dr Gundlach criticised Dr Garcia-Pineda’s use of confidence levels. He said that these were “highly interpretative”. These interpretations were visual, not computer-based, assessments of the features of interest which, Dr Gundlach argued, were not replicable and led to inconsistent findings across images. He illustrated this point by reference to overlapping images. Dr Gundlach compared these images and said that almost all of the images where overlaps occurred showed a disparity in assessment by Dr Garcia-Pineda and his team. For example, the same feature shown on an overlapped area on one image might be defined with “High” confidence while, on the next image, it might be defined with “Medium” confidence, or the feature might be defined with “Medium” confidence on one image but with “Low” confidence on another. Dr Gundlach said that he observed this kind of disparity throughout the area analysed. Dr Gundlach also observed that Dr Garcia-Pineda and his team had not consistently identified the same feature across images. One analyst analysing one image might find that a feature was not oil whereas another analyst analysing the same feature on a different image might find, with “High” confidence, that the feature was oil.

394    Dr Garcia-Pineda did not dispute that different levels of confidence in respect of findings of possible oiling were offered in respect of adjacent images or that there might be seemingly inconsistent findings with respect to the one feature shown on multiple images. However, he did not regard these as inconsistencies, as such. He said that each image was analysed individually and separately from each other image. The quality of one image in respect of a particular feature may differ in an adjacent image in respect of the same feature. It would therefore affect the confidence with which the feature could be identified.

395    Dr Gundlach also remarked that, of the 122 images analysed by Dr Garcia-Pineda and his team, false positives were identified in five images—all located close to the shorelines of Rote and Timor. By comparison, of the 14 images where possible oiling was identified with the Australian coastline (a roughly equal distance from the H1 Well as Rote and Timor), none had been identified as false positives. The suggestion was, plainly, that there had been unequal treatment of the images.

396    Dr Gundlach also examined images in Dr Garcia-Pineda’s report that he (Dr Gundlach) had not commented on in his main report. He concluded that several images indicated the possibility of oiling near or at the shorelines of Rote and Timor which, in his view, had not been “fully recognised” by Dr Garcia-Pineda.

397    For his part, Dr Garcia-Pineda was critical of certain aspects of Dr Gundlach’s evidence. He disagreed with the enhancement techniques that Dr Gundlach employed to examine some satellite images. He criticised Dr Gundlach’s use of, and reliance on, Google Earth images. He argued that certain satellite images that he and his team analysed contradicted Dr Gundlach’s interpretation of a corresponding satellite image. In a Joint Report prepared by Dr Gundlach and Dr Garcia-Pineda, Dr Garcia-Pineda noted (with cross-references to his report) the respects in which he disagreed with Dr Gundlach’s interpretation of possible oil features in satellite images discussed in Dr Gundlach’s report, although there was agreement between the experts that some images revealed features that might contain oil.

398    Dr Garcia-Pineda also examined images collected before and after the Montara oil spill by the same satellites used during the spill. He presented several cases where similar features (aspect, shape, texture, size, and colour) observed at the time of the spill in a particular area could also be observed outside the timeframe of the spill. As Dr Garcia-Pineda put it:

13    … by analysing imagery from years prior to or many months after the oil spill, I was able to identify some features that could not be oil from the Montara oil spill, but that somewhat could resemble oil look alike. This type of comparison exercise is critical and confirms the importance of including ancillary data and records, as well as oceanographic and meteorological data, in the analysis of satellite images in order to confirm for each of the features detected if they are in fact oil or an oil-look alike.

399    The thrust of this evidence was not merely to emphasise the need to have regard to ancillary information and in situ observations; it was also to contend that what Dr Gundlach had identified as possible oiling in some of the satellite images on which he relied could, by features shown in images of the same locations taken outside the oil spill timeframe, be explained as other, natural phenomena and not oil.

400    In a presentation he gave in the course of concurrent evidence, Dr Garcia-Pineda said that, based on the remote sensing data he and his team analysed, there was no evidence that oil could have reached the shorelines of Rote and Timor. In his Joint Report with Dr Gundlach he went further to say that he could:

10 (a)    express with high confidence that oil did not reach the coastal areas of Rote Ndao or Kupang.

401    I do not accept that, in this part of the Joint Report, Dr Garcia-Pineda was intending to express such an absolute opinion. When cross-examined in the course of the concurrent evidence session with Dr Gundlach, he accepted that this opinion was given by reference to the satellite imagery that he and his team analysed. Further, he readily agreed with the proposition that the fact that oil cannot be discerned from a satellite image does not demonstrate, determinatively, that oil is not present. He also agreed that his opinion—that oil did not reach the coastal areas of Rote or Timor—was based solely on the analysis of the satellite imagery and did not take into account the in situ observations of the lay witnesses who had given evidence in the case.

402    In some ways, the two experts were as ships passing in the night. On the one hand, Dr Garcia-Pineda analysed satellite images and, although acknowledging—indeed, stressing—the limitations in interpreting these images, and confirming the criticality of consulting ancillary data and in situ observations, did not consult such information before expressing the opinion that there was no evidence from the images he analysed that oil could have reached the coastlines of Rote and Timor. Contrary to Dr Gundlach, he did not accept that computer trajectory modelling was a means of satellite image verification.

403    On the other hand, Dr Gundlach analysed satellite images and, without assigning any particular level of confidence (he did not, in practice, use confidence levels), concluded that they showed the possibility that oil was present at the locations he identified, which conclusion, in his view, was supported by other data he consulted, including computer trajectory modelling. He eschewed Dr Garcia-Pineda’s analysis referring to confidence levels (which he said was highly interpretive and led to inconsistent findings) and false positives (which, in his view, could not be demonstrated by an analysis of computer images alone). To his mind, even Dr Garcia-Pineda’s false positives were all sightings of possible oil features. Dr Gundlach accepted that anomalies were thrown up by the satellite images taken before and after the Montara oil spill of certain features at certain locations but, in referring to those anomalies (as they might affect the interpretation of the images at the time of the spill), he said:

… That’s the exceptional part. There’s a large oil spill out there and these have oil-like features.

404    Dr Gundlach also expressed his confidence that, when viewing satellite images, he could distinguish between features that showed possible oiling and features that were obviously natural phenomena, like seaweed wrack.

405    Towards the end of the concurrent evidence session on this topic, there was a narrowing of the differences between the two experts, reflected in the following exchanges with Senior Counsel for the respondent:

MR SCERRI: Yes. Now, in this case – because this is the reason I say this is a different context, Dr Gundlach, in this case, this happened more than 10 years ago, so we can’t send the US Coastguard out to get a sample. You accept that?

DR GUNDLACH: Yes.

MR SCERRI: So aren’t we left with the position that you look at the images and you say they’re possible oil features?

DR GUNDLACH: Yes.

MR SCERRI: But you don’t know? I think you said - - -

DR GUNDLACH: That’s correct.

MR SCERRI: - - - several times in your presentation, “I don’t know, but they’re possible features.”

DR GUNDLACH: Correct.

MR SCERRI: And I think Dr Garcia-Pineda, you say the same thing?

DR GARCIA-PINEDA: I say the same thing, but I also say I know which features are not oil.

MR SCERRI: Yes. I was going to get to that. The difference between you is that you say I can identify some as false positives and Dr Gundlach doesn’t do that. Again, you might- - -

DR GUNDLACH: That’s good. I – that’s correct.

MR SCERRI: That’s correct.

DR GUNDLACH: Yes.

MR SCERRI: And so the verification can’t occur in this case, whereas it could occur if we were in real time and Dr Garcia-Pineda was working at NOAA and identifies something that has a 75 per cent confidence level, and the coastguard or whoever happens to have a plane or a helicopter or a ship nearby and they think it’s worth – and the weather is - - -

DR GARCIA-PINEDA: Yes.

MR SCERRI: - - - good weather and they think it’s worth risking someone’s life to go out and get a sample, or to have a look at it.

DR GARCIA-PINEDA: Yes, correct.

MR SCERRI: So isn’t that really what it boils down to, that you, sir, Dr Gundlach, you say there are possible features. In respect of some of those Dr Garcia-Pineda says they’re false positives.

DR GARCIA-PINEDA: Yes.

MR SCERRI: And others I give them a confidence rating.

DR GARCIA-PINEDA: All right.

MR SCERRI: And you, sir, Dr Garcia-Pineda, you don’t give a high confidence level a confidence level to anything close to the coast?

DR GARCIA-PINEDA: I didn’t do a confidence level – sir, can you repeat, please.

MR SCERRI: You don’t ascribe a high confidence level to any possible oil close to the coast.

DR GARCIA-PINEDA: That is correct.

MR SCERRI: And you disagree, and you say that the observations or the images that Dr Gundlach says are oil on the beach. You say they’re definitely not; they’re false positives.

DR GARCIA-PINEDA: That is correct.

406    However, at the end of the concurrent evidence on this topic, there was no greater bridging of the gap between the two experts on the interpretation of individual satellite images. What can be said is that, regardless of their differences in relation to the interpretation of the images, the experts agreed that, even though there were hundreds of satellite images potentially available for analysis, a limited number of them showed oil; a smaller number gave aerial coverage along the shorelines of Rote and Timor. Neither expert contended that satellite imagery alone could establish that features identified as possible oil (levels of confidence aside) were, in fact, oil. By the same token, they also agreed that it could not be concluded that oil was not present at the location at which an image was taken simply because a possible oil feature was not imaged. In short, the satellite images alone could not prove that oil was present or absent at a particular location.

407    While, in submissions, each party called in aid aspects of Dr Gundlach’s and Dr Garcia-Pineda’s evidence to support or, alternatively, to refute the proposition that oil from the Montara spill reached the coastlines of Rote and Timor in September/October 2009, sensibly neither party sought to elevate that evidence beyond its obvious limitations.

408    Before departing from this topic, I should record certain other criticisms of Dr Gundlach’s evidence not related to his interpretation of satellite imagery. Dr French-McCay, who was called by the respondent, and whose evidence in respect of oil spill trajectory modelling is discussed in the next section of these reasons, advanced two principal criticisms of Dr Gundlach’s analysis.

409    First, Dr French-McCay criticised Dr Gundlach’s use of winds and currents data obtained from published analyses of satellite data. She argued that this data was of low resolution and was not accurate near shore.

410    With respect to the QuikSCAT wind data, Dr French-McCay noted that Dr Gundlach himself had acknowledged that “(t)he analysis of wind flow does not work within a range of 15-30 km from the coastline”. Dr Hubbert, who was called by the applicant, and whose evidence in respect of oil spill trajectory modelling is also discussed in the next section of these reasons, also observed that QuikSCAT winds are not accurate within the 200 m bathymetric line near shore. Dr French-McCay said that Dr Gundlach’s predictions of oil movements within ~30 km of the Indonesian coastline based on QuikSCAT winds (either nearshore or assuming winds offshore to apply to the nearshore) were not reliable.

411    With respect to the IMOS currents data, Dr French-McCay said that the IMOS maps are based on geostrophic calculations derived from the slope of the satellite’s measurement of the topography of the ocean’s surface and the Coriolis effect of the rotation of the earth on motion across its surface. She noted Dr Gundlach’s acknowledgement that “these currents explicitly do not include tides nor are they accurate close to shore (inside the 200 m depth contour)”. She also noted the ~30 km between vectors in Dr Gundlach’s figures, which she said was of low resolution. Thus, Dr French-McCay said that Dr Gundlach’s predictions of oil movements within ~30 km of the Indonesian coastline based on IMOS currents were not reliable.

412    In summary, Dr French-McCay said:

… Dr Gundlach used winds and currents from satellite data that are of low resolution and not accurate near shore. He did not perform computer modelling using meteorological modeled winds or hydrodynamic modeled currents. Meteorological models solve physical equations and also “assimilate” (calibrate to) measurement data (e.g., temperature, pressure, wind speeds and directions, etc.). Hydrodynamic models include consideration of additional forces to the sea-surface pressure gradient as measured by the topography and the Coriolis effect (e.g., tides, bottom drag, wind stress on the surface, etc.). Thus, Dr Gundlach’s calculations are rough (of low accuracy) and not reliable for modelling potential shoreline impacts of Montara oil.

413    Dr Gundlach’s response was to say that he used winds and currents information derived from satellite data as a “first-approach”, which found that both winds and currents were favourable for oil movement towards Rote and Timor for much of September and October 2009. He said that while QuickSCAT data are not reliable for detailed spill forecasting, they do indicate a trend that influences oil movement. He also noted that no data from ground stations or current meters in the Montara area were offered as calibration of the hydrodynamic and meteorological models that Dr French-McCay had presented.

414    Secondly, Dr French-McCay criticised Dr Gundlach’s extrapolations based on the APASA modelling. In this connection, Dr French-McCay noted that most of APASA’s modelling comprised forecasts that had been undertaken in support of the response to the spill. She argued that hindcasts are considerably more accurate than forecasts (although I note that APASA also performed hindcasts on which Dr Gundlach explicitly relied).

415    More importantly, Dr French-McCay contended that it was not appropriate for Dr Gundlach to project a new trajectory for the oil based on APASA’s modelling, which was aligned with a source from a different location. To explain, in his first report, Dr Gundlach noted that the APASA modelling which he had considered had, in two places, displaced the centre of modelled oiling patches further eastward than actually observed by aerial surveillance. He therefore adjusted (in his words, “artificially moved”) the modelled results to obtain an “idea” of the possible oiling of Rote. By making this adjustment in the two cases, he observed that there was a likelihood of oil impact with Rote in both cases, with a timing difference of one day in each case.

416    Dr French-McCay explained:

21    … in Figure 5-35 of his report, Dr Gundlach asserts that a northwestward forecast trajectory APASA made, which passed to the southwest of Rote Island, infers that a patch of oil starting from a position farther northeast would move to Rote Island. This amounts to an inappropriate displacement of current and wind data underlying the trajectory. Currents and winds vary spatially and temporally, particularly near coastlines and where there is mountainous terrain on land. Thus, the transport between two points cannot be legitimately shifted in space or in time.

417    In short, Dr French-McCay’s criticism was that Dr Gundlach’s methodology relied on a form of extrapolation that could lead to error by implying currents that might not be possible under the laws of physics. She argued:

21    A far more accurate approach is to use hydrodynamic modelling, which is constrained to obey the laws of physics (i.e., mass and water flow momentum are conserved in the system such that water flow is continuous and water mass is not lost or gained).

418    In closing submissions, the respondent criticised Dr Gundlach’s evidence by relying on these particular criticisms from Dr French-McCay. Amongst other things, the respondent submitted that the QuickSCAT data have “a coarse time and space resolution” and the IMOS data are of “very low resolution”. Based on comments made in the Joint Report on Trajectory Modelling, the respondent submitted that all the modelling experts agreed that the satellite-derived winds and currents data are not reliable and should not be relied on. This submission somewhat overstates the experts’ opinion. Their opinion was not directed to the reliability in general of these data but of their reliability for use in modelling. This, however, is not how Dr Gundlach used the data. He used the data as part of an evaluative approach to addressing whether it was possible for oil from the H1 Well blowout to impact Rote and Timor.

419    The respondent criticised Dr Gundlach’s reliance on the APASA modelling on the basis that Dr Gundlach did not conduct the modelling himself, with the consequence, according to the respondent, that assumptions and equations used by APASA were not explained and not tested. The respondent also pointed to the fact that the purpose of APASA’s modelling was, essentially, to forecast using worst-case parameters and overestimations to provide search parameters, rather than to accurately predict the trajectory of the oil.

420    The fact that Dr Gundlach did not conduct the modelling himself is not a reason for criticising his reliance on it. I do accept, however, that APASA’s purpose in carrying out the modelling is relevant to be taken into account in assessing the weight to be given to the evaluative task that Dr Gundlach was carrying out. However, as I later explain, Dr Hubbert’s consideration of the APASA modelling trajectory analyses he examined indicated to him that they were unlikely to have identified all positions to which oil may have travelled following the H1 Well blowout. In other words, some of the APASA modelling most likely under-reported the locations to which the oil had drifted. No other expert examined these reports.

421    The respondent also criticised the extrapolation of APASA’s modelling that Dr Gundlach made. I accept that this could lead to error of the kind referred to by Dr French-McCay. But, once again, Dr Gundlach was not seeking to model oil trajectory. He was relying on a range of data sources with a view to looking for information and trends to explore, and express an opinion on, the possibility of oil from the H1 Well blowout impacting Rote and Timor. The utility of, and weight to be given to, that approach are matters I must consider, but I am not persuaded that Dr Gundlach’s evidence is so unreliable, as the respondent seems to contend, that I should reject his opinions out of hand.

422    I do note, however, that the experts agreed that the APASA modelling has been superseded by the evidence on modelling specifically undertaken for the purposes of this case. It is appropriate, therefore, to turn to that evidence. As the evidence reveals, the results of this modelling are dependent on the outputs of the particular hydrodynamic models that the respective experts employed.

The oil spill Trajectory Modelling Evidence

Introduction

423    An oil spill trajectory model is, in essence, a set of equations, solved by a computer, that quantifies, for given oil, its movement, and the processes that affect its physical characteristics and compositional changes, based on physical-chemical laws and other scientific information. The changes in oil location and properties are calculated over time and the projected trajectory (path) and concentrations of the oil are then mapped.

424    In his report, Dr Gundlach provided a convenient description of oil spill trajectory computer modelling:

4.2    Oil spill models define the primary movement of oil spilled on water based on winds and currents.

4.2.1    Wind and ocean current information is critical to determining the on-water transport of spilled oil. Currents are the dominant force, moving surface oil at the speed and direction of the current. Winds are a smaller force, moving oil at a commonly accepted rate equal to 3% of wind speed. The resultant movement of surface oil is the vector addition of the (100%) current speed and 3% of the wind velocity.

Subsurface oil (oil incorporated into the water column) is essentially not affected by wind speed and will move with the overall subsurface current flow of the area.

425    The ITF is of particular significance when discussing the influence of currents in the Timor Sea. I will return to discuss the evidence on the ITF in a later section of these reasons. For present purposes, it will suffice to understand that the ITF is a highly variable and complex flow of water from the Pacific Ocean that weaves through the Indonesian seas and out into the Indian Ocean.

426    Dr Gundlach continued:

4.3    Computer-based oil spill models assimilate current, wind, and oil characteristic data and propose outcomes that attempt to reflect the actual movement of spilled oil on the ocean surface.

The accuracy of computer-based models is dependent on many factors. As a comparison, weather models are used to predict the weather on an hourly, daily, weekly and monthly basis. Weather models use high-speed super computers to process data derived from numerous ground stations and satellite images, and still may provide a wrong forecast. The further out the forecast, the more prone it is to error.

We then can look at an oil spill model that has access to few ground wind stations and no direct current measurements within several tens if not hundreds of kilometres. In addition, the amount of oil spilled and evaporated is not accurately known. While modelling close to the wellhead may have some degree of accuracy, modelling tends to become increasingly inaccurate the further in time and distance from the wellhead. Small errors in movement at the beginning of the spill will become amplified over time and space.

Oil spill models essentially have two components: (a) spill trajectory that predict[s] the movement of oil and (b) oil spill weathering, which includes such factors as evaporation, emulsification, dissolution, sedimentation, and entrainment in the water column. Advanced spill models can predict both components in three-dimensions to reflect potential oiling within the water column as well as on the surface.

427    Oil spill trajectory models can be useful tools, but the results of their simulations cannot be taken as anything more than indicative of outcomes considered to be likely. Any kind of ocean modelling is an attempt to represent the behaviour of the real world as accurately as possible, but no mathematical model can be expected to simulate the complex physical processes in the atmosphere or the ocean with perfect accuracy. Modelled predictions must not be mistaken for a true representation of what happens in the real world.

428    Not all models are the same. They do not have the same degree of uncertainty. Moreover, they provide a limited representation of the behaviour of a continuous spill. One of these limitations is the representation of oil as collections of particles or “spillets”. These are model artefacts which, in the model, are released intermittently to represent the continuous flow of oil into (here) the ocean as it is affected by winds, ocean currents, and turbulence.

429    Errors can also arise from the resolution of the model not being sufficiently fine to resolve important features such as islands, channels, and complex coastlines. Take, for example, a model run on a regular grid with approximately n km between each grid point. Features which are not able to be represented by points every n km will not be well represented in the model. The modelling of currents close to the coast or along narrow channels are key examples of this concern.

430    Speaking with respect to grid points 12 km apart, Dr Hubbert, explained the position as follows:

78    Accurate calculations of the currents at a given grid point requires accurate calculations at surrounding grid points in order to calculate sea level gradients and the rate of change of the current speeds. As a result, calculation at the nearest grid point to the coast, which necessarily has a land point on one side of it, are not as accurate as calculations further out to sea. This becomes a significant issue when calculating currents along narrow channels because there are land points on either side of the sea points. It has been shown … that realistic currents are not able to be represented until the third grid point out from the coast (or river bank) and that a channel must be at least 7 grid points wide to obtain a reasonable representation of the flow along the channel. For the present case, with ocean model grid points 12 km apart, this means that the flow along channels less than 84 km wide will not be well represented and currents closer to the coast than 30 km will not be well represented.

79    An example of this is the channel between Timor and the Island of Rote which is approximately 12 km wide. …

431    Further uncertainty is introduced by the model inputs. Typically, the key inputs in oil spill trajectory modelling are: environmental forcing throughout the potentially affected region (such as wind speeds and directions, large scale ocean current speeds and directions, and tidal currents and directions); data concerning the oil itself (such as its physical and chemical properties, and the volume rate of the oil spilled over time); and information about the local environment (such as water depths (bathymetry) and regions of environmental concern (such as coral reefs, seagrasses, fisheries, and the like)).

432    It is also important to understand that these data can also be the outputs of other models. For example, one of the important underlying inputs in the trajectory modelling relied on in the present caseocean currents—was, in each case, calculated by a hydrodynamic model which was, in turn, dependent on winds calculated by a meteorological model. As to hydrodynamic models, Dr French-McCay explained:

29    … Hydrodynamic models use the laws of physics (i.e., conservation of mass and momentum) to calculate water movements and currents, based on mapped locations of coastlines, bathymetry (water depths), winds, inflows, outflows and water levels (heights) at the boundaries of the model domain due to tides and atmospheric pressures, and related information. As such, the hydrodynamic model provides comprehensive current data that is calibrated (adjusted) to observational data and is internally consistent by obeying physical laws (e.g., ocean non-tidal currents cannot run up on land). …

433    The point of present importance is that all these models differ, to some degree, from the real world. Thus, the oil spill model runs at the end of a chain of processes, all of which are subject to error that can accumulate along the chain.

434    Oil spill trajectory modelling is used for (a) stochastic (probabilistic) studies, which involve multiple model simulations to simulate a large number of random events over time to determine oil impact probabilities, and for (b) deterministic studies, which are case-specific studies involving real-time event forecasting.

435    Oil spill trajectory modelling is most commonly used by oil exploration and mining companies for stochastic studies undertaken for risk assessment purposes. These studies are generally required by government regulators for the purpose of assessing the likelihood of affection of significant environment features (such as reefs, seagrasses, and coastlines) by spilled oil. As I have discussed, they may also be used for the preparation of oil spill contingency plans.

436    In the event of a spill, oil spill trajectory modelling is also used to provide forecasts of where the oil will travel, to assist in response operations. Such modelling was carried out by APASA in real-time at the time of the Montara oil spill and, as I have already noted, Dr Gundlach relied on some information from these reports in his evidence. APASA also carried out hindcast modelling on which, as I have said, Dr Gundlach also relied.

437    In analysing the outcomes of oil spill fate simulations, it is essential to nominate a threshold oil concentration below which it is considered there is no (or no acceptable) risk to the marine environment. If a minimum concentration threshold is not defined, the results would show probabilities of hydrocarbon strike for miniscule concentrations over very large areas. This could lead to the misinterpretation of results.

Dr Hubbert’s modelling

438    Dr Hubbert reviewed five reports on oil spill trajectory modelling undertaken by APASA. Three reports (as identified by him, Reports 1, 2 and 3) were undertaken relatively shortly after the beginning of the Montara oil spill and contain information directly relating to the fate of the oil from that spill. Report 1 was a stochastic oil spill study. It showed a probability of 0% to 20% of oil impacting Rote. Reports 2 and 3 were deterministic studies. They did not show oil impacting Rote or Timor.

439    The other reports (identified by Dr Hubbert as Reports 4 and 5) were of a different character. Report 4 was an oil spill environmental risk assessment study which was carried out in 2010. It focussed on identifying the areas at potential risk of future spills from the Montara well. It was a stochastic study. Report 5 was a full oil spill contingency plan for all the respondent’s operations in the Timor Sea. It was a stochastic risk assessment for each operation carried out in 2013. It included the study of a sea bed blowout at the Montara wellhead platform.

440    Dr Hubbert was of the view that the oil spill trajectory modelling in each of the five reports suffered from shortcomings. He identified and discussed, in some considerable detail, the significant errors he saw in data input affecting all the reports, particularly Reports 2 and 3. He concluded that all the APASA modelling trajectory analyses in these reports were unlikely to have identified all positions to which oil may have travelled following the H1 Well blowout. Indeed, he said that the five reports, particularly Reports 2 and 3, could have significantly underestimated the locations to which the oil had drifted, such that APASA’s predictions lacked “real world credibility”. The respondent’s experts did not deal with APASA’s modelling in the reports that Dr Hubbert reviewed.

441    Dr Hubbert carried out his own modelling using the OILTRAK3D oil spill model and the GCOM3D hydrodynamic model. These are proprietary models associated with GEMMS of which Dr Hubbert is Managing Director and Head of Oceanographic Studies. The two models are run sequentially, with output currents from the GCOM3D model being used as input to the OILTRAK3D model.

442    OILTRAK3D is designed to simulate the fate of particular hydrocarbon types when spilled into the marine environment. It calculates the spreading, evaporation, entrainment, dissolution, transport, and stranding of defined oil types over time, taking account of prevailing wind conditions, water currents, temperatures, and sea conditions.

443    GCOM3D is part of AMSA’s search and rescue operational system. Dr Hubbert’s evidence was that it is run almost daily to determine search areas for marine incidents in the Australian region and neighbouring countries.

444    In Dr Hubbert’s modelling, ocean currents and sea levels were modelled with GCOM3D using large scale currents and tides derived from the US Navy Global Coastal Ocean Model (NCOM). Atmospheric pressures, and wind speeds and directions, were derived from the BoM’s high resolution atmospheric model called the MesoLAPS Mesoscale Limited Area Prediction System (MesoLAPS). This is a hindcast model which outputs wind speeds and atmospheric pressures every 3 hours on a 0.125° grid (approximately, 12 km). Dr Hubbert compared the BoM winds against data from Browse Island. He said that the model winds showed excellent agreement with the observed data.

445    Dr Hubbert obtained his data on oil properties, composition and weathering characteristics from the data that had been used in the APASA modelling.

446    Dr Hubbert’s simulations were carried out for the full period of the Montara oil spill, treating the spill period as 74 days. Four scenarios were investigated based on daily spill rates of 400 bbl/day; 800 bbl/day; 1,200 bbl/day; and 2,500 bbl/day. He plotted the impacts of crude oil > 0.1gm/m2 in various Figures included in the report (Figures 24 to 83).

447    The modelling showed oil reaching the shores of Rote in early September 2009 and in late October 2009. The two most notable periods when this occurred were 9 to 13 September 2009 and 25 to 31 October 2009. Subsequent predictions, which introduced the application of dispersants to the oil (assumed to have particular degrees of efficacy) still showed oil reaching the shores of Rote in late October 2009. Adopting certain assumptions, Dr Hubbert concluded that the application of the dispersants would have resulted in the concentrations of oil being reduced slightly from the results without dispersants, and to spread the oil over a slightly larger region.

448    Dr Hubbert found that the differences observed between his modelling and the APASA modelling were (what he said was) the excessively high current speeds in the ITF in the APASA modelling—which led APASA to conclude that the fast-moving currents moved weathered oil patches some significant distances south-westwardly from their source—and the fact that, according to him, the APASA modelling did not allow for potential errors in the wind and current fields.

449    Based on his modelling, Dr Hubbert expressed his belief that:

239    … it is highly likely that oil from the Montara Oil Spill impacted on the Regency of Rote.

240    Whilst it is also likely that oil reached Kupang it is beyond the scope of the APASA modelling or the modelling I have undertaken to derive firm conclusions on this matter because of the resolution of the large-scale ocean currents (12 km). The channel between Rote and mainland Timor through which the oil would have to pass is too narrow for the modelling on a 12 km grid to be accurate.

Dr French-McCay’s modelling

450    Dr French-McCay carried out modelling using the Spill Impact Model Application Package (SIMAP) oil spill model, developed by the professional services firm simply called RPS (formerly, Applied Science Associates). Dr French-McCay commenced employment with Applied Science Associates in 1984. Currently, she is the Director of Research & Model Development at the Ocean Science office of RPS at Rhode Island in the United States of America.

451    Dr French-McCay’s evidence was that, under her direction, SIMAP has been “developed and thoroughly tested over more than three decades” and that it is “internationally accepted as the most advanced state-of-the-art model for calculation of oil transport, fate, exposure, and biological effects of spilled oil”.

452    SIMAP sums oil movements from currents and wind data (which accounts for most of the oil movement) with movements by turbulence and buoyancy. The latter two components are calculated within the model itself. In undertaking her modelling, Dr French-McCay used oil property and compositional data provided by Dr Stout, and currents, water temperature, and salinity data provided by Professor Ivey, who had, in turn, derived this data from, or used it in, the Stanford Unstructured Nonhydrostatic Terrain-following Adaptive Navier-Stokes Simulator (SUNTANS) hydrodynamic model. SUNTANS was developed at the Environmental Fluid Mechanics Laboratory at Stanford University.

453    Dr French-McCay modelled the trajectory from 21 August to 10 December 2009. In doing so, she assumed that oil was released from the Montara H1 wellhead platform at the rate of 400 bbl/day. She also assumed that the spill into the water was purely oil. She was instructed to assume that dispersants applied to the oil were 80% effective. The basis for this assumption is not clear to me.

454    At her request, Dr French-McCay was provided with, and consulted, Dr Garcia-Pineda’s interpretations of the satellite imagery, as an aid to determining (for herself) whether her modelled results were reliable.

455    Based on her modelling, Dr French-McCay concluded that much of the oil mass from the Montara oil spill would have evaporated rapidly to the atmosphere since the oil was a light crude and the weather was warm with light breezes. She said that the oil that did not evaporate and was not effectively treated by dispersants, would have been primarily on the water surface and would not have mixed into the water given the light winds and the absence of breaking waves.

456    In her primary report, Dr French-McCay said of her modelling:

4    During August 2009, the modeled floating oil initially remained close to the Wellhead Platform (WHP), within 35 km of it until August 28, and then moved east and north to about 70 km from the WHP by August 31. During September 1-13, the modeled floating oil stretch up to 140 km towards the southeast and 160 km to the northwest of the WHP (which is 90 km from the coast of Rote Island). During September 14-25, the closest floating oil was 35-65 km and the closest residual oil was 37-73 km from the coast, respectively. After September 25, most of the floating oil and subsurface oil was further south (>85 km from Rote Island) and only occasionally did patches of oil or waxy residuals pass within 45-85 km of the Indonesian coastline. In all months, oil and residuals that reached <85 km of the coast of Rote Island were caught up in the westward Indonesian Throughflow current (which extends ~85 km from the Indonesian coast … in the area of the Timor Trench …) and were swept to the southwest into the eastern Indian Ocean. Over time, the oil was dispersed throughout the Timor Sea and eventually carried by currents westward into the Indian Ocean.

457    Dr French-McCay also said:

18    The closest that floating oil came to the Indonesian coastline in my (base case) model simulation was ~35 km to the southeast of Rote Island (on 18 September). The closest that highly-weathered waxy residual oil came to the Indonesian coastline during August to November 2009 in my base case model simulation was ~32 km to the south of the Indonesian shoreline. In the period of 6-10 December, one spillet representing ~28-26 kg of highly-weathered waxy residual oil approached the coastal waters near Rote Island. At the time, it had weathered for 77 days at sea, so was comprised of microscopic particles of very highly weathered residual oil at concentrations of ~0.03 mg/m3.

458    Further, Dr French-McCay said:

13    Based on the modelling I performed, and comparisons to observations and remote sensing, neither oil nor dispersants from the Montara spill reached the seaweed cultivation areas along the coast of Nusa Tenggara Timur. The quantities of oil, and any dispersants carried with it, that passed within ~30-40 km of the coastline (the closest approach) were very small, highly weathered, and in extremely low concentrations. Regardless of the quantity, any oil or dispersant carried with oil reaching the area of the ITF (i.e., within 85 km of the Indonesian coast) would have been carried southwestward by the ITF, away from Rote Island.

459    Dr French-McCay argued that her model predictions are supported by the observations of responders in the field at the time. She said that the closest observations to Rote made by responders or identified by AMSA were patches of sheen observed about 30 to 40 km southeast of Rote. She also remarked that the AMSA records also reported on observations of natural phenomena which, she suggested, might also have occurred in Indonesian nearshore waters.

460    These remarks must be viewed in the context that AMSA’s observations did not involve incursions significantly into Indonesian territory. Flight surveillance records at the time show that aircraft changed tracking to miss Indonesian-controlled airspace. AMSA’s principal focus was on Australian territorial waters, in particular Australia’s northern coastline and the area of Ashmore and Cartier Islands to the west of the H1 wellhead platform. Thus, the absence of observations of oil by AMSA in areas close to the Indonesian coastline is explained by the fact that this area was not under close surveillance. Therefore, care must be exercised in accepting the contention that the evidence of AMSA’s observations supports Dr French-McCay’s modelling results. In a later section of these reasons I also deal with the evidence adduced by the applicant which compares Dr French-McCay’s modelled predictions with actual observations in AMSA flight surveillance reports, maps, and other records that were made at the time of the oil spill.

461    Dr French-McCay also argued that her modelling results were supported by “interpretations of satellite imagery”. By this, Dr French-McCay appears to have included her own interpretations of the satellite imagery:

6    Based on the modeled trajectory and sensitivity analyses, I evaluated whether “possible oil features” identified by Dr. Garcia-Pineda in remote sensing-based images could have been oil from the Montara spill, as they could have been other phenomena or other oil releases from a different source than the Montara WHP. The polygons denoting “possible oil features” are locations where it is possible that some amount of oil was present. This does not confirm oil was actually present, the source of any oil present, or the amount and patchiness of oil if present. There were varying degrees of confidence associated with the areas identified by Dr. Garcia-Pineda … as “possible oil features” in the images. …

462    Dr French-McCay went on to discount “possible oil features” interpreted from certain satellite images:

7    … I do not believe the LANDSAT and MODIS interpretations of “possible oil features” are in all cases accurate or reliable indicators of Montara oil. I think that some of the areas identified in the LANDSAT and MODIS imagery could be the Montara oil, or more likely contain patches of oil, but other sources of oil and causes for these features were likely present in those data sets, compromising their reliability.

463    She then expressed what seems to have been her preference for other satellite images which, she said, were in agreement with her model predictions that no oil reached the Indonesian coast. Even then she said:

8    Possible oil features were identified by Dr. Garcia-Pineda … in six SAR images from September and October of 2009 in areas that were >35 km from Rote Island. Based on the modeling, some of those features may not have been due to Montara oil and none of the features identified would have subsequently moved to the Indonesia nearshore.

464    There are a number of comments that should be made about this evidence.

465    First, for the purposes of this case, Dr French-McCay was not qualified as an expert in the analysis and interpretation of satellite imagery. She was qualified as an expert in oil spill trajectory modelling. Indeed, in a report responding to Dr Gundlach’s primary report and Dr Hubbert’s primary report, Dr French-McCay made clear that the interpretation of satellite imagery was not her area of expertise.

466    Secondly, Dr French-McCay’s stated reason for referring to the satellite imagery examined by Dr Garcia-Pineda was to compare her modelling results with the interpretations given to that imagery by Dr Garcia-Pineda, as a check on the reliability of her results, not as a stepping stone to proffering alternative or possible interpretations of the imagery.

467    Thirdly, in proffering her interpretations of the satellite imagery, Dr French-McCay was seeking to do what Dr Garcia-Pineda did not do—interpret the imagery by reference to ancillary data and in situ observations. However, in doing this, she adopted a means that Dr Garcia-Pineda unequivocally rejected—the use of a computer model (in this case, SIMAP) as verification. According to Dr Garcia-Pineda, a numeric simulation model (i.e., a computer model) should never be considered as a way of verifying satellite imagery.

468    Fourthly, there is an element of circularity involved in Dr French-McCay’s self-appraisal of the accuracy of her modelling: she relied on some images that did not identify any possible oiling features; she discounted as unreliable some images that showed “possible oil features”; and, importantly, she interpreted other images according to her own modelled results. Having done that, she reached the overall conclusion that her modelled results were consistent with the interpretations of the satellite imagery.

469    In a report responding to criticisms of his analysis by Dr French-McCay (which I have discussed above), and in the concurrent evidence session dealing with trajectory modelling, Dr Gundlach demonstrated, persuasively, that, in fact, Dr French-McCay’s modelling was not in agreement with Dr Garcia-Pineda’s analysis of the satellite imagery, including images in respect of which Dr Garcia-Pineda had expressed high confidence that oiling was present.

470    Dr French-McCay estimated the degree of uncertainty of her model in the transport of oil in offshore areas of the Timor Sea to be about 50 km, taking into account Dr Garcia-Pineda’s high confidence interpretations. She expressed her expectation that her model would have less uncertainty in nearshore areas, and also across the breadth of the ITF in the area of the Timor Trench south of Rote. Dr French-McCay said that transport in the ITF would be much less than 50 km because of “the certainty that the ITF exists and flows westward into the Indian Ocean”.

471    Dr French-McCay found that her trajectory results were not sensitive to the assumptions she was instructed to make regarding the use and effectiveness of dispersants, or the recovery of spilled oil. Further, although her modelling assumed an oil release rate of 400 bbl/day (63.6m3/day), her results were not dependent on that assumption. She said:

14    … If the oil release rate were higher or lower than I assumed, the spillets representing the oil would still have been transported to the same locations. Each spillet would simply have represented more or less oil mass. Thus, the concentrations of oil and waxy residuals represented by the spillets would vary proportionately with the oil release rate, but the trajectory (locations the oil moved, both floating on the surface and entrained in water) would not change. As such, my opinions and conclusions with respect to the oil trajectory would not change if I assumed a different or time-varying volume spill rate.

Criticism of Dr Hubbert’s modelling

472    Like many other areas of the expert evidence in this case, the disagreements on the oil spill trajectory modelling were legion and, ultimately, remained substantially unresolved between the experts.

473    Dr French-McCay contended that Dr Hubbert’s modelled simulations were not realistic or credible. She advanced four reasons.

474    First, Dr Hubbert’s hydrodynamic model (GCOM3D) was run as a barotropic model rather than as a baroclinic model. This is, perhaps, an oversimplification of the position, as I will later explain. Dr French-McCay argued that a baroclinic model was the more appropriate model to use for oil spill trajectory modelling in the present case.

475    To explain, a baroclinic model is a numeric model that incorporates as many of the physical processes operating on the ocean as is possible, such as winds, tides, and the thermodynamic processes within the ocean. As to the latter, a baroclinic model recognises that pressures within the ocean can vary with both depth and local water density. Water density is a function of water temperature and salinity. Recognition of these processes requires a connection with the lower atmosphere to simulate the heat fluxes between the atmosphere and the ocean. A barotropic model leaves out these thermodynamic processes. It assumes that the ocean is well-mixed and relies on a single value for density. It proceeds on the basis that the pressure inside the ocean varies only with depth.

476    Dr French-McCay argued that, in undertaking his modelling, Dr Hubbert did not consider the effect of temperature and salinity variations across the Timor Sea and the ITF. Dr French-McCay said that ocean currents such as the ITF are density-driven (due to spatial variations in water temperature and salinity) and can only be simulated using a three-dimensional baroclinic model. In substance, Dr French-McCay’s contention was that, in his modelling, Dr Hubbert had not considered the major current in the area—the ITF—which, in her view, would have carried the spilled oil westwardly away from the Indonesian coast.

477    In advancing this contention, Dr French-McCay deferred to comments made by Professor Ivey, who was the expert called by the respondent on the topic of currents. Given its significance to the modelling that was undertaken, I deal separately with the topic of currents in a later section of these reasons. Nonetheless, for the purpose of discussing oil spill trajectory modelling and commenting on Dr Hubbert’s modelling in particular, Dr French-McCay said:

43    Based on my decades of experience modeling oil spills in the oceans around the world, I would not depend on a hydrodynamic model run in barotropic mode for current data when modeling a spill in oceanic waters such as the Montara oil spill. Barotropic models are much less accurate than baroclinic models, and do not predict ocean currents such as the ITF. Barotropic models are typically used for near-shore coastal regions where tidal flows dominate and water density-driven flows are weak.

44    Therefore, I also agree with Dr Ivey’s conclusion and I do not consider the model used by Dr. Hubbert to be capable of predicting the ocean currents in the Timor Sea with any accuracy, particularly not in the area of the ITF (i.e., within 100 km of the Indonesian archipelago including Rote Island). Using his model, Dr. Hubbert underestimated the westward current transport near Indonesia.

478    Secondly, and relatedly, Dr French-McCay contended that the currents from GCOM3D directed to Rote appeared to be unrealistically high given that, in her view, winds (which were directed southeastwardly and away from Rote) could not account for oil being carried there.

479    On the question of winds, Dr French-McCay noted that Dr Hubbert used only the BoM’s wind models and had not performed a sensitivity analysis using various wind fields, including those taken from the models provided by the European Centre for Medium Range Weather Forecasts (ECMWF), which Dr Hubbert had also identified, in his primary report, as an appropriate wind data source for modelling the Montara oil spill.

480    Dr French-McCay evaluated and compared wind data from five meteorological data sets, which included the BoM’s wind models (including MesoLAPS) and the ECMWF. Dr French-McCay found that the BoM winds were consistently of higher speed than the four other wind models (which she found to be in closer agreement), so much so that she classified the BoM winds as an outlier. She also noted what she considered to be “spurious data” in Dr Hubbert’s wind data set (a sudden strong westerly wind sustained in the data set over 6 hours), which she suggested could be due to a post-processing error, as opposed to being present in the original BoM models. Dr French-McCay noted further that the BoM winds were primarily from the southwest in August and September 2009 at the spill site, whereas the ECMWF showed variable or southeast winds in the period, and the other three models (which were United States’ models) showed variable wind directions. Whilst noting the primacy of ocean currents in transporting oil (given the light winds during the spill period), Dr French-McCay concluded:

54    The implication of the stronger wind speeds and prevailing southwesterly winds (from the southwest) in the BoM winds used by Dr. Hubbert is that in the oil spill model (OILTRAK3D) the BoM winds would push the oil farther to the east and northeast than any of the other wind models.

481    To reinforce her criticism of Dr Hubbert using only the BoM winds, Dr French-McCay noted that she considered the sensitivity of her oil trajectory model results to the choice of wind model used as part of her uncertainty analysis.

482    As to the strength of Dr Hubbert’s modelled currents, Dr French-McCay used the Figures in Dr Hubbert’s report to observe that his model showed oil moving north-westwardly at about 75 km/day or 0.9 m/s (1.7 knots) in the period 11 to 13 September 2009. Assuming that the BoM winds carried the oil toward Rote (as Dr Hubbert had said), Dr French-McCay analysed the available wind data and concluded that wind drift could not account for this transport rate: the winds would have been blowing consistently from the south-east over this period at speeds typical of, and exceeding, those in tropical cyclones. Dr French-McCay observed that these speeds were not shown in any of the five wind models (including the BoM models) to which I have referred, and could not be accounted for in other data (QuikSCAT winds).

483    Dr French-McCay concluded that the currents predicted by Dr Hubbert’s modelling must have been the driver for the oil’s movement towards Rote on 11 to 13 September 2009, as depicted in his Figures. However, this would mean that, taking into account wind-induced drift, GCOM3D had predicted a current averaging about 1 m/s (2 knots) from the site of the spill towards Rote, which crossed the ITF and lasted for several days during the 11 to 13 September 2009 period. Dr French-McCay argued that this was not a reasonable result because this current would be four times the average speed of the ITF, which is about 0.25 m/s (0.5 knots).

484    Thirdly, Dr French-McCay observed that the oil transport and fate algorithms in OILTRAK3D were undocumented and, in her view, could be contributing to errors in Dr Hubbert’s modelled results. She criticised what she saw as Dr Hubbert’s failure to describe how oil fate processes, such as weathering and entrainment, were applied in the model. She also criticised what she saw as Dr Hubbert’s failure to run sensitivity analyses—varying the input assumptions over at least several model runs—to estimate uncertainty. She remarked that Dr Hubbert appeared to have undertaken a single model run for each case (i.e., the assumed spill volume rate) and had not undertaken stochastic modelling (undertaking multiple model runs varying the assumed inputs within a range of possibilities).

485    Fourthly, Dr French-McCay argued that OILTRAK3D had not been validated or compared to observational data from the spill to support Dr Hubbert’s results. Dr French-McCay noted that Dr Hubbert had provided no comparisons of his modelled results to any oil observations, either from responders or from remote sensing data (satellite imagery) to evaluate the accuracy of his model (as she had done in respect of her modelling).

486    Dr French-McCay also contended that Dr Hubbert’s depiction of oil impacts in the Figures accompanying his primary report made it appear that the oil spreads further when a higher volume rate of oil is spilled. Dr French-McCay observed that, in the Figures with lower spill rates, oil concentrations fell below the threshold (> 0.1gm/m2) at the edges of the mapped distribution. In the Figures with higher spill rates, concentrations above the threshold were mapped. According to Dr French-McCay, this led to the visual effect that there would have been a larger area of oiling when Dr Hubbert’s model was run with the higher spill rates. Dr French-McCay said that, in reality, this larger area was not indicative of “spreading”, just that higher concentrations of oil were present in the same locations. Dr French-McCay argued that the oil movements calculated in Dr Hubbert’s model should not change with the volume of oil spilled. This is because oil transport cannot be a function of oil volume, given that Dr Hubbert used the same hydrodynamic and wind model for all cases he modelled, and had stated that wind drift is not related to oil volume. Dr French-McCay said that if Dr Hubbert had mapped all concentrations, without a threshold, the mapped areas with > 0 mass would have been the same.

487    Dr French-McCay also questioned the amount of oil depicted in Figures 24 to 47 of Dr Hubbert’s report. Based on the digitisation of Dr Hubbert’s Figures 24 to 32, and making certain calculations, Dr French-McCay argued that Figures 24 to 32 in Dr Hubbert’s report depicted at least 8 to 28 times more oil than was actually spilled (as assumed for each case run in his model). This led Dr French-McCay to conclude:

70    Either Dr Hubbert’s model has an error in the mass balance or possibly the modeled oil amounts have been contour-mapped in the figures (such that patchy discontinuous oil distribution were filled in and smoothed in space using his mapping program), which essentially creates more oil than is actually present and misrepresents the spill’s impact.

488    Finally, Dr French-McCay undertook additional runs of her model (SIMAP) to test Dr Hubbert’s propositions that he used the most accurate wind data available, and that his modelled oil impacts moved closer to Rote when a larger volume of oil was assumed (remembering that Dr Hubbert carried out simulations from 400 bbl/day to 2,500 bbl/day whereas, in her base model, Dr French-McCay assumed a spill rate of 400 bbl/day, as she had been instructed).

489    Dr French-McCay found that using the BoM winds in her model produced results that were similar to those produced by the four other wind data sets to which I have referred. In none of the simulations with any of the five wind models did oil approach or reach the coastal areas of Indonesia.

490    Dr French-McCay performed additional model runs using the highest spill rate assumed by Dr Hubbert—2,500 bbl/day—both with and without the inclusion of response activities (dispersant use and oil recovery). She said:

77    My results demonstrate that if the oil release rate were higher than I assumed, the spillets representing the oil would still have been transported to the same locations and each spillet would simply have represented more oil mass. Thus the concentrations of oil and waxy residuals represented by the spillets would vary proportionately with the oil release rate, but the trajectory (locations the oil moved, both floating on the surface and entrained in water) would not change. As such, my opinions and conclusions with respect to the oil trajectory would not change if a different volume spill rate were assumed. …

Dr Hubbert’s response

491    Dr Hubbert made a number of responses to Dr French-McCay’s criticisms of his modelling.

492    Dr Hubbert said that GCOM3D was run in barotropic mode for two reasons. First, the methodology he has developed for incorporating large scale ocean influences with local wind and tidal effects is sound, as it has been run repeatedly over 30 years in hundreds of projects producing reliable results. Secondly, Dr Hubbert said that his methodology is similar to the way in which GCOM3D has been run at AMSA for search and rescue purposes with great reliability for the past 20 years. In short, Dr Hubbert said that he thought that running GCOM3D in barotropic mode was appropriate for the task at hand. The dominant forces on surface oil slicks are derived from winds, the tide, and large scale ocean currents, all of which (Dr Hubbert said) were incorporated in his simulations. I will return to the characteristics of GCOM3D when dealing with the topic of currents later in these reasons.

493    With respect to the oil concentration threshold used in his modelling, Dr Hubbert referred to the fact that his primary report was directed, in part, to commenting (as he was instructed to do) on the reliability of APASA’s modelling, including that undertaken at the time of the Montara oil spill. Dr Hubbert said that the oil concentration threshold he adopted was the same as that adopted by APASA. It was used in his modelling in order to make valid comparisons between his modelling and APASA’s modelling. I note, incidentally, that Dr French-McCay agreed that this threshold was, in any event, objectively reasonable.

494    With respect to Dr French-McCay’s comments on Dr Hubbert’s mapping as it reflected the different trajectories of oil released at different and greater spill rates, Dr Hubbert said:

33    To claim that the fate of larger volumes of oil spilled is the same as smaller volumes suggests that Dr French-McCay’s modelling does not account for non-linearity in the movement of oil droplets and that it also has a very poor representation of oil dispersion. Obviously the more oil droplets available, the broader the range of outcomes from the non-linear dispersive processes.

495    With respect to Dr French-McCay’s observations on the wind data used by Dr Hubbert, Dr Hubbert said that it was important to note that Dr French-McCay’s use of the BoM winds in her model (SIMAP) produced results that were similar to her base model using other wind data. Dr Hubbert observed that the corollary of Dr French-McCay’s additional modelling results was that the choice of wind source data does not appear to explain the major differences in the modelling undertaken with OILTRAK3D and SIMAP (or, for that matter, the modelling undertaken by APASA using the OILMAP model). Dr Hubbert said that Dr French-McCay’s statement that she could not account for oil reaching Rote since the winds were directed southeastwardly was “simply not correct”. As to his choice of the BoM winds and Dr French-McCay’s criticism that, in his modelling, Dr Hubbert did not also use (at least) the ECMWF winds, Dr Hubbert said:

47    … I routinely use ECMWF winds but not the winds used by Professor Ivey which are too coarse in their spatial and temporal resolution.

    There is rarely a significant difference between the high resolution ECMWF winds and the BoM high resolution model winds as both model winds use the same physics package and the two centres collaborate very closely. Accordingly, I use the BoM model winds in the Australian Region and the ECMWF winds in other parts of the world.

496    With respect to Dr French-McCay’s criticism that Dr Hubbert had failed to describe how oil fate processes, such as weathering and entrainment, were applied in his model, Dr Hubbert said that he did not think it necessary to offer such a description because it was an assumed input made to align his inputs with those used in APASA’s modelling (which he was instructed to analyse).

497    On the question of currents, Dr Hubbert said that his model did consider the effect of the ITF and that Dr French-McCay’s statement that the ITF was not taken into account was an apparent misunderstanding of his modelling methodology. Further, Dr Hubbert said that GCOM3D did not “predict” the ITF. Rather, it took that information from the NCOM model. With respect to that model, Dr Hubbert said:

66    One of the features of the large scale currents derived from NCOM is that the region south of Rote often exhibits eddies which can in fact drive oil to the northeast or north against the general flow of the ITF.

The ebb tide can also carry oil towards the west as it is following the relative movement of the moon to the earth.

498    On this question, Dr Hubbert also said that the wind calculations that Dr French-McCay performed to illustrate the proposition that “something other” than the BoM winds was moving the oil towards Rote in Dr Hubbert’s simulations, were “not applicable”. This was because, in Dr Hubbert’s view, “something other” than the BoM winds was involved. This “something other” included the tides and the large-scale ocean flows that Dr Hubbert used.

499    As to Dr French-McCay’s calculation of the strength of the current that Dr Hubbert’s simulations predict for the 11 to 13 September 2009 period, Dr Hubbert said:

69    Dr French-McCay’s calculation of current averaging is incorrect. This is because oil was being continuously released from the Montara Wellhead and an accumulation of oil at a given location can consist of oil released at many different times and having travelled many paths. The oceanic conditions can contribute to oil concentrations dropping below the defined threshold and then joining up with new oil to generate concentrations above the threshold.

500    With respect to Dr French-McCay’s comment that Dr Hubbert had not addressed uncertainty in his modelling, Dr Hubbert said that Dr French-McCay had not accounted for his comments in his primary report which explained that error was explicitly accommodated in his modelling by applying Gaussian distributions (bell curves) to the input data. Dr Hubbert said that this was the method he developed 20 years ago when setting up AMSA’s search and rescue system.

501    To explain further, Dr Hubbert said that, in order to provide some mitigation against errors in inputs to OILTRAK3D, random errors are generated into its model simulations. Multiple simulations are then run for each spillet, with each simulation applying randomly chosen “reasonable” errors individually to the wind speed, wind direction, current speed, and current direction. The errors are derived by applying a Gaussian error distribution to each variable. In this manner, the simulation no longer produces a single prediction of the fate of an individual oil spillet, but predicts a number of locations where the spillet might be, depending on the errors in the inputs. This method feeds into an analysis of the regions likely to be impacted by oil and enables a calculation of the probable maximum oil concentration.

502    Dr Hubbert also explained that it is logically impossible to calculate accurate concentrations of oil impacting a given region from the results of oil spill trajectory modelling (which represents oil as a collection of particles) for two main reasons: the release of particles at intervals is a poor representation of a continuous oil spill at the wellhead; and the problem of trying to determine oil slick concentrations in the ocean from a limited number of particles.

503    With respect to the latter problem, Dr Hubbert explained that oil spill model particles drift across the ocean. At a given time, they are dotted across large areas, with each particle carrying a given volume of oil. To determine concentrations, the volume of oil per unit area (not per particle) needs to be calculated. But what area is to be used? If the analysis is carried out on a grid, then grid cells with small dimensions will have no particles in them and show, therefore, zero concentrations of oil. However, the cells that do have particles in them are unlikely to represent the full spread of a slick and will likely overestimate the concentrations of oil in those cells. If the size of the cells is increased, then more particles will be included within them. However, this may also include patches of ocean which, in the real world, are not impacted by oil. In this case, the concentration of oil is likely to be underestimated, and also show oil impacts where in fact there were none. In this context, Dr Hubbert stressed that, in simulating oil spills, the object is to simulate trajectory, not to model oil concentrations as such. He said that OILTRAK3D does not attempt to predict oil impact concentrations but, simply, the likely highest concentrations which may impact a region.

504    It is in this context that Dr Hubbert addressed Dr French-McCay’s questioning of the amount of oil depicted in his Figures, and the fact that the amount of oil depicted appeared to be in excess of the assumed amount of oil spilled. Dr Hubbert said:

Table 3

70    I agree that these points are essentially true. The combined effects of including randomised error and contouring the results can lead to the result reported by Dr French-McCay.

Given that it is impossible to accurately forecast exactly where an oil spill will go the purpose of such contouring is to show the likely concentrations which may be found in regions of the ocean. To put it more simplistically, if I am swimming in the ocean and an oil slick passes 100 metres away I would have to assume that it could have engulfed me and I was lucky that the exact environmental forcing conditions prevailed such that it missed me.

Criticism of Dr French-McCay’s modelling

505    Dr Hubbert argued that Dr French-McCay’s report conveyed the impression that the predictions generated by her model were actual outcomes that occurred following the Montara oil spill, noting that her model trajectories were, in fact, based on the simulations of four mathematical models (BRAN (as to which, see below), SUNTANS, ECMWF and SIMAP).

506    He observed that Dr French-McCay’s report included no discussion of potential errors in the model inputs. He instanced, in particular, a lack of discussion on the SUNTANS modelling of currents, which Dr French-McCay accepted as accurate. In particular, Dr Hubbert said that Dr French-McCay should not have accepted hindcast monthly averages of surface currents (as Professor Ivey had done) as adequate verification of that data for use in oil spill trajectory modelling. Dr Hubbert said that Dr French-McCay should have attempted to verify hourly predictions because tidal currents can vary significantly in hours.

507    To exemplify what Dr Hubbert said was Dr French-McCay’s failure to differentiate between her modelling and real world outcomes, he pointed to the fact that, on proper analysis (having regard to the resolution of the grid of currents driving the SIMAP model), Dr French-McCay’s modelling showed that the Ashmore and Cartier reefs and islets were enveloped in oil. This did not correlate with the actual observations made as part of a five day shoreline assessment which was carried out on behalf of AMSA and completed on 25 October 2009, or from a subsequent longitudinal study carried out by the respondent in 2013 of bird life, both of which reported that there was no visible evidence of oil impact in that area. Dr Hubbert said that these findings indicated that Dr French-McCay’s modelling was “in serious error” and that “no useful conclusion” could be made from her studies as to where the oil drifted.

508    With utmost respect, I think that Dr Hubbert’s observations concerning the confidence with which Dr French-McCay presented her findings are well-made. An incautious reader of her reports might not appreciate that, when Dr French-McCay was speaking of her modelled results, she was actually speaking in the realm of prediction, not the realm of fact. For example, in her primary report she said that her model was “definitive” as to where oil from the Montara oil spill could have been transported. I do not accept that Dr French-McCay’s modelling or, for that matter, any of the other modelling, could be or was definitive of the transport of Montara oil during the time of the spill. The best that each model could do was to predict, by hindcasting (or, in the case of Reports 1, 2 and 3 of the APASA reporting, forecasting) what the trajectory of the oil might have been. Plainly, much depends not only on the accuracy of the model and the methodology used to run it, but also the accuracy of the data used to carry out the simulations—hence the appropriateness of Dr Hubbert’s warning about the potential for, and effect of, cascading errors in a given model’s results.

The Modelling of Ocean Currents

Introduction

509    In light of the criticisms that had been made of his modelling, Dr Hubbert modelled the Montara oil spill with OILTRAK3D for a three month period using the ocean currents data that Dr French-McCay used in SIMAP—namely, the SUNTANS data provided by Professor Ivey. When he did this, Dr Hubbert found that the simulation produced modelling results which he described as “very similar” to the results that Dr French-McCay obtained. In particular, Dr Hubbert found that OILTRAK3D produced a trajectory in which oil did not impact the coastal waters of Rote but did impact the Ashmore and Cartier Reefs. This led him to conclude that the major differences between his work and Dr French-McCay’s work was not due to the specific model used (OILTRAK3D v SIMAP). Rather, Dr Hubbert reasoned that the modelled results were a function of differences in the ocean currents and/or the atmospheric winds used in the respective models.

510    However, as to atmospheric winds, Dr Hubbert noted that Dr French-McCay had investigated a range of wind sources, including the BoM winds that he had used, only to find that similar results were produced—no oil was predicted to enter the coastal waters of Rote. Given that the source of the winds data seemed to make little difference to the results, Dr Hubbert reasoned that the ocean currents, and particularly the accuracy of the representations of the ITF, were the more significant drivers of the differences observed in the respective oil spill model projections.

511    Dr Hubbert said:

30        … it is my opinion that the speeds of the ITF input to the Dr French-McCay oil spill model are excessive and have resulted in oil movements being erroneously modelled as unlikely to have reached the coastal waters of Rote. Instead, once the SUNTANS predictions are adopted the oil spill is assumed to have been swept to the southwest.

31    The above opinion is starkly illustrated by the simulations I undertook with OILTRAK3D. In the case provided in my original report … where OILTRAK3D was driven by currents from GCOM3D (which derives its information for the ITF from NCOM), oil was found to enter the coastal waters of Rote. Using the exact same method, except, using the SUNTANS currents, my recent studies showed oil being swept to the southwest and avoiding impact on the coastal waters of Rote.

512    Dr Hubbert’s opinion was that the SUNTANS current speeds in the Timor Passage appeared to be too strong, particularly in the vicinity of Rote. While the NCOM current speeds used in GCOM3D were slightly stronger than suggested by Dr Sprintall (see below), Dr Hubbert nevertheless considered that data to be “closer to representing reality” than the SUNTANS data.

Ocean circulation in the Timor Sea region: the ITF

513    Before dealing with Professor Ivey’s modelling of the currents, it is convenient to address, firstly, some aspects of the evidence dealing with ocean circulation in the Timor Sea region and the significance of the ITF.

514    Essentially, the ocean, which covers 70% of the Earth’s surface area, is a series of interconnected ocean basins. Globally, the average ocean depth is around 3.7 km, while the width of the ocean basins can range up to 15,000 km. In light of these dimensions, the ocean basins can be seen as shallow water bodies, relatively speaking.

515    The water in these basins has variable temperature and salinity, and hence water density, with typically less dense water at the surface and heavier, more dense water at depth. Given the external forcing of the wind blowing on the less dense water surface, the horizontally non-uniform heating by the sun, and the gravitational attraction by the moon and the sun, the water in the basins is in motion. These external forces create internal forces, referred to as pressure gradients.

516    As a result, the ocean is in constant motion at its surface and throughout its depth (ocean currents). The term “ocean circulation” refers to the strength and the direction of these currents over the width and depth of the basins as they respond to these processes.

517    The geographic region which is the focus of this case (so far as ocean current modelling is concerned) has a total area of about 8 million km2. This region is bounded by the Indonesian archipelago to the north and the Australian coastline to the south and it is connected to the open ocean to both the west and the east. The Timor Sea lies within this region. It is bounded by the island of Timor to the north, the northwest coast of Australia to the south, the Arafura Sea to the east, and the Indian Ocean to the west.

518    Professor Ivey said:

4.1    The Arafura Sea and the Gulf of Carpentaria are both shallow marginal seas, with large areas and water depths of less than 50 m, and the Gulf of Carpentaria is connected to the southwestern Pacific Ocean via the shallow (less than 20 m deep) Torres Strait. To the southwest, the Timor Sea connects with the Australian North West Shelf (NWS). Shelf regions are typically defined as having water depths of less than 200 m, and the NWS width approaches 600 km offshore of the Kimberley region. To the west the eastern portion of the Indian Ocean has water depths extending to more than 4000 m. To the north is the complex of islands of the Indonesian archipelago, with numerous ocean passages, ultimately connecting to the western Pacific Ocean.

    The Region is large and geographically complex with the large number of islands of differing sizes in the Indonesian region and the bays and inlets of the northern Australian coast, as well as … large variations in depth … The ocean circulation is influenced by the land masses; the highly variable ocean bathymetry (depth); and the external forcing from large-scale pressure gradients, wind stresses and tidal forcing; and is highly complex. …

519    Professor Ivey described the ocean waters in this region as follows:

3.4    The ocean waters in the Region are continually in motion, driven by a number of forces external to the water body. These forcing agents include the atmospheric winds, which apply a stress (force per unit area) on the ocean surface; tidal forcing due to the gravitational attraction or pull from the moon and sun acting on the entire water body; and large scale horizontal pressure (force per unit area) differences internal to the ocean. The oceanographic processes in the Region are both spatially complex (on physical scales ranging from a few m to 100’s of km) and variable in time (from scales of minutes to years).

3.5    Due to the combined effects of differences in water surface elevation and lateral differences in density, there is a large-scale pressure gradient between the western Pacific Ocean and the eastern Indian Ocean. This pressure gradient forces a flow southwards through the Indonesian archipelago, and this is what is known as the Indonesian Throughflow, hereafter denoted as the ITF. The ITF is the dominant current system of the Region, and best estimates from the scientific literature suggest it has an average annual flow rate of approximately 15 Sverdup (Sv), where 1 Sv is equal to 1 million cubic metres per second (m3s- 1). To give a sense of scale, this flow rate is approximately 75 times the flow in the Amazon River, the largest river in the world.

3.6    The local wind fields are highly seasonally variable and strongly influenced by the monsoons in summer (Northwest Monsoon) and winter (Southeast Monsoon). The monsoons are very large-scale atmospheric systems that extend from South and South East Asia down into Northern Australia and west over the Indian Ocean. During the winter months in the Southern Hemisphere, the Southeast Monsoon (SEM) drives winds from the southeast over the entire Region. During the Southern Hemisphere summer, the Northwest Monsoon (NWM) drives strong winds from the west and monthly averaged wind speeds exceed 5 m per second (ms-1) over the Region. During the cyclone season from December to April, there are on average about 5 tropical cyclones that occur over the warm ocean waters of North Western Australia.

3.7     The tides, which are very strong in the area, are forced by the gravitational attraction of the moon and sun acting on the water in the ocean. Along the northwestern Australian coast, tidal forces drive strong surface tides with vertical displacements of the air/water interface (or free surface) that exceed 3 m above and below still water levels during a single tidal period. In some near coastal regions, these movements (or tidal amplitudes) are considerably larger. The tides are modulated on a cycle of 14.7 days known as the spring-neap cycle. The vertical movement of the free surface drives horizontal flows which reverse every tidal cycle. The dominant tides in the region are the M2 (period 12.4 hours) and S2 (period 12 hours) tides. Internal tides (driven by the combination of these surface tides, the strong vertical density gradient and the sloping bottom bathymetry) generate reversing horizontal currents which vary in magnitude over the depth. These internal tides also contribute to both ocean transport and mixing in the Region.

3.8    The combined action of these diverse processes conspire to make the Region a complex and energetic ocean region.

520    All witnesses agreed that the most important large-scale current system in the region is the ITF. As I have said, the ITF is a highly variable and complex flow of water from the Pacific Ocean that weaves through the Indonesian seas and out into the Indian Ocean. After noting an observation by Dr French-McCay concerning the “certainty” that the ITF flows westward into the Indian Ocean, Dr Sprintall observed that while, in an average sense over long-time scales (greater than a year), the ITF pathway through the Timor Passage flows into the Indian Ocean, on shorter time scales (less than a year) the ITF shows strong spatial variability and that, through the Timor Passage, it would rarely, if ever, appear as a cohesive “river” type flow that fills the Timor Passage and presents a barrier. Indeed, citing her own published (and, I would add, accepted) work Sprintall et al 2009, (Sprintall, J, Wijffels SE, Molcard R, and Jaya I “Direct estimates of the Indonesian Throughflow entering the Indian Ocean: 2004-2006” (2009) 114 J. Geophys. Res. C07001), Dr Sprintall said that physical observations show that there can be opposing flows on either side of the Timor Passage that can shift laterally northwardly and southwardly, with time. Some of this variation is due to eddies; some may be due to the response to winds. Dr Sprintall said that, in the “real world” (in contradistinction, I assume, to the “modelled world”), eddies and wind-driven flow are responsible for significant variability in ocean currents.

521    Dr Sprintall also took issue with Dr French-McCay’s statement that the ITF extends 85 km from the Indonesian coast in the area of the Timor Trench. Dr French-McCay cited Sprintall et al 2009 for that proposition. However, Dr Sprintall observed that there is no discussion at all in Sprintall et al 2009 about the distance of the ITF flow in the Timor Passage from the Indonesia coast or that (as Dr French-McCay also said) the ITF was persistently present as a southwestward flow. Dr Sprintall said that there was strong observational evidence that the ITF in the Timor Passage would, at times, be at very different distances from the coast of Rote and exhibit different current strengths within and across the Timor Passage.

522    In summary, Dr Sprintall said:

17    The ITF cannot be considered as a single coherent southwestward stream occupying a stable position in width and depth within Timor Passage. Such a picture only exists in a long-term average or idealistic sense. Rather, the ITF is a dynamic oceanographic feature with flow that on times scales less than a month would constantly change in width, depth, strength, direction and distance from the Indonesian coastline.

523    I accept this evidence. It is based on Dr Sprintall’s undoubted expertise in relation to the characteristics of the ITF. Her expertise on this subject was not matched by any other witness in this case.

524    Dr Luick, who is also a physical oceanographer, gave evidence to similar effect. He disagreed with Dr French-McCay’s statement (if considered as a general statement) that, regardless of quantity, any oil or dispersant carried with oil reaching the area of the ITF would have been carried southwestward by the ITF, away from Rote Island. He rejected the notion that the ITF is a barrier to surface-floating oil or substances that would otherwise reach Indonesia. He said that, during the period of the oil spill, large eddies frequently spanned the width of the ITF, creating strong northward flow paths across the ITF. These so-called “mesoscale eddies” were, typically, between 20 and 200 km in diameter. He said that an eddy moving in a 0.5 m/s current travels more than 400 km in 10 days. These loops and eddies are the primary pathways that carry floating material across the Timor Passage. During the spill period, floating oil would have received a northward boost from the prevailing southerly winds.

525    In his report, Professor Ivey gave the following evidence in relation to the ITF and the physical processes operating in the region of interest from August 2009 to January 2010:

3.9    All of the physical processes discussed above operated during the particular period of interest from August 2009 to January 2010. There was only one major tropical cyclone (TC Magda) during this time. TC Magda was observed over the Kimberley region from the period 19 to 24 January 2010, but after forming in the Timor Sea it travelled south over the Kimberley coast where it subsequently dissipated. This short lifespan, combined with the occurrence near the end of the period of interest, indicates TC Magda was only important for a small region near the Kimberley coast.

3.10    The wind fields acting over the Region can be divided into two distinct periods. From August to November 2009, the SEM controlled the wind fields. Over the Region itself, the winds were light with monthly means of less than 1 ms1 and came from the southeast. Winds were stronger in the regions surrounding the Region. In December 2009 and January 2010, the NWM was the dominant factor and strong winds with magnitudes exceeding 5 ms1 came from the west over almost the entire Region.

3.11    In terms of the ocean currents, the monthly-mean currents from the data-assimilating BRAN model confirm the dominant feature of the Region is the ITF. From August to October 2009, the ITF flowed along the southern coast of the island of Timor and then headed directly west, where it merged with the South Equatorial Current (SEC) in the eastern Indian Ocean. During this period, the BRAN results indicate the ITF had a monthly-mean surface velocities up to 1 ms1 and a width, in the upper 10 m , in the range of 50 to 100 km in the north-south direction. BRAN results also indicated that by December 2009 the ITF had started to slow, and by January 2010 the ITF was hard to detect.

3.12    While these periods coincided with changes in the magnitude of the wind velocities, the wind- induced surface currents were likely too small to arrest the ITF. The ITF is driven by large scale pressure forces associated with differences in sea surface height and water density. In the scientific literature, the years 2009 and 2010 were predicted to be years with stronger than historical average annual flows near 20 Sv (i.e. 20x106 m3s1 ). The weakening of the ITF starting in December 2009 is thus most likely related to the influence of remotely-generated eastward propagating large-scale planetary waves arriving in the Region, as was hypothesized in the scientific literature.

526    Dr Sprintall expressed concern about Professor Ivey’s use of monthly averages in relation to winds and currents. Her evidence was that there is significant variability in both the wind and ocean currents on time scales much less than a month:

3    … By averaging the wind or currents over a month-long period this variability will be masked and invisible. The ocean currents and winds can change significantly in both speed and direction on much shorter time scales than a month. This sub-monthly variability could have been represented for example through typical statistical depictions that oceanographers use such as standard deviations or variance ellipses (representing variation in both speed and direction) that show how much each day’s wind or current field differed from the typical values of that month. These statistical representations and their associated information about the variability were not provided by Professor Ivey.

527    Dr Sprintall pointed out that the winds in the Timor Sea region are monsoonal, meaning that they reverse direction twice a year (southern hemisphere winter, from the southeast; southern hemisphere summer, for the northwest). Professor Ivey’s report noted that the wind fields acting over the region can be divided into two distinct periods. However, Dr Sprintall observed that the two periods referred to by Professor Ivey encompassed a transition period (typically, September to October) when the monsoon winds are reversing direction:

4    … During the transition month of September, there may be day to week-long periods when the winds blow from either direction before making the complete change to be from the northwest direction. Thus, if the winds had reversed by 180 degrees during a month, then the average monthly winds could appear as very light or near zero, when in fact they could have been strongly from either direction. Monthly averaging will obscure this day-to-day variability in the wind direction and strength which would be [an] important contributor to the wind-driven flow.

528    Dr Sprintall observed that averaging currents over a month-long period will also mask any higher variability that can occur on time scales less than a month:

5    … although the large-scale current in the region – the [ITF] – over long-time scales of around a year flows from east to west, from the Pacific Ocean into the Indian Ocean, on any day along [its] pathway the current might be weak or even reversed and flow from the west to east. Reversals in the currents were frequently observed in the Timor Passage during the 2003 – 2006 field campaign that I participated in …[referring to Sprintall 2009]. During that field campaign much of the variability in the ocean currents of the Timor Passage was found on time scales less than 1-2 months most likely related to the higher frequency wind forcing but also due to instabilities (eddies) in the currents. Monthly averaging of the ocean currents will obscure this higher frequency variability of the flow direction and strength that would be important for understanding the higher frequency (hours to days) transport of oil contaminant on the ocean surface.

6    Currents averaged over a month, such as presented in the Ivey report, can conceal the presence of eddies. Timor Passage is a known generation site for eddies that can perturb the flow so much that it can cycle back on itself … Strong eddies can develop as a result of instability in the currents that can influence the direction and strength of the ITF currents. The eddies might exist in the region for periods of days to weeks. When these eddies are present they can influence the flow and sea level not only in Timor Passage but in the entire south-east Indian Ocean region between the Australian Northwest Shelf and the southern Indonesian islands of Nusa Tenggara (e.g. Sumba, Sumbawa, Lombok, Bali, Java). This region is known as the Indo-Australian Basin … It is feasible that the eddy reversals also influence the direction of flow in Sumba Strait that lies between the northern side of Rote Island and Sumba Island. In this way, eddy variability can impact the direction and strength of the flow both north and south of Rote Island.

529    I accept these qualifications to Professor Ivey’s evidence, once again based on Dr Sprintall’s undoubted expertise with respect to the characteristics of the ITF.

530    Professor Ivey used a theoretical model, Andersson and Stigbrandt (2005), (Anderson, H.C. and Stigbrandt A. 2005. Regulation of the Indonesian Throughflow by baroclinic draining of the North Australian Basin. Deep Sea Res: 52: 2214-2233), to estimate the width of the ITF in the Timor Passage (~50 km, although in other parts of his report Professor Ivey said 50 to 100 km), which was then used in his modelling. However, the Andersson/Stigbrandt model estimates the total transport (volume flow) of the ITF, which is fed by multiple pathways and has three major outflow passages: the Lombok Strait; the Ombai Strait; and the Timor Passage. The model predicts the ITF transport within a “Downstream Buoyant Pool” which lies in the Indo-Australian Basin within the Indian Ocean. As such, it represents the integrated ITF flow from the Timor Passage and the flow contributions made by the Lombok Strait and the Ombai Strait. It does not predict, therefore, the ITF transport, or its characteristics, within the Timor Passage.

531    Dr Sprintall said that the Andersson/Stigbrandt model is only appropriate to describe the ITF flow within the eastern Indian Ocean itself on large space and long-term scales, and that it is erroneous to assume that the model can resolve the width and depth of the ITF within the Timor Passage. I accept this evidence.

532    In other published work Hu and Sprintall 2016, (Hu S and Sprintall J “Interannual variability of the Indonesian Throughflow: The salinity effect” (2016) 121(4) J. Geophys. Res. Oceans 2596-2615), Dr Sprintall and her colleague estimated the flow rates of the ITF over a ten-year period. They predicted that 2009 and 2010 would have been years with stronger volume transport with transport rates of ~20 Sv, compared to the historical average of ~15 Sv. Professor Ivey relied on this prediction in his modelling.

533    Dr Sprintall expressed three concerns about this reliance. First, as discussed above, the Andersson/Stigbrandt model represents the integrated ITF comprising contributions of the flow from Lombok Strait, Ombai Strait, and the Timor Passage combined. Therefore, while 2009 and 2010 might have had stronger flow rates than the typical ITF transport, this may have been attributable to the contributions of the outflows from Lombok Strait and/or Ombai Strait, not necessarily from the Timor Passage. Secondly, the stronger than normal transport might have occurred at any time during the 2009 and 2010 years. For example, the flow might have been enhanced in July to August when the ITF is strongest, but then weaker than normal for the rest of the year, bearing in mind that the ITF exhibits strong variability on many time scales. Thirdly, (and Dr Sprintall said most significantly), the Andersson/Stigbrandt model provides no information on the vertical profile of the current (how current speed and direction changes with depth). Dr Sprintall said that there is strong observational evidence that the vertical profile of the various streams of the ITF change on seasonal and yearly time-scales. This means that, at times, the surface flow can be weaker than the flow at depth, and vice versa. It also means that a 20 Sv transport might be unevenly distributed throughout the water column with cores of stronger and weaker flow. I accept this evidence.

Professor Ivey’s modelling

534    Professor Ivey used the SUNTANS model to provide a detailed hindcast description of the ocean circulation in this region for the period August 2009 to January 2010. He described SUNTANS as follows:

3.13    SUNTANS model was used to provide a very detailed description of the ocean circulation in the Region over the period August 2009 to January 2010. SUNTANS is a state-of-the art baroclinic ocean circulation model with the ability to resolve small scale time varying flow features in the large, complex and energetic ocean environment of the Region. As configured for this study, SUNTANS provides predictions of all ocean properties (i.e., currents, temperatures and salinities) at horizontal scales as small as 1.5 km over a total domain of size 4000 km in the zonal (east-west) and meridional (north–south) directions, respectively, and over the entire depth. This represents prediction of ocean currents and ocean properties at a total of 10,414,110 grid cells or locations, every hour, for six months. The model is of necessity complex as it describes all the diverse forcing processes, the complex bathymetry and physical geometry of the Region. It thus provides an accurate description of the ocean circulation in both the horizontal and vertical directions over time. It requires considerable judgment, expertise and time to set up the model, to conduct the model runs for six months, and to analyse and interpret the large and complex data set produced from the model runs.

535    In carrying out his modelling, Professor Ivey used data from a number of publicly-accessible databases. The data included Bluelink ReANalysis (BRAN) data on average monthly currents, published by the CSIRO. BRAN assimilates (i.e., locally corrects) its predictions from available ocean observations (e.g., from satellite and Argo profiling floats) to derive a dynamically consistent best estimate of the ocean state. However, it does not include information on tidal forcing. For this information, Professor Ivey used tidal flows computed from a model called Oregon State University Tidal Inversion Software (OTIS). These flows were added to the velocities obtained from BRAN to establish the initial conditions, and also the boundary conditions on the open ocean boundaries, of his simulations. Professor Ivey also used the ECMWF winds data.

Professor Ivey’s criticism of GCOM3D

536    Professor Ivey said that the ITF was not just a near-surface flow; it has a large vertical and lateral extent which flows for many hundreds of kilometres in an approximately east to west direction. He said that this three-dimensional character can only be captured with “a baroclinic modelling approach”. In his report, he described the difference between his ocean current modelling using SUNTANS and Dr Hubbert’s ocean current modelling using GCOM3D, as follows:

E3.3     As Dr Hubbert uses the GCOM3D hydrodynamic model in barotropic mode, the pressure gradient that drives the flow in the horizontal direction is determined only by the free surface slope (Unless there is a total calm with no motion in the ocean, there are always differences in the height of the free surface of the ocean from one location to another. These differences in height correspond to a local slope of the free surface, and the ocean responds to this driving force by flowing from regions of relatively high elevation towards regions of relatively low elevation – that is flow is “down the slope”). In contrast, my primary report relies on the SUNTANS hydrodynamic model used in the more complex baroclinic mode. In the baroclinic mode, the pressure gradient that drives the flow in the interior of the ocean in the horizontal direction is determined by both the free surface slope and by the slope in the density gradients inside the ocean. The density is determined by local salinity, temperature and depth, and a baroclinic model must thus compute, at all depths and horizontal locations, additional equations describing conservation of heat and salinity. While these requirements add greatly to the computational requirements of a baroclinic model as compared to a barotropic model, the results are a much more accurate description of the ocean flow in the region.

E3.4     The two modelling approaches are quite different in a number of ways. The overall model domain of the SUNTANS model is slightly larger, and the duration of the SUNTANS run is slightly longer, than the GCOM3D model. In the horizontal plane, the two models have approximately the same grid resolution, but SUNTANS uses an unstructured model with the grid size decreasing (which yields higher spatial resolution) in the area of interest around Montara. The models are very different in their treatment of the vertical direction: GCOM3D only has up to 22 layers in the vertical, whereas SUNTANS has up to 100 layers. Density varies particularly strongly in the ocean in the vertical direction, and a primary reason why SUNTANS uses the 100 layers in the vertical is to capture this variation in density. As a consequence of this model set up, the GCOM3D model has (estimated) 220,112 cells, whereas SUNTANS has 10,414,110 cells. This is almost 50 times the number of cells or, equivalently, 50 times finer resolution in the computational domain. The finer the resolution, the better a model is able to describe the movement of the ocean.

E3.5     All models have boundaries at the edges of the model domain, and an example of the SUNTANS model domain is shown in Figure 12a in my primary report. The boundaries are of two types: closed boundaries and open boundaries. A closed boundary is when the model boundary coincides with with a land mass, in which case there is no ocean flow entering or leaving the model. At open boundaries, the model connects to the open ocean – such as the western side of the model domain shown in Figure 12a, for example. At open boundaries, ocean flow can (physically) both enter and leave the model domain. This information must be provided in order for the model to run, and this information is typically obtained from a coarser resolution ‘outer model’ that is providing predictions in the region outside – and these coarser large-scale models can be for the entire global ocean. At the open boundaries, as the GCOM3D model used by Dr Hubbert is barotropic, it only needs to transfer velocity information, in particular the two horizontal velocity components from the outer model (NCOM is the outer model in the Hubbert Report). The baroclinic model SUNTANS transfers not only these two horizontal velocity components but also salinity and temperature from the outer model (BRAN2016 in my primary report). This transfer is done at every cell on the boundary and at every time step, as the model moves forward in time to predict the ocean flow.

E3.6     More cells means better resolution and accuracy in describing motions. The SUNTANS model requires more computational time both because it has a greater number of cells and also because it must solve additional equations representing the conservation of heat and salt (hence density) at every location at every time step. The effect of these differences is seen in the differing run times required by the models: the GCOM3D model required only 15 Central Processor Units (CPU) hours, whereas the SUNTANS model required 26,352 CPU hours – almost 2,000 times greater computing effort.

E3.6     I provide a detailed comparison of Dr Hubbert’s GCOM3D model and the SUNTANS model used by myself, covering all aspects of model set up and forcing, in Section E9.

E3.7     The barotropic approach of Dr Hubbert’s GCOM3D model assumes that, at all times and at all locations within the computational domain, there are no variations in density of the ocean. In other words, this assumes that the ocean is well mixed in temperature and salinity, and hence density, over the entire depth and horizontal extent of the model domain at all times – the “well-mixed washing machine”. This assumption is not consistent with the field observations made in the relevant region (see, for example Hu and Sprintall (2016) and numerous references therein, and my primary report).

E3.8     While the barotropic modelling approach used by Dr Hubbert has the advantage of greatly reducing the computational times and computing resources required, it has the great disadvantage of not being able to describe any ocean motions that are dependent on density differences. This includes such important features in the relevant region as the large-scale Indonesian Throughflow (ITF) discussed in more detail in section E4, as well as any local baroclinic motions, internal tides, and the depth of penetration of wind-forced motions. A model running in barotropic mode for the Timor Sea region does not provide an accurate prediction or description of all the physical oceanographic processes of importance in the region.

537    It is important to understand that, in making these observations, Professor Ivey was proceeding on the basis that Dr Hubbert had relied only on barotropic processes in his modelling. As the evidence emerged, the deployment of GCOM3D was somewhat different.

538    Professor Ivey said that the equation for the flow rate, and the equation for the flow width, of the ITF each rely on the quantification of density differences. However, a barotropic model assumes that, at all times, the ocean is well mixed and that there are no variations in ocean density, vertically or horizontally—a “well-mixed washing machine”, as Professor Ivey described it. According to Professor Ivey, GCOM3D in barotropic mode will assume that there are no differences in ocean density (i.e., the ocean density difference is zero). Therefore, using the equations to which Professor Ivey referred, GCOM3D will predict that both the flow rate and the width of the ITF is zero. Thus, according to Professor Ivey, GCOM3D cannot predict any characteristics of the ITF, including its total flow rate, width, depth or location. Further, Professor Ivey said that GCOM3D cannot predict near-surface currents associated with the ITF. He continued:

E4.5    As the scientific literature (as summarised in my primary report) makes clear, the ITF does not cover the entire Timor Sea. Rather, the ITF is a broad and deep flow moving in an (approximately) east to west direction along the southern side of the Indonesian archipelago. The example calculation [referenced in Professor Ivey’s report] demonstrates that over a region extending at least 50 km south of the Indonesian coastline, the GCOM3D model running in barotropic mode will be unable to resolve the ocean currents in this region (contrary to what is said in the Hubbert Report).

539    In summary, Professor Ivey said that the use of GCOM3D in barotropic mode will not allow any description of the ITF (a proposition also advanced by Dr French-McCay). Further, he said that GCOM3D in barotropic mode will not allow description of any smaller scale flows driven by density differences, or the description of any internal tidal motions. Moreover, Professor Ivey said that GCOM3D is likely to prevent the accurate description of near-surface wind driven currents.

540    According to Professor Ivey, the use of ocean current predictions obtained from GCOM3D would, consequently, adversely affect any predictions from OILTRAK3D in respect of the movement of oil. Once again, it is important to understand that Professor Ivey was proceeding on the basis that Dr Hubbert had relied only on barotropic processes.

Dr Hubbert’s response

541    Dr Hubbert’s immediate response to Professor Ivey’s criticism was that the simulations he undertook with GCOM3D did, in fact, take information from NCOM, which is a large-scale baroclinic model. He said:

Table 1

E2 & 3    There is no doubt that a baroclinic model represents a broader range of physical processes in the ocean than a barotropic model. In this regard, I developed one of Australia’s first baroclinic ocean models back in the early 1980’s and GCOM3D is a derivative of those studies. Back then however such a model was only useful for research studies as there was not enough data on the processes occurring below the ocean surface to initialise a baroclinic model for real-time ocean forecasting. From 1985 I was a senior scientist in the BoM Research Centre tasked with developing their first ocean forecast models (waves, storm surges, tides, currents) which became the groundwork for the Bluelink baroclinic ocean modelling program used by Professor Ivey. The development and use of baroclinic models has to date been primarily limited to long term climate change studies. The accuracy of baroclinic models for use in short term forecasting and hindcasting is still limited by the lack of sufficient data to properly describe the ocean thermodynamic structure from seabed to surface. As my focus has always been on short term ocean forecasting (I developed the US Navy’s first coastal ocean forecasting system from 1995 to 1999 and worked on the last two of Australia’s America’s Cup campaigns in 1992 and 1995) I needed to find a way of incorporating what we can learn about large scale ocean currents and thermodynamic structures such as eddies in order to incorporate their influences in a timely fashion for real-time forecasting. The system I developed, which is encapsulated in GCOM3D, was to take the information from a large scale baroclinic ocean model (in these studies that is NCOM) and not pretend to have enough data to improve on it but incorporate the information from NCOM with the hydrodynamics simulated within GCOM3D to produce a composite result which was as close to the real-time situation as possible.

This is the system which has been running at AMSA Search and Rescue for the past 20 years and has been instrumental in saving a large number of lives that would otherwise have been lost if the GCOM3D system was not able to define an accurate search area.

GCOM3D has also been used in hundreds of environmental impact studies, each of which included verification of model predictions against data.

One of these cases was in the area of the Timor Sea whilst working for Woodside at Scott Reef. Extracts from the study report which are specifically relevant to verification of GCOM3D are included in Appendix A.

Dr Hubbert’s criticism of SUNTANS

542    Dr Hubbert also expressed concern about the SUNTANS model. I deal with a number of these criticisms below when discussing the Joint Report on Currents. However, it is convenient to draw attention to two matters now.

543    The first matter is that the only quantification of the currents output by SUNTANS reported by Professor Ivey was in the form of monthly averages. Like Dr Sprintall, Dr Hubbert said that the problem with monthly averages is that they do not immediately allow comparison with hourly data for short term forecasting. A further problem is that averaging tends to remove any indication of the existence of eddies which can have a significant effect on the hourly influences on the path of drifting oil slick. Dr Hubbert drew attention to the uniformity of the currents predicted by SUNTANS and the absence of eddies. He explained that monthly averaging would not pick up eddies that are present in the region if those eddies are not persistent in one location for the majority of the month which will, more often than not, be the case:

50    … Monthly averaging is a process not sensitive enough to pick up on eddies and other short term influences such as tides. However, this is not to say that eddies and other short term influences do not play a significant role in oil spill movement. To the contrary, an eddy can impact on oil movement significantly.

544    Dr Hubbert observed that NCOM had shown several eddies in existence in and around Rote on 9 September 2009. Dr Hubbert noted one eddy which he considered to be of particular importance. It was located south of Rote:

51    … I consider that eddy to be important because, like all eddies in the Southern Hemisphere, it possesses an anticlockwise movement which, on the eastern side, was forcing water in the opposite direction to the ITF. I believe that it would also have forced oil and dispersants, if any, in the opposite direction to the ITF. I believe that eddies such as this one were important mechanisms enabling oil to reach the coastal waters of Rote. Even if I am wrong in this regard, the influence of eddies should not be ignored if as accurate as possible a trajectory is to be obtained.

545    The second matter was Dr Hubbert’s concern that SUNTANS current speeds were overestimating the strength of the ITF. To investigate this issue, Dr Hubbert compared the current speeds from SUNTANS with the current speeds from NCOM and GCOM3D at three locations where current speeds were reported in Sprintall et al 2009. He found that SUNTANS speeds were consistently higher than NCOM and GCOM3D at the location closest to Rote (-11.1613(S), 122.7801(E)) and regularly higher at the second location (-11.2768(S), 122.8584(E)), southward. Further south into the Timor Passage, at the third location (-11.3699(S), 122.9567(E)), the current speeds from the three models were weaker and showed better alignment.

546    Dr Hubbert then compared the mean values of the current speed of the ITF in the Timor Passage measured by Sprintall et al 2009 with the mean values predicted by NCOM, GCOM3D, and SUNTANS. Although the different time periods did not allow a direct quantitative comparison, the mean current speeds predicted by NCOM and GCOM3D were slightly higher than measured by Sprintall et al 2009. However, SUNTANS predicted significantly higher current speeds, particularly on the Rote side of the Timor Passage.

547    Dr Hubbert performed a further analysis which compared the current speeds predicted by NCOM and SUNTANS at the Puffin and Jabiru locations north of the Montara oil field along the southern edge of the Timor Passage. At both locations the mean current speeds predicted by SUNTANS were more than double those predicted by NCOM.

Dr Sprintall’s criticism of SUNTANS

548    Dr Sprintall expressed the opinion that Professor Ivey’s modelling with SUNTANS would not accurately represent the true conditions that existed in the Timor Passage at the relevant time. This was because (as discussed above) the use of monthly averages obscures the higher frequency variability that is characteristic of the region, particularly with respect to eddies. Dr Sprintall argued that the Andersson/Stigbrandt model had been incorrectly applied by Professor Ivey to determine the width of the ITF in the Timor Passage. Dr Sprintall also said that there was a lack of validation of the modelling with regional observations. As I will come to explain, Professor Ivey subsequently sought to support his modelling with some such observations.

The Joint Report on Currents

549    In order to consider in more detail the differences in applications of the two hydrodynamic models (GCOM3D v SUNTANS) it is necessary to refer to the Joint Report on Currents to which Professor Ivey and Dr Hubbert contributed substantially. To explain, other experts (Dr Sprintall and Dr Luick) participated in the conclave that led to the production of the Joint Report. Therefore, when I refer to the “experts” in relation to this Joint Report, I refer to all the experts, not just Professor Ivey and Dr Hubbert.

550    The focus of the Joint Report on Currents was the respective modelling of the currents in the Timor Sea between the Montara H1 wellhead and Indonesia for the three month period of August to November 2009. The experts directed their attention to three main topics: the methodology adopted by each model (GCOM3D v SUNTANS); the relative accuracy of the data relied on in each model; and the relative accuracy of each model’s predictions.

551    Although the Joint Report contains some areas of agreement on these topics, the areas of disagreement were far greater and more fundamental, particularly as between Dr Hubbert and Professor Ivey in relation to the reliability and accuracy of their respective modelling. Ultimately, the Joint Report served the function of identifying, but not resolving, these disagreements.

Methodology

552    The experts agreed that, for the purpose of modelling currents in the Timor Sea between the Montara H1 wellhead and Indonesia in the period August to November 2009, both baroclinic and barotropic processes are important. Professor Ivey argued that a baroclinic model is more accurate than a barotropic model. I do not think that there was any substantial disagreement with that general proposition by the other experts. However, Dr Hubbert pointed out (in an earlier report responsive to Dr Ivey’s report and Dr French-McCay’s primary report) that, even though thermodynamic processes operate in coastal zones and on the continental shelf, their influence is usually minor compared with the dynamical forcing from the winds and tides. According to Dr Hubbert, forecasting and hindcasting in coastal zones is normally undertaken with barotropic models because they simulate these dominant processes and can be run in acceptable times on modern computers.

553    The experts agreed that GCOM3D and SUNTANS employ substantially different methodologies in their modelling. It is important to understand these differences because they formed the basis for Dr Hubbert and Professor Ivey challenging the accuracy of or, at least, casting doubt on, the results of each other’s modelling of the currents in the region and for the period in question.

554    For this purpose, it is convenient to start with the parent or outer models for GCOM3D and SUNTANS that, in the present case, provided their currents data. As I have noted, for the modelling undertaken in the present case, GCOM3D used data from NCOM. SUNTANS used data from BRAN. The experts agreed that NCOM and BRAN are similar in terms of their global coverage, resolution, computational approach, types of forcing, and the processes they simulate. These models also have similar limitations. They do not have flexible grids. They are limited by their grid size (10 km). They do not have information on tides. They each rely on observational data for assimilation and verification or, as the experts put it, “to nudge them towards reality”.

555    The experts agreed that both NCOM and BRAN perform well in specific regions of Australia (e.g., the East Australian Current, and the Leeuwin Current off Western Australia). They attributed this to the investment of time, money, and effort in collecting significant observational data to verify and improve the models so that they are, in these regions, reliable predictors of ocean currents. However, they observed that, in the Timor Sea region, there has been little concentrated modelling effort to validate the currents predicted by the models and there are few observations for verification. The experts also noted that, in this region, eddy features can be more variable in size.

556    According to the experts, the main difference between NCOM and BRAN is the time-step at which they make information available to set the boundary conditions used in GCOM3D and SUNTANS, respectively. The boundary conditions are the ocean properties at the boundaries of the three-dimensional domain that is being modelled. The top boundary is the air-water interface. The lower boundary is the bottom of the ocean. Water can enter or leave the model through the horizontal boundaries if they are connected to the open ocean. If a boundary is land, no flow occurs into or out of the model at this boundary. The frequency of updating the boundary flow to GCOM3D is every 6 hours (with an instantaneous or snapshot figure). In SUNTANS, it is every 24 hours (with a daily average figure).

557    GCOM3D and SUNTANS treat this information differently. As the experts described it, SUNTANS imports the baroclinic and barotropic information at the boundary and computes everything locally within the interior of the domain. Professor Ivey described SUNTANS as a “fully baroclinic model” or “single hydrodynamic model” that describes all barotropic and baroclinic processes for the relevant area at the same model grid at high temporal and spatial resolution. This grid is an unstructured or irregular triangular element mesh grid consisting of 218,035 cells in the horizontal. The horizontal resolution (the distance between cell centres) transitions from 10 km near the open boundaries of the model to 1.5 km in the central portion of the domain. Professor Ivey said that a key advantage of this unstructured model—and one of the reasons for using SUNTANS—is that it provides greater horizontal resolution, and hence a more detailed description of the flow, in the region of interest.

558    SUNTANS works by solving coupled, non-linear equations that describe the ocean’s movement at every grid point, advancing at 30 second time-steps from 1 August 2009 for a period of six months. This took 26,352 computer processing hours to complete. As if to underscore the reliability of his modelling, Professor Ivey said that the more refined a baroclinic model is, the more accurate its results will be.

559    However, as Dr Hubbert pointed out, SUNTANS relies on its own internal calculations without data assimilation. Professor Ivey explained that SUNTANS does not use data assimilation because this process is heavily dependent on the availability of high resolution data. Further, there is no single method or algorithm to “nudge” an exact or scientific solution towards an available measurement. Professor Ivey said that a lack of high resolution data is one reason why data-assimilating models are generally not useful for small time, high resolution ocean modelling—the kind of modelling which, he said, was required in the present case.

560    In comparison, GCOM3D combined with NCOM, as used by Dr Hubbert, is a hybrid model in the sense that (a) GCOM3D is a non-linear model that was, in the present case, run in barotropic mode, with forcing by winds and tides; and (b) NCOM is a separate, non-linear baroclinic model, with forcing by large-scale pressure gradients and wind. GCOM3D and NCOM have very different resolutions and time-steps. GCOM3D has a constant grid resolution of 2 km, and a time-step of 20 seconds; NCOM has a constant grid resolution of 10 km and, as I have noted, a time-step of 6 hours.

561    Dr Hubbert said that GCOM3D is run in hybrid mode by deriving thermodynamic information from NCOM rather than by solving internal algorithms, like SUNTANS. He said that GCOM3D combines local dynamical influences (wind, tide, bathymetry) with thermodynamics at each grid point from NCOM, and assimilates NCOM currents with its own internal calculations. He did not elaborate on how this is done, except in very broad terms during the course of concurrent evidence. He argued that GCOM3D’s hybrid modelling can simulate the hydrodynamics of the Timor Sea “very well”. He pointed to its use by AMSA’s search and rescue system (SAR) for the past twenty years (noting that AMSA uses another currents model, OceanMAPS, instead of NCOM). He said that this hybrid method of modelling was developed specifically for forecast modelling because it was impossible to run a full baroclinic model in a timely fashion for real-time forecasting. He said that, for search and rescue purposes, it had been determined that a more accurate prediction could be obtained by taking in temperature and salinity information from a global model rather than trying to solve these variables internally in GCOM3D itself.

562    Professor Ivey questioned the GCOM3D/NCOM hybrid model’s attempt to emulate the SUNTANS approach by combining two independent model outputs. He noted that Dr Hubbert had not provided any explanation or detail as to how the outputs of GCOM3D and NCOM were, or indeed could be, combined. This was significant because, as Professor Ivey explained, each model is non-linear and the principle of superposition states that only the outputs of linear models can be combined. Professor Ivey argued that, without explanation or further detail, the GCOM3D/NCOM hybrid model “cannot be understood or reproduced by an independent party”. For this reason, Professor Ivey did not recognise the GCOM3D/NCOM hybrid as a scientific model. He described it as merely an empirical one whose output has not been shown to be a valid solution of the equations describing ocean motion, and whose representation of the hydrodynamics of the Timor Sea was not valid. Professor Ivey argued that only the SUNTANS model is a reliable estimator of the currents in the Timor Sea in the time period addressed in the Joint Report.

563    Further, Professor Ivey said that, as NCOM does not describe tidal forcing, the hybrid GCOM3D/NCOM model is unable to model one particular baroclinic process that is significant in the Timor Sea—internal tides. He also noted that Dr Hubbert’s original simulation using the GCOM3D/NCOM hybrid model took 14.75 processing hours, which he compared unfavourably with the far greater processing time (and, implicitly, the far greater accuracy) of his simulation with SUNTANS (26,352 processing hours).

564    In the Joint Report, the experts discussed how the resolution of a hydrodynamic model can affect the accuracy of its results. They agreed that the resolution of the model should be fine enough to resolve the processes of interest in terms of both spatial and temporal variability in the region, and over the time period, of interest.

565    Models cannot describe processes and features on scales smaller than their grid resolution. In general, higher resolution models resolve more features. The experts agreed that, as both models in the present case use very different methodologies and spatial resolutions, they simulate small scale and mesoscale flows and eddies differently. These flows may be of importance in transporting oil and dispersants across the Timor passage.

566    The experts agreed that, as a rule of thumb, features such as eddies or flow variability along coastlines or in narrow straits are only well-resolved by a model if they are around 7 to 10 times the size of the smallest grid spatial resolution. Professor Ivey noted that temporal resolution was accommodated by the same rule of thumb. Using 10 as a rule of thumb, he pointed out that baroclinic features 15 km or larger, on a time-scale 5 minutes or longer, are resolved by SUNTANS. Further, the model output data is saved every hour, thus resolving features with time-scales of 10 hours or longer. Using the same rule of thumb, he said that the GCOM3D/NCOM hybrid model can only resolve baroclinic flow features 100 km or larger on a time-scale of 60 hours or longer. He argued that SUNTANS was, therefore, more accurate than GCOM3D/NCOM, particularly in relation to its spatial resolution.

567    Dr Hubbert accepted that the difference in spatial resolution between GCOM3D and SUNTANS might mean that SUNTANS resolves current speeds generated by small eddies better than GCOM3D. But the main effect would be that GCOM3D would underestimate the current speeds of the small eddies (because, in the modelling, they would be slightly “smoothed out”).

568    On the issue of model resolution, Dr Sprintall observed that increased complexity and resolution does not necessarily imply enhanced accuracy of the modelled result. She said that all results need to be verified by observations that might be difficult to obtain on the scales of the models in issue.

569    Further, Dr Sprintall noted that boundary conditions are provided to GCOM3D every 6 hours whereas these conditions are provided to SUNTANS only once a day. These properties are considered for the entire period the model is running. After noting that boundary conditions are provided more frequently in GCOM3D than in SUNTANS, Dr Sprintall said:

3    This will have an impact on the model behaviour. This means that the GCOM3D model gets to feel and subsequently adjust to the upstream conditions provided by the bounding model 4 times more often than the SUNTANS model.

570    The experts addressed the extent to which other (observational) data was available to verify each model’s results. They agreed that in order to test a model, such data must be continuous and of good quality; independent of the model (i.e., not assimilated into the model in any way); and contemporaneous with the period being modelled. Although identifying some data that was available to test the models (such as regional and local wind observations, satellite sea level measurements, drifting buoys, and regional tide gauges), the experts agreed that, overall, there was a lack of observational data during the time period in question, which made it difficult to test each model’s performance. Importantly, no currents meter data for the region was available for the time period in question.

571    Dr Hubbert referred to the fact that, for his modelling, he compared the BoM winds he used with meteorological data from Browse Island (he said that the BoM winds verified well against this data). He also said that GCOM3D produced a good prediction of sea levels at Scott Reef (about 350 km southwest of the Montara oil field—see the chart reproduced in Schedule A). However, in neither case did he provide the relevant statistics.

572    Dr Hubbert criticised Professor Ivey’s modelling because no such verification had been undertaken. In this connection, he said that it was critically important that the sea surface temperatures simulated by SUNTANS be compared with the satellite data. This is because, although SUNTANS used data from BRAN, it discarded the BRAN predictions for temperatures and relied entirely on its internal algorithms to predict temperatures in the region. He said that, to assess the overall performance of SUNTANS, it was important to know how well these internal calculations were simulating temperatures. Notwithstanding this criticism of Professor Ivey’s modelling with SUNTANS, Dr Hubbert said that it was not necessary to compare the GCOM3D sea surface temperatures with satellite data because those data were already included in GCOM3D from NCOM.

573    Professor Ivey’s answer to this criticism was that the best data to predict currents in the Timor Sea during the period in question would have been actual current measurements taken at that time. But, as no currents meter data was available, he made no comparisons.

574    Further in this regard, Professor Ivey relied on a peer-reviewed paper (of which he was a co-author), Rayson et al (2018), (Rayson MD, Ivey GN, Jones NL and Fringer OB, “Resolving high-frequency internal waves generated at an isolated coral atoll using an unstructured grid ocean model” (2018) 122 Ocean Modelling, 67-84), which provided a formal statistical comparison of the SUNTANS model predictions with measured currents and other flow properties, and hence quantitative measures of the SUNTANS model accuracy. Professor Ivey expressed the opinion that this paper provided “an excellent guide to the accuracy of SUNTANS predictions in the present application”. This was because the model described in the paper overlapped the domain that Professor Ivey modelled in the present case, and was centred on Scott Reef which, as I have noted, is about 350 km southwest of the Montara oil field. Of this region, Professor Ivey said:

4    It is a region with strong tides, currents and buoyancy effects. The paper makes extensive quantitative statistical comparison between the model predictions and measurements collected in 2015 in water depths from near surface and down to 400 m from two through-the water–column moorings which measured currents, temperatures and sea-surface elevation.

575    After referring to a number of statistical comparisons made in the paper between the model and observed quantities, Professor Ivey noted the paper’s main conclusions, which included findings that the BRAN model used by SUNTANS was better than the NCOM model used by GCOM3D in representing background large-scale ocean temperatures and currents, and that ocean temperatures were predicted with a Murphy Skill score of 0.8 (a very high skill score).

576    Also, apart from noting that Dr Hubbert had not provided quantitative or statistical metrics in relation to the comparison he had made, Professor Ivey also noted that Dr Hubbert’s measurements at Scott Reef were collected in 2006 at two sites located in a shallow lagoon at the reef (the implication being that Dr Hubbert’s comparisons on this score were not as reliable as the findings in the paper that Professor Ivey co-authored).

577    Under this topic, the experts also addressed the question of whether tracking buoy data can be used to verify the results of the two models. This question was considered by Dr Luick in particular. I will deal with it when discussing Dr Luick’s evidence below.

578    The experts then addressed the question whether the size of the model domain will affect a model’s results. They agreed that it did, noting that the domain must be of sufficient size to cover the key physical processes and the scale of the features of interest. They agreed that the placement of the model boundaries is important, particularly having regard to the input of tides and other information at those boundaries. The question of model boundaries arises because the boundaries of GCOM3D and SUNTANS are at different locations.

579    Whilst agreeing that the model domain must be large enough to simulate the key physical processes and features of interest, Dr Hubbert said that the domain should not be so large that the information applied at the boundaries has minimal connection with those processes and features. He saw this as a major problem with the way in which SUNTANS had been set up for the modelling undertaken in this case:

6    The boundaries (particularly the eastern boundary) are so far from the region of interest that the information obtained by nesting in BRAN is of little help in simulating the physical processes in the region of interest.

BRAN included the assimilation of sea surface temperatures from satellite observations and data from other sources within the SUNTANS domain.

However, any benefit which could be derived from the BRAN simulation of the temperature and salinity structure across the region of interest and eddies in the Timor Passage is entirely lost.

As a result the thermodynamic structure must be simulated for a three month period by SUNTANS without the benefit of any input of data across the domain.

This raises the question as to how well the internal algorithms of SUNTANS were able to simulate temperature and salinity variations across such a large area for 3 months without any input of data.

The mathematical algorithms within SUNTANS may well be state of the science but by no means can they be expected to accurately simulate the real world over the 3 month time period of the Montara oil spill.

By contrast we do not expect an atmospheric model to accurately forecast 7 days ahead of time and definitely not 3 months!

This point underpins my view that it was essential for SUNTANS predictions of sea surface temperatures to be compared with satellite observations to give some understanding of the reliability of the model predictions.

GCOM3D does not have this problem as it constantly updates the temperature and salinity within the model domain with the data from NCOM, which in turn has assimilated all available thermodynamic data in the region.

580    For his part, Professor Ivey argued that the size of the model domain must be large compared to the scale of the process of interest—here, the potential size of an oil spill from the Montara H1 Well over 6 months. He said that, in the case of his modelling, there was a finite size region between the coupling of his outer model BRAN with the inner model SUNTANS. This region was about 50 km in from the edge of the domain boundary. He said that this was tiny compared to the overall model domain size, which was 4,000 km east to west and 2,000 km north to south. Professor Ivey said that the important thing was that the correct boundary conditions are applied at the edge of the model domain. He repeated his opinion on the accuracy of the internal algorithms of the SUNTANS model (particularly with respect to its predictions of sea surface elevations, temperatures, and currents) which, he said, demonstrated that the location of the boundaries does not influence the model’s performance in the domain of interest.

581    The experts also addressed the of question whether the time-step update of boundary conditions in each model could have affected the potential for error in the modelling that was undertaken. They agreed that this may be important because the boundary inputs must be updated regularly enough to capture the important physical changes in the ocean region external to the model domain. These include changes in temperature, salinity, and velocity (both tidal velocity and large-scale velocities such as the ITF), and changes at the free surface (the surface which defines the boundary between the ocean and the atmosphere), such as winds and free surface elevation.

582    Dr Hubbert saw the frequency of updating the boundaries of SUNTANS as a major problem. As I have noted, in Professor Ivey’s modelling, SUNTANS only obtained new data every 24 hours (as a daily average) compared to GCOM3D’s updates from NCOM every 6 hours (as an instantaneous or snapshot figure). Dr Hubbert said that significant changes can occur within a 24 hour period, which he believed could have significant impact on the accuracy of the SUNTANS predictions. He argued that the infrequent updating of the boundary conditions of the model placed even more reliance on the ability of the SUNTANS algorithms to accurately predict, without the benefit of data assimilation, the ocean thermodynamics in the region for the three month period in question. I have already noted Dr Sprintall’s observations concerning the frequency of updating in GCOM3D compared to the frequency of updating in SUNTANS.

583    Professor Ivey was of a different view. He said that the time-step of updating the boundary conditions in his modelling with SUNTANS should make no significant difference in the interior of the model domain. He said that the information coming in at the lateral boundaries, from both NCOM and BRAN, is very coarse information (in that these models can only resolve flow features or eddies of a scale of about 100 km or larger). He emphasised his view that the model structure and algorithms, model spatial resolution, and model temporal resolution within the model domain, are much more important considerations.

Data

584    The experts agreed that physical measurements of the observed currents are the most valuable data concerning the ITF. Surface drifters observe the near-surface flow. Profiling current meters measure the velocity of the currents in the upper ~30 to 500 m of the water (an assumption being made that the velocity of the 0 to 30 m layer is similar).

585    As I have noted, the difficulty in the present case is that, in the period of interest, and in the region of interest, there is a lack of contemporaneous observational data available for direct comparison with the currents data that Dr Hubbert and Professor Ivey relied on. This then returned the experts to consider the differences between the currents data that Dr Hubbert and Professor Ivey relied on, and whether those differences affected the accuracy of their model outputs.

586    Overall, Dr Hubbert said that he had three main concerns with the SUNTANS methodology. First, SUNTANS takes on the role of simulating a large region of the ocean but discards information over that region predicted by BRAN and relies on its internal algorithms to accurately predict variations in temperature and salinity over the three month period. Dr Hubbert argued that, given the limitations of ocean forecasting, this was an impossible task. Secondly, he referred to the fact that SUNTANS derives updated boundary data from BRAN once a day, whereas GCOM3D derives its updated boundary data from NCOM four times a day. Thirdly, the winds driving SUNTANS are on a very coarse grid (100 km) and are likely to produce errors in the wind-driven currents near the coastlines of Indonesia (amongst other places). Dr Hubbert expressed his belief that these are the major causes for the currents predictions from SUNTANS possessing significant errors and showing very poor agreement with the drifting buoy tracks (discussed below).

587    Professor Ivey said that, in modelling with GCOM3D, Dr Hubbert relied on currents from NCOM at all points and at all times in the interior of his model domain. These currents are only coarsely resolved in time and space. In SUNTANS, total currents data (both barotropic and baroclinic data) from BRAN are relied on at the outer boundaries of the model, but these boundaries are remote from the flow regions of interest. SUNTANS computes all the internal dynamics. Therefore, the currents data at the boundaries is not that important. As I have noted, Professor Ivey said that model structure and algorithms, and model spatial and temporal resolution within the model domain, are much more important.

588    Lastly on this topic, the experts discussed the extent to which the winds data affected each model’s output. They agreed that the winds sets used by GCOM3D (the BoM winds) and SUNTANS (the ECMWF winds) are recognised as state-of-the-art science products by the scientific community. However, these model winds have different spatial and temporal resolutions—12 km (approximately) and 3 hours for the BoM winds; and 100 km and 6 hours for the ECMWF winds. The experts agreed that the different wind resolutions may impact the model’s ability to accurately drive the wind-driven component of the total current close to the coast. This is less significant on the regional scale but, closer to the coast, the local wind effects (island topography, land-sea breezes) may be more significant. The experts agreed that, for this reason, higher resolution winds may be more important.

589    Professor Ivey said that, in the hydrodynamic model, currents are dominated by the influence of large-scale pressure gradients, other baroclinic processes, and the strong tides. Wind is a secondary influence.

Model predictions

590    The experts were asked to give an opinion as to the extent to which Professor Ivey’s SUNTANS model and Dr Hubbert’s GCOM3D model accurately predicted the characteristics of the ITF in the relevant region and period of interest in this case (including current speeds, flow rates, width, depth, and location of the ITF). In answer, the experts noted that the two models differ substantially in their methodology and spatial resolutions. In particular, the two models simulate differently the small scale and mesoscale flows and eddies that may be of importance for transporting oil and dispersants across the passage. In the end, the experts simply agreed to disagree on the accuracy of the two models.

591    The “agreement to disagree” was in the context of the experts agreeing that the characteristics of the ITF in the region and during the period of interest were “largely unclear” because, apart from data obtained from tracking buoys, no other observational data was available. I will deal with the tracking buoy data shortly. It is, however, convenient to now record the following evidence given by Dr Luick:

15    JL: In my opinion, no hydrodynamic model of the Timor Sea during 2009 can predict the fate and transport of oil and dispersants with any practical reliability.

It is not the fault of the models themselves. In other regions, with stable flow regimes and abundant support data, both of the models (SUNTANS and GCOM3D/NCOM) are provenly capable of excellent fidelity to natural surface flows. The problem here is twofold: in the Timor Sea in 2009, the support data (by which I primarily mean the three-dimensional structure of temperature and salinity) is non-existent, and the regional currents are so complex and variable that even with good support data, modelling would be a challenge.

592    Dr Sprintall expressed a similar sentiment in her oral evidence.

593    For the purposes of the experts’ conclave leading to the Joint Report on Currents, Professor Ivey produced three Figures (in addition to those in his report) to show predicted currents (by SUNTANS) at short time scales over various depths. The purpose of preparing these Figures was to support the accuracy of the SUNTANS modelling by comparing outputs of the model against other measured observational data.

594    The first Figure (Figure 24) depicted currents predicted at the same location as the deepest physical mooring of an array of instruments referred to as The International Nusantara Stratification and Transport program (INSTANT), which were deployed in 2003 to 2006 in the vicinity of, but downstream from, the region of interest. Professor Ivey noted a number of limitations with respect to the data from this array:

8    The main limitations of the INSTANT measurements were: measurements were only made in 2003-2006 and not at the relevant time; moorings containing instruments for the purpose of taking measurements were only installed in 8 locations, which is not a large sampling given the complexity of the region; and the instruments were unable to make direct measurements in the strong flow region from the surface down to about 30-50 m depth (which is an important area given the strong flows in the region).

595    Notwithstanding these limitations, Professor Ivey used the INSTANT data for comparison. He said that Figure 24 showed results from the SUNTANS model in August, September, and October 2009 that were very consistent with what might be expected from the measurements made during INSTANT in 2003 to 2006.

596    In the course of concurrent evidence, Dr Sprintall argued that Figure 24 shows the model current results at one location in the middle of the passage, and that there is no concurrent observational data to confirm that this model output behaved like the real world over that period. She also said:

...we have no spatial context to put this time series in … and it’s only the flow in the east-west direction … We have no information spatially. It could be in the middle of an eddy. We have no information about the north-south flow on this figure. So it could be the north-south direction. If there is an eddy present, it’s stronger than what we are seeing here.

597    The second Figure (Figure 25) depicted predicted surface currents at the same location for Figure 24, and compared those predictions with predictions made by satellite altimetry-estimated geostrophic currents from two models—OceanCurrent and SSALTO/DUACS. Geostrophic currents are steady, large-scale ocean flows that are driven by a balance between pressure gradient forces and the Coriolis force. Professor Ivey said that the surface currents predicted by SUNTANS were “in good agreement” with the predictions from the satellite observations. In examination in chief, Dr Sprintall said:

Figure 25 includes only model geostrophic currents – that is, that they’re calculated from the temperature and salinity of the – from the model and then compares them to currents predicted from two satellite products. But there seems to be – there – of around the same magnitude but there really is no coherence between the variability that those three products show, so there’s no correlation between what we see at any one time in any of the products that agrees. In addition, the absolute currents which would include this geostrophic estimate – but they also include estimates from ageostrophic flow, such as might be driven by the winds and also from eddies, and that is not included in that Figure 25.

598    Later in oral evidence, Dr Sprintall said:

Figure 25, again as I said earlier, Professor Ivey is correct in that the magnitude is in the ballpark of the – of these two estimates that are derived from altimetry, but again there is no – if we look at the lines themselves over that three month period, there’s no correlation with the movement, with the variability …So again we only have one location. We don’t have any information about the other component to the flow.

599    The third Figure (Figure 26) showed flow rates for the ITF which are actually smaller than the flow rate of 20 Sv estimated in Hu and Sprintall 2016. The point of Figure 26 was to illustrate Professor Ivey’s contention that SUNTANS did not over-predict currents in the ITF in his modelling. With regard to Figure 26, Dr Sprintall said that:

... again there are no concurrent observational data provided to verify that the model output is representative of the real-world currents … they are simulations for the currents and not the actual currents, so we have no idea how representative they are of the flow conditions in terms of their strength, timing, location or direction of the currents in the real ocean.

600    Further, Dr Sprintall said that:

... from this figure we don’t get a sense of whether the flow is happening on one side of the passage or the other or at what depth level it is occurring.

601    In response to Dr Sprintall’s criticisms of Figures 24, 25, and 26 in oral evidence, Professor Ivey said:

I think my overall point [in presenting] the information in three alternate ways in the three figures was ... to give a number of points of view, both a view as to what the model is predicting at a particular location and time, in terms of velocity, what it’s predicting in a region for the total flow rate, and what is the depth variation in time. So the sum total is to give a sense of the model’s capability and to show generally that the model is very consistent with the estimated magnitudes of flow rates and currents that have been noticed or observed from actual moorings in previous years.

Tracking buoy data

602    More than a dozen tracking buoys were released during the Montara oil spill, each being tracked over periods of days or weeks. Buoy 1F24 was released on 31 August 2009, about 10 days after the spill commenced, about 50 km to the northeast of the H1 Well. It immediately began moving towards the north. It was tracked for nine days.

603    Dr Luick carried out an analysis whereby he compared the hindcast trajectories of eight hydrodynamic models, including SUNTANS, against the observed movement of Buoy 1F24, using tracking software called ATRANS.

604    To undertake this comparison, Dr Luick released 400 virtual buoys (i.e., spillets) into each model at approximately the time and place at which Buoy 1F24 was released. The 400 spillets were then allowed to drift in accordance with the particular model’s surface currents for the same number of days as the buoy.

605    Of the eight models tested, SUNTANS was the least successful at reproducing the track of Buoy 1F24. SUNTANS predicted the spillets would go to the south, whereas Buoy 1F24 went to the north.

606    Dr Luick observed that Buoy 1F24 followed the regional ocean currents to the north, rather than the local wind stress, which was to the west. This reinforced that, no matter how sophisticated the oil trajectory model, the trajectories they predict are only as good as the underlying advective currents.

607    The SUNTANS currents were also compared against a second buoy, Buoy 1F5E, which was released on the same day as Buoy 1F24, about 35 km to the east. Buoy 1F5E went west for three days, then turned sharply north. The spillets in SUNTANS went due south at approximately the same average speed as those released at the 1F24 site.

608    Dr Luick noted that, of the eight models tested, mesoscale eddies in the ITF were most visible in the OceanMAPS model. This is an Ocean General Circulation Model (OGCM) run operationally by the BoM. It is run in the same way as a weather forecast system, where the model is run successively in predictive mode, with new data assimilated each day, making the predictions increasingly reliable as the day approaches the current day.

609    Dr Luick considered that two other models, HYCOM and ozROMS, were equally disrupted by mesoscale eddies at different times and captured the high spatial and temporal variability of the northern Timor Sea, although the three models disagreed on the timing of the mesoscale eddies. To Dr Luick, the other models, which included SUNTANS and BRAN, seemed unrealistically “smooth”.

610    Given that mesoscale eddies are a ubiquitous feature of the region, and were referred to in reports at the time, Dr Luick concluded that OceanMAPS, HYCOM, and ozROMS were likely to be the most accurate in showing the size and frequency of mesoscale eddies, if not accurate in their exact timing. OceanMAPS, in particular, showed the combination of a counter-clockwise eddy followed by a clockwise eddy, generating a particularly strong north-westward flow in between (just north of) the H1 Well and ending at Rote Island.

611    Dr Luick said that, when mesoscale eddies are present, the ITF is not a barrier to northward transport; in fact, the eddies can provide surface pollution with “a rapid ride to the north”. He considered that the currents from the track of Buoys 1F24 and 1F5E from 31 August to 9 September 2009, taken with the OceanMAPS surface currents, indicated a high probability of oil crossing the ITF in the first half of September 2009, and reaching Indonesia.

612    Dr Hubbert also carried out a comparison of trajectories predicted by GCOM3D/NCOM surface currents with drifting buoy tracks. One comparison was made with the track of an AMOSC buoy, which headed north and then southwest in the ITF. The predicted GCOM3D track showed excellent agreement with the recorded track of the buoy. However, SUNTANS predicted a track that headed away from the ITF, towards the Australian coastline.

613    Other comparisons were made with the trajectories of buoys released by AMSA (1F2F, 1F5E, 1F24, 1F57, and 1F59). In the Joint Report, Dr Hubbert referred to a comparison with seven AMSA buoys, but only five were illustrated in his presentation to the Court. The comparison showed reasonable agreement, in that the GCOM3D/NCOM predictions showed movement in a similar direction, although there were variations in the extent of movement.

614    The tracking buoy evidence given by Dr Luick and Dr Hubbert respectively must be weighed with acceptance of the fact that the behaviour of one tracking buoy may not be representative of the behaviour of other tracking buoys released during the course of the spill. Further, the experts (including Dr Luick and Dr Hubbert) agreed that the deterministic prediction of a single buoy is not necessarily indicative of a predictive model’s skill. Sophisticated statistical tests are needed to assess the accuracy of a model’s representation of any buoy trajectories. Model results are sensitive to the location and times where the numerical buoy is deployed in the model’s finite grid. Professor Ivey, in particular, pointed to literature on the subject, Sebille et al (2009), (van Sebille E, van Leeuwen PJ, Biastoch A, Barron CN, and de Ruijter WPM, Lagrangian validation of numerical drifter trajectories using drifter buoys: application to the Agulhas system” (2009) 29 Ocean Modelling 269-276), which defined a quantitative method of assessing the skill of an ocean hydrodynamic model in predicting drifter or physical buoy paths.

615    The experts agreed that, in this case, the behaviour of drifting buoys are a qualitative, but not definitive, indicator of the current trajectories in the Timor Passage at the time of the spill. A valid statistical comparison of the model currents with the observed buoy trajectories would provide a good assessment of how well the model currents represent real world conditions. However, this evidence is not available to the Court.

Lagrangian Coherent Structures

616    Dr Luick gave evidence that, following the Deepwater Horizon spill in the Gulf of Mexico, beginning on 20 April 2010, there was a growing recognition that when oil, or a large number of drifting objects, are spilled into the ocean’s surface, they do not spread out evenly over a wide area. Rather, they aggregate into filaments called Lagrangian Coherent Structures (LCSs). These filaments are also referred to colloquially as “tiger tails”. The filaments capture oil or buoyant objects, which then move along the path of the filament. The filaments are often associated with the edges of mesoscale eddies.

617    The significance of this evidence is that it provides part of the context for the evidence that has been adduced concerning the visual sightings of oil during the time of the oil spill. For example, the absence of sightings of oil during the course of aerial surveillance does not necessarily mean that Montara oil did not travel northwardly towards Rote/Kupang. The absence of sightings could simply mean that the surveillance was not in the immediate area where the oil might have travelled to, and this could be due to the fact that oil might have travelled via conduits created by the activity of mesoscale eddies interrupting the flow of the ITF—a ubiquitous feature of the region. Indeed, the extent of aerial surveillance was limited in any event because of international borders. As Dr Luick opined:

7.3    Once out of the Montara vicinity, a search aircraft travelling north from Montara would not have observed oil on the conduit pathway following such a path until the surveillance track came within perhaps 30 kilometres of the coast of Indonesia. In other words, the surveillance aircraft may have been looking in the wrong place.

GIS Mapping

618    Nathan Kearnes is a principal consultant at Eco Logical Australia Pty Ltd (Eco Logical). Eco Logical is an environmental consultancy agency which provides a range of environmental consultancy services including Geographic Information System (GIS) mapping and remote sensing.

619    Eco Logical was engaged by the applicant to prepare static maps that depicted data obtained from a variety of sources. The sources included: Dr Hubbert’s modelling; Dr French-McCay’s modelling; data extracted from Dr Gundlach’s reports; observations made by lay witnesses; AMSA tracking buoy data; an AMSA dispersant map (showing the locations AMSA said were sprayed with dispersant during the spill); INSTANT moorings data (Dr Sprintall’s work); aerial observational data; flight path data; and vessel observational data in relation to the oil spill, to name just a few of the sources.

620    Of most relevance to this proceeding is a series of composite maps which were prepared by taking a data set from Dr French-McCay’s daily trajectory model output and then mapping that output alongside other data (for example, AMSA observations) onto a series of maps whose purpose was to compare Dr French-McCay’s model predictions with actual observations made at the time of the oil spill. These maps comprise Annexure NK – 23 to an affidavit made by Mr Kearnes on 26 November 2019. Some of the maps were relied on by the applicant to contradict Dr French-McCay’s evidence that the observations made by responders in the field at the time of the spill supported her model’s predictions. I discuss the results of this exercise in the next section of these reasons.

621    In response to Mr Kearnes’ evidence, Dr French-McCay produced a supplementary report containing additional Figures comparing her modelled results to observational information contained in AMSA documents. This is the report to which specific objections as to admissibility were made by the applicant, and on which I ruled in an earlier section of these reasons. The model runs presented were the same ones presented in Dr French-McCay’s primary report. No new model runs were performed.

Findings and conclusions on Trajectory Modelling

622    As I have previously remarked, oil spill trajectory models can be useful tools, but the results of their simulations cannot be taken as anything more than indicative of outcomes considered to be likely. Model predictions must not be mistaken for a true representation of what happens in the real world.

623    To produce the predictions of the trajectory of the oil in the Montara spill, Dr French-McCay and Dr Hubbert used a hierarchy of four models: (a) a large scale ocean model (BRAN or NCOM), feeding information into (b) a local scale ocean model (SUNTANS or GCOM3D), driven by (c) an atmospheric model (the ECMWF or the BoM winds); and (d) an oil spill model (SIMAP or OILTRAK3D), driven by (b) and (c). As Dr Hubbert put it, individually these models are unable to exactly simulate events in the real world. Inherently, the outputs of each model possess errors. The hierarchy leads to the final results being subject to compounding errors from the four models used. I accept that evidence.

624    In his evidence, Dr Luick referred to literature which discussed modelling undertaken as part of the Deepwater Horizon spill. Four different models were each initialised with virtual drifters. Four maps, showing the drifters two days after initialisation, exhibited major differences in the shapes of the oil patches as well as in the underlying current fields. Dr Luick employed this example to highlight the risk involved in relying on a single model to claim that oil cannot have reached a particular target area. I accept that caution.

625    The assistance which the trajectory modelling provides in the present case should not be overstated. While the modelling has the appearance of scientific and mathematical precision, and has been carried out by experts whose qualifications are impeccable, the Court has been presented with starkly different results. Dr Hubbert’s modelling shows that there is a real possibility, and a scientific basis for concluding, that oil from the H1 Well blowout reached Rote/Kupang. Dr French-McCay’s modelling shows that there is a scientific basis for concluding not simply that there is no real possibility of Montara oil reaching Rote (in closing submissions the respondent said the modelling by Dr French-McCay and Professor Ivey suggested it was “highly unlikely” that it did so), but a practical certainty that it did not and could not reach Rote. In this connection, it should be remembered that, in her initial report, Dr French-McCay claimed that her modelled results were definitive. Dr French-McCay conceded the possibility that small, inconsequential oil remnants reached the coastal waters of Rote and Kupang. However, this was the limited extent of her concession.

626    Each side advanced cogent criticisms of the other’s modelling, not only as to the accuracy of the models employed but also as to the reliability of the inputs used on which the modelled results depend. The state of the evidence is such that I cannot resolve each of the myriad debates that were raised and argued in the course of the evidence. But it is not necessary for me to do so. The trajectory modelling simply contributes to the pool of evidence before the Court to assist in determining whether Montara oil, in various weathered states, reached the coastal areas of Indonesia, in particular the Rote/Kupang region, as a result of the H1 Well blowout. As such, it has no precedence over the other evidence including, in particular, the actual observations of witnesses in that area at the time. There are, however, some matters going to my assessment of the trajectory modelling evidence to which I wish to draw attention.

627    First, the respondent advanced a number of criticisms of the OILTRAK3D oil spill model and GCOM3D hydrodynamic model to discount the reliability of Dr Hubbert’s modelling and to advance the reliability of the SIMAP oil spill model and the SUNTANS hydrodynamic model used in Dr French-McCay’s modelling. These propositions are not, of course, corollaries. The reliability or accuracy of the models used by Dr Hubbert in his modelling in no way reflects on the reliability or accuracy of the models used by Dr French-McCay’s in her modelling.

628    The evidence amply demonstrates that the dominant feature in the region of interest is the ITF. This is, undoubtedly, a highly complex body of water. The experts agreed that, overall, there is a lack of observational data during the time period in question, which makes it difficult to test the respective performances of OILTRAK3D/GCOM3D and SIMAP/SUNTANS in that region. At the end of the day, the experts simply agreed to disagree on the reliability and accuracy of the two combined models, as I have noted. I am placed in no better position. It is, however, ironic that the respondent criticised OILTRAK3D/GCOM3D as such, given that these models were used to carry out modelling in 2003 and 2005 that informed the OSCP on which the respondent, in fact, relies.

629    Dr Hubbert also gave evidence that he was involved in the development of the SAR prediction system which uses GCOM3D, and has been run daily at AMSA since 1999. He said the method by which GCOM3D derives temperature and salinity directly from NCOM was developed over 20 years of SAR operations at AMSA and that, given the high stakes of SAR operations, he would not use GCOM3D in the SAR system if he thought there was a better method of providing reliable predictions of ocean currents.

630    Secondly, I am persuaded that the most likely explanation for the difference in the modelled results is the data used for currents rather than the specific technical differences in the combined models themselves (despite the criticisms advanced). Indeed, the experts agreed that the respective models produced similar trajectories when run with the same inputs for currents and winds. On Dr French-McCay’s own analysis, the choice of winds data was of secondary significance and seems to have made little difference to the results. In the Joint Report on Trajectory Modelling, all experts agreed that the wind data used in the models did not have a material impact on the modelled results. Therefore, the ocean currents, and particularly the representations of the ITF, appear to have been the more significant drivers of the differences observed in the respective oil spill model projections.

631    I do not doubt that SUNTANS is a sophisticated and respected hydrodynamic model. In closing submissions, the respondent pointed to Dr Hubbert’s and Dr Luick’s endorsement, in oral evidence, of it as a model. The respondent quoted Dr Luick’s statements, in answer to questions put to him in the concurrent evidence session, that it is “a highly respected model” and he had “100% respect” for it. However, what the respondent failed to point out was Dr Luick’s immediately following remark:

…but, I mean, in this particular application I feel that it was unable to reproduce the important features of the Indonesian Throughflow, through no fault of the model numerics or physics or anything like that, just because of this particular application being 2009 and … before there …was a good dataset … to constrain it and before … the satellite altimetry was available … to be assimilated into it. So it depends what exactly you’re asking.

632    I have specific concerns about the output of SUNTANS that was used for the purposes of Dr French-McCay’s modelling. The ITF is a dynamic oceanographic feature with flow that changes constantly in width, depth, strength, direction, and distance from the Indonesian coastline, on time scales less than a month. This is significant because, as I have noted, the currents data used in SUNTANS for the purposes of Dr French-McCay’s modelling was based on monthly averages (as, indeed, were the winds data). I accept the real possibility that such data may well have masked the high variability of the ITF in Dr French-McCay’s modelling, particularly in relation to the presence and action of eddies.

633    Another concern is the adoption in SUNTANS, for the purposes of Dr French-McCay’s modelling, of the Andersson/Stigbrandt model of currents flow. I accept the real possibility that, by using this theoretical model, the flow rates used in SUNTANS may well have overestimated the strength of the ITF at the time in question, despite Professor Ivey’s attempt to show otherwise by other data. I have discussed the criticisms of using this theoretical model in some detail in earlier sections of these reasons. I will not repeat that discussion here other than to note my earlier acceptance of Dr Sprintall’s evidence on this topic, and Dr Hubbert’s similar observations in his evidence.

634    A further concern, albeit unrelated to the currents data, is the fact that SIMAP, as used by Dr French-McCay, does not appear to have been responsive (in terms of oil trajectory) to the volume of oil spilled. I am not persuaded that the volume of oil released from the H1 Well at the time of the spill was of no consequence as to its likely subsequent trajectory. I prefer, and accept, Dr Hubbert’s evidence that this proposition defies common sense. This lack of responsiveness in Dr French-McCay’s modelling seems to be a consequence of SIMAP’s limitation on the maximum number of spillets periodically released in the model. SIMAP released five spillets every 30 minutes. OILTRAK3D released around 200 spillets every two minutes, to represent the continuous spill.

635    As I have previously recorded, I am satisfied on the balance of probabilities that, over the period in question, oil was being discharged from the H1 Well at an uncontrolled rate in excess of the ranges considered in Dr Hubbert’s modelling (i.e., in excess of, on average, 2,500 bbl/day)many times greater than the assumption (400 bbl/day) used in Dr French-McCay’s base case modelling. As I have also recorded, a very large part of this oil was not treated with dispersants. Even when treated, the effect of the dispersants was not always successful, or completely successful, in dispersing the oil. Given the prevailing temperature and sea conditions at the time, I am not persuaded that much of the untreated oil or partially treated oil would have been dispersed by natural forces, particularly as it underwent increased weathering over time.

636    Thirdly, Dr French-McCay’s modelling led her to conclude that the closest that highly-weathered waxy residual oil came to the Indonesian coastline during August to November 2009 was approximately 32 km to the south of the Indonesian shoreline. (There was one exception shown by her modelling in the period 6 to 10 December 2009 represented by a single spillet of highly-weathered waxy residual oil which approached the coastal waters near Rote). According to Dr French-McCay, all oil and waxy residuals that reached within 85 km of the coast of Rote were caught up in the westward flow of the ITF and swept into the eastern Indian Ocean.

637    As I have noted above, I accept Dr Sprintall’s evidence that, contrary to the presentation given in Dr French-McCay’s reports, the ITF cannot be considered as a single coherent southwestward stream occupying a stable position in width and depth within the Timor Passage. I note that, by the time of the Joint Report on Trajectory Modelling, Dr French-McCay accepted that the ITF is not a barrier to oil flow, at least in the sense of acting as a wall.

638    Further, in the Joint Report on Trajectory Modelling, all the experts accepted that it was, indeed, possible that oil from the H1 Well reached the coastal waters of Rote/Kupang. However, Dr French-McCay’s acceptance of this proposition was highly qualified. She posited a number of conditions by which she thought that this might be possible. These conditions were of such a character, and expressed in such terms in the Joint Report, that it seems that Dr French-McCay really considered that eventuality to be implausible. Moreover, she said that any such oil would have the appearance of scattered microscopic pieces of waxy oil residual material in concentrations <0.1mg/m3 (parts per billion) or, as she also described it, “some small inconsequential oil remnants”.

639    Fourthly, these opinions expressed by Dr French-McCay—no doubt genuinely based on her faith in and adherence to the accuracy of her modelling—are inconsistent with the many observations of, for example, the waxy globules/residues seen floating at various locations just off the shoreline of Rote/Kupang or amongst a number of seaweed farms (assuming of course that the evidence of those observations is accepted). A number of the experts—in particular, Dr Sprintall, Dr Luick, Dr Thorhaug, Professor Ball, and Dr Fingas—considered these observations to be important data that could be used in forming a conclusion on whether Montara oil was present at these locations. I agree. In the next section of these reasons, I discuss the evidence of the weathering of Montara oil and how it weathers into waxy material. Conspicuously, Dr French-McCay did not review—because she was not asked by the respondent to review—the observations of the lay witnesses, which stood as observational data against which to test the accuracy and reliability of her modelled results.

640    In addition to these matters, there is also the analysis of the observations of oil, as presented through Mr Kearnes’ affidavit, compared with Dr French-McCay’s modelled results as originally presented in her primary report. It is to be recalled that, in that report, Dr French-McCay called in aid the observations of responders in the field at the time of the spill to support the accuracy of her modelling. The analysis provided through Mr Kearnes’ affidavit casts doubt on Dr French-McCay’s contention.

641    I have already mentioned the maps that comprise NK – 23. The applicant’s submissions focussed on some of them, namely those comparing Dr French-McCay’s modelled results (as mapped by her) with AMSA’s observation of oil sightings on 5, 13, 23, 25, 26, 27, and 29 September 2009. These comparisons reveal the following matters.

642    As early as 5 September 2009, AMSA’s surveillance recorded sightings of oil well north of the line that demarks the northernmost point of Australia’s Exclusive Economic Zone, and to the east of Dr French-McCay’s mapping of the modelled projections. The observation is of “infrequent and patchy slick/sheen”. As depicted in the AMSA map for that day, the oil extended north-eastwardly from the Montara wellhead in a substantially different configuration to the depiction of Dr French-McCay’s modelled results. However, the AMSA map does not show with any specificity where the infrequent and patchy slick/sheen was located other than within a hatched area on the map.

643    AMSA’s daily map for 13 September 2009 recorded “patches of sheen” well north of the demarcation line, over a significant area, south of Rote. AMSA’s daily map is no more specific than locating the broad area of the sighted patches. However, Dr French-McCay’s mapped results shows oil in substantial concentrations well south of this line, with only some low concentrations on or approaching this line.

644    The differences between the depiction of AMSA’s observations and Dr French-McCay’s modelled results are even more pronounced by 23 September 2009. Dr French-McCay’s modelling for 23 September 2009 shows that the highest concentrations of oil from the H1 Well were north of the Montara wellhead but significantly south of the demarcation line for Australia’s Exclusive Economic Zone. However, AMSA’s surveillance flight observations for 23 September 2009 record significant sightings of oil north of this line, in Indonesian waters.

645    Dr French-McCay explained that the mapping of the oil concentrations from her modelling for that day showed light pixelation which could represent the oil sighted by AMSA. She said that if, in her modelling, the floating oil was patchy and of very low concentrations, then the yellow spots used to map them would be small and not visible on the scale of the map.

646    Taking that to be the case, it is informative to understand that AMSA’s sightings on 23 September 2009 included: sheen 3 nm in length and ½ nm across, running north-east; a “large patch of orange oil” approximately 1 nm in length (the Lady Vasilia, referred to in other reports as the Lady Valisia, in company with another vessel was observed working on this patch); “long white coloured oil in windrows” running west to east and “another patch 3 NM” to the north-east; patches of “orange oil” (also described covering 2 to 3 nm and as “significant”, with the observation that the Lady Vasilia “may move to boom it on the 24th); and “orange oil … approximately 2 nm in length”. The surveillance report also records that, generally throughout the flight, oil was seen to be “in large columns tracking NE” and extending “beyond [the] observable range in this flight level”.

647    AMSA’s records for 23 September 2009 include a map showing sheen, and sheen with “orange patches”, well north of the demarcation line. The extent of this observation covers a substantial area that is, generally, to the south-east of, but approaching, Rote/Kupang, and is markedly different to the depiction and location of oil in Dr French-McCay’s map of the modelling.

648    AMSA’s surveillance report for 25 September 2009 records a number of other oil sightings well north of the demarcation line, once again generally to the south-east of, but approaching, Rote/Kupang. Once again, the depiction and location of these sightings is markedly different to the depiction and location of oil in Dr French-McCay’s map of the modelling, where the main concentrations of oil continue to be well south of the demarcation line. AMSA’s report records the following general observation:

From Lady Valisia to East oil sheen is continuous until flight track crossed Sahul Bank. Thickens as water depth decreases in lead up to bank. Limited oil sighted on bank – when bank fully crossed the sheen resumes and tracks to East. Appears a division between streams has developed either side of bank to NE of rig.

649    The following specific observations are made at different locations within this area: “Lady Valisia on orange oil patch booming. Sheen extends to NE horizon 1%”; “White/orange patches/lines in sheen 1-2% to horizon N-S extensive – centred on position. This info passed to L Valisia”; “White/orange oil patches 200-300n long 10m width – Passed to L Valisia”; “Sheen 1% with heavy metallic patches, intermittent white and orange oil continuous to horizon 90 degrees either side of track. Horizon line between sea and air indistinguishable”; “Heavy metallic patches – 2-3nm long x 1nm wide, on light sheen”; “Oranges patches 500m long 20m wide”; “Metallic patches 5% in sheen 1%” (two locations); “Multiple orange oil lines running N-S to horizon, 10m width spaced at 1/4nm intervals” (two locations); and “Vessel sighted in sheen. Sheen thins immediately E to this [position] …”.

650    Another daily report on 25 September 2009 records that a vessel (the Calypso Star) was carrying out tow boom operations in part of this area.

651    The following day, 26 September 2009, flight surveillance was carried out to determine the extent of slick coverage north and east of the Sahul Banks (the location of the Sahul Banks is shown in the chart reproduced in Schedule A). The following observations (amongst others) were recorded: “Lady [Valisia] directed to concentrations of orange oil …” (two locations indicated); “Orange and white oil 400m v 20m width – many patches in this size in this area …” (two locations); “Orange oil patches streaks 200-300m x 10m” (four locations); “Isolated metallic oil in sheen” (two locations); “Large concentration of oil – 2-3 m long 100 m wide”; “Isolated patches small concentrations orange oil in sheen”; “Orange oil streaks tending NW-SE”; “Small isolated patches yellow/white oil 50m diameter max”; and “Isolated orange oil in Sheen”. These observations were made at locations far removed from the locations of oil depicted in the map of Dr French-McCay’s modelling for that day. Once again, the depictions of her modelling show concentrations of oil well south of the demarcation line.

652    A flight report for 27 September 2009, which includes a flight path even closer to Rote/Kupang (approximately 56 km), includes the following observations: “Slick breaks to 75% continues in the follow locs” (three locations are given); and “Thin white and orange lines in 75% slick” (seven locations are given). Once again, these observations were made at locations far removed from those depicted in the map of Dr French-McCay’s modelling, which shows concentrations of oil well south of the demarcation line.

653    A flight report for 29 September 2009 once again records observations of oil far removed from the depiction of oil in the map of Dr French-McCay’s modelling for that day. The report includes observations north of the demarcation line in Indonesian waters, significantly closer to Rote/Kupang (and Timor more generally) than Dr French-McCay’s modelling shows. Relevant observations include: “Long windrow of white waxy heavy sheen”; “Windrows of weathered oil (yellow/orange) 1000m v 20m”; “Extensive patch of waxy windrows 1000m x 10m”; “Long windrows of weathered oil 2000m x 20m …”; “Breaking up of windrows with some heavier concentrated patches. Prime Target Area for Vessel Operations …”; “Extensive patch of oil in windrows and apparent accumulation point. Prime Target Area for Vessel Operations …”; and “Long windrows of weathered oil along the edge of the shoals with some scattered heavier concentrations in the middle of the mass 2000m x 20m”.

654    The respondent submitted that the maps of Dr French-McCay’s modelling used in these comparisons (even though prepared by Dr French-McCay) were not appropriate because they do not show entrained (subsurface) oil or oil residuals at all, both of which can be visible and appear as oil from aircraft and sea vessels. The respondent submitted that a more appropriate comparison is shown by supplementary maps which Dr French-McCay compiled and provided in response to Mr Kearnes’ analysis, because these maps show the actual extent of floating oil, entrained oil, and residuals (described by Dr French-McCay as “globs, flakes, tar balls and waxy residuals”), predicted in the modelling.

655    Dr French-McCay cautioned that not every observation on the AMSA maps would be Montara oil. She noted that, on 27 September 2009, the AMSA observers recorded observations of sheen with discolouration, and windrows at three locations. Dr French-McCay said that sheen and windrows are frequently seen on the ocean, in the absence of oil. I observe, however, that these are not the observations I have noted above for that day (which are, in any event, far removed from, and significantly closer to, Rote/Kupang than the observations referenced by Dr French-McCay).

656    Dr French-McCay’s supplementary maps were provided on two time bases: 5 day intervals in September and early October 2009 (Appendices A and C of her supplementary report) and on 5, 13, 23, 25, 26, 27, and 28 September 2009 (Appendices B and D of her supplementary report). Appendices A and B show modelling using the ECMWF winds and SUNTANS currents (as in Dr French-McCay’s base case) and Appendices C and D show modelling using NOGAPS winds (the US Navy model) and SUNTANS currents. Dr French-McCay said that the NOGAPS winds better capture the movements of oil north-eastwardly over the Sahul Banks in late September 2009. Modelling using these winds shows that oil residuals over the Sahul Banks moved into the ITF and southwestwardly, south of and past Rote.

657    Dr French-McCay said that the maps in NK – 23, with which she was provided, do not affect her opinion about whether Montara oil could have reached Rote or West Timor seaweed farms in September to November 2009. She said that AMSA observations in September 2009 are greater than 100 km from the Indonesian coastline and that the ITF was located between those observations and Indonesia. She said that any oil residuals entering the ITF from the Sahul Banks would have been swept to the south-west into the Indian Ocean. She concluded by saying:

Based on the modelling, I performed, it remains my opinion that neither oil nor dispersants from the Montara oil spill reached the seaweed cultivation areas along the coast of Nusa Tenggara Timur. The quantities of floating oil, and any dispersants carried with it, that passed within ~30-40 km of the coastline (the closest approach in my modeled base case) were very small, highly weathered, and in extremely low concentrations. Floating oil and subsurface microscopic oil residual particulates reaching <50 km off Rote Island were swept southwest-ward along the Timor Trench by the ITF.

658    Once again, this conclusion was expressed without reference to the observations of the lay witnesses.

659    What can be drawn from these comparisons? The immediate and obvious conclusion is that the sightings of oil reported in the AMSA records do not accord with the depictions of oil in Dr French-McCay’s modelling, as shown in the maps in NK – 23, to which the applicant drew particular attention. On the other hand, the supplementary maps provided by Dr French-McCay show a greater spread of oil and, on the whole, a greater correlation with the AMSA observations than the particular maps in NK – 23. But even then, the depictions in the supplementary maps are not coextensive with the sightings made by the observers.

660    These comparisons underscore the importance of looking at the recorded observations of oil that were actually made at the time to gain an appreciation of the extent of the oil from the H1 Well blowout that was moving towards Timor and the Rote/Kupang regionsome of it warranting booming operations where vessels could be located to undertake that task. Set against all the evidence of oil observations, Dr French-McCay’s broadly-expressed contention—that her model predictions are supported by the observations of responders in the field—needs to be treated with caution. Her contention is not a complete statement of the position because the contemporaneous observations indicate the presence of oil outside areas where Dr French-McCay’s modelling predicted it would be. To this extent, her model predictions are not supported by the observations of responders in the field.

661    In response to Dr French-McCay’s further contention that her modelling was also supported by interpretations of satellite imagery, the applicant relied on comparisons in NK – 23 between the depictions of Dr French-McCay’s modelling and the various satellite imagery in evidence. I have already commented in an earlier section of these reasons on Dr French-McCay’s own interpretations of that imagery, on which I am not prepared to act. The applicant directed my attention to maps in NK – 23 which show numerous patches of oil, which Dr Garcia-Pineda identified with “high confidence”, outside the depictions of Dr French-McCay’s modelling. I have also previously commented on the fact that Dr Gundlach demonstrated, persuasively, that Dr French-McCay’s modelling was not in agreement with Dr Garcia-Pineda’s analysis of the satellite imagery. These comparisons throw into further doubt the accuracy and correctness of Dr French-McCay’s modelling.

662    Taking all these matters into account, I am not persuaded that Dr French-McCay’s modelling shows, reliably, the trajectory of all the oil that was spilled as a result of the H1 Well blowout. I certainly do not accept that her modelling demonstrates, as a matter of scientific fact, that oil from the H1 Well blowout could not reach the Rote/Kupang region. Further, even though the respondent does not bear the ultimate burden of proof, I do not consider that Dr French-McCay’s modelling establishes, on the balance of probabilities, that oil from the H1 Well blowout did not reach the Rote/Kupang region.

663    By the same token, I do not consider that Dr Hubbert’s modelling, alone, establishes, on the balance of probabilities, that oil from the H1 Well blowout did reach that region. Of course, Dr Hubbert did not advance so definite a position himself. At all times he candidly presented oil trajectory modelling (including his own modelling) as subject to the limitations of the models and the data used in them. He emphasised that model simulations should not be confused with the real world. However, even given Dr Hubbert’s caveats, the respondent submitted that his modelling was not reliable for a number of reasons and could not support a finding that any oil from the H1 Well blowout reached Rote or Kupang.

664    First, the respondent submitted that the GCOM3D model was fundamentally flawed because, as it was run in barotropic mode, it could not describe ocean motions that depend on density differences and could not, therefore, provide an accurate description of the oceanographic processes in the Timor Sea region. The respondent pointed to the fact that the relevant experts on the modelling of ocean currents (including Dr Hubbert) agreed that, throughout the relevant domain, both baroclinic and barotropic processes are important and need to be included in any model of the region of interest.

665    As I have recorded above, Dr Hubbert addressed this criticism by explaining that GCOM3D is run in hybrid mode by deriving thermodynamic information from NCOM rather than by solving internal algorithms, as SUNTANS does. However, as I have also recorded, Dr Hubbert did not explain, to the satisfaction of the other experts (particularly Dr French-McCay and Professor Ivey), how this is done. The respondent pointed to Professor Ivey’s criticism that GCOM3D/NCOM could not be considered to be a scientific model, as opposed to an empirical one, because it combined non-linear outputs and, without explanation or further detail, could not be understood or reproduced by an independent third party. The respondent submitted that, even if combining the model outputs could produce reliable results, the hybrid model does not describe internal tides, which is a significant baroclinic process in the Timor Sea.

666    Set against these criticisms is the evidence that GCOM3D’s hybrid modelling has been successfully used for many years. Dr Hubbert pointed to its use by AMSA’s SAR system for the past twenty years. I doubt that AMSA would rely on that modelling for its search and rescue operations over that period if its predictions were found to be inaccurate in the field. Also, as I have pointed out above, GCOM3D was used in the modelling adopted in the OSCP on which the respondent relies. It seems unlikely that the respondent would have prepared its OSCP on the basis of a model recognised as dubious.

667    For completeness, I note that, in his oral presentation to the Court, Dr Hubbert called in aid certain tracking buoy data which illustrated some broad agreement, in terms of direction and extent, between the tracks of AMSA buoys and buoy tracks derived from GCOM3D currents. Whilst this provides support for Dr Hubbert’s modelling, I nevertheless bear in mind the caution to which I have previously expressed with respect to the tracking buoy data, namely that the trajectories of single drifting buoys provide only a qualitative, but not definitive, indicator of the trajectories of currents.

668    The respondent also submitted that the OILTRAK3D model also has not been explained properly. By reference to Dr French-McCay’s criticisms of Dr Hubbert’s modelling, the respondent submitted that Dr Hubbert had not explained key assumptions on which he had relied and that OILTRAK3D was an undocumented model which had not been peer-reviewed, verified or validated. In answer to these criticisms, it worth repeating that the experts agreed that when OILTRAK3D and SIMAP are run with the same inputs for currents and winds, they produce similar trajectories.

669    The respondent also criticised Dr Hubbert’s modelling on the basis that there was no adequate explanation of the OILTRAK3D’s outputs or results, with only static maps being provided for a small number of days which did not explain the trajectory of the oil on other days or explain what Dr Hubbert meant when he said that Montara oil was in the “vicinity” of Rote and West Timor. It is true that Dr Hubbert’s modelling was illustrated by static maps for certain time periods. However, this is hardly a criticism of the accuracy or reliability of his modelling as such.

670    Secondly, the respondent submitted that Dr Hubbert’s modelling included predictions that made “little if any sense”. The respondent pointed to two matters.

671    The first matter was Dr French-McCay’s observation that Dr Hubbert’s modelling appeared to predict more oil in the water than he assumed was released during the spill. Dr French-McCay expressed this view on the basis of an exercise she carried out, which estimated the areas within each of the concentration contours that Dr Hubbert had depicted in certain Figures in his report and the volume of oil in each contour interval. These volumes were then added to show a total volume. I have referred to this criticism earlier in these reasons.

672    Dr Hubbert addressed this criticism by explaining that the purpose of contouring is to show the likely concentrations of oil which might be found in regions of the sea (it being impossible to accurately forecast exactly where the oil will go). He accepted that the combined effects of randomised error (purposefully introduced into OILTRAK3D to mitigate against errors in inputs) and contouring the modelled results can give the appearance that Dr French-McCay had remarked upon. But this acceptance by Dr Hubbert must be seen in the context of his explanation that OILTRAK3D does not attempt to predict oil impact concentrations but, simply, the likely highest concentrations that might impact a region, and that OILTRAK3D’s modelling is directed to predicting oil trajectory.

673    The second matter was Dr Hubbert’s modelling for 11 to 13 September 2009, which showed oil travelling north-westwardly towards Rote at about 75 km/day. Given the wind direction and speed, Dr French-McCay calculated that Dr Hubbert’s model assumed that currents were heading north-westwardly at speeds averaging 1 m/s through the ITF, which Dr French-McCay did not consider to be reasonable. Once again, I have referred to this criticism earlier in these reasons. Dr Hubbert responded by arguing that Dr French-McCay’s calculation of current averaging was incorrect, for the reasons he explained. This is another example of the many disagreements between the experts which, ultimately, cannot be resolved by the evidence before the Court.

674    Thus, the respondent’s argument that Dr Hubbert’s modelling includes predictions that made “little if any sense” must be considered against the explanations given by Dr Hubbert that were responsive to the criticisms advanced by Dr French-McCay. Those explanations provide important context in which to understand what Dr Hubbert’s modelling shows.

675    Thirdly, the respondent pointed to the fact that Dr Hubbert’s modelling did not show oil reaching the northern shores of Rote, and could not predict whether oil reached Kupang. Both of these observations are correct. However, this submission attributes to trajectory modelling a precision, and the character of scientific infallibility, which I do not accept the modelling, advanced by each party, has. Further, Dr Hubbert explained the absence, in his modelling, of oil impacts in Kupang. He nevertheless expressed the opinion, based on his modelling, that it was likely that the spilled oil did, in fact, reach Kupang. That opinion is to be weighed with all the other evidence.

676    Fourthly, the respondent pointed out that Dr Hubbert’s modelling was sensitive to the various assumptions made with respect to the rate of oil released from the H1 Well and the effectiveness of dispersants that were applied. The respondent submitted that, depending on certain assumptions, some of Dr Hubbert’s modelled results would show oil impacts in late October 2009 “well after the seaweed died”. This submission, however, is based on the acceptance of two propositions. The first is that substantially less than 800 bbl/day of oil was released during the spill. The second is that a substantial volume of dispersant was effectively applied to the oil that was released. I have rejected both propositions.

677    As I have stressed, the trajectory modelling, with all its limitations, simply contributes to the pool of evidence before the Court to assist in determining whether Montara oil, in various weathered states, reached the coastal areas of Indonesia, in particular the Rote/Kupang region, as a result of the H1 Well blowout. The accuracy and reliability of Dr Hubbert’s modelling will fall to be assessed against all the evidence.

The weathering of Montara oil

Introduction

678    Crude oil is a complex heterogeneous liquid mixture of a very large number of hydrocarbon fractions and organic compounds. Organic compounds are any class of chemical compounds in which one or more atoms of carbon are covalently linked to atoms of other elements, most commonly hydrogen, oxygen, or nitrogen. Hydrocarbons are organic compounds which contain only carbon and hydrogen.

679    The organic compounds in petroleum crude oil are composed of linear and branched-chain volatile and non-volatile aromatic and aliphatic fractions, ranging from light gases with a small number of carbon atoms (1 to 4 carbon atoms, C1 – C4 compounds) to heavy residues (35 to 40 carbon atoms, C35 – C40 compounds). Aliphatic hydrocarbons are the main compounds of petroleum crude oil, comprising about 50% of hydrocarbon products, with aromatic hydrocarbons comprising about 26 – 30% of hydrocarbon products. Resins and asphaltenes constitute the remaining constituents.

680    In addition, crude oil contains some non-hydrocarbon, inorganic elements such as sulphur, inorganic nitrogen, and oxygen, and traces of metallic compounds, including phosphorus, lead, nickel, arsenic, and vanadium.

681    The composition of crude oils is not fixed. It may vary depending on the age and location of the oil field, and upon the depth of each individual oil well. Every oil well contains crude oil which differs in composition from oil sourced from any other well.

682    Montara oil is a mixture of thousands of different hydrocarbons. It is formed from a range of organic compounds, chemically converted under varying geological conditions over millions of years. As a result, Montara oil has a distinct chemical composition, different from other oils.

683    The weathering of Montara oil is relevant to two broad areas of inquiry in this case: first, the observations of Montara oil at the time of the spill and whether the observations of lay observers in the Rote/Kupang region in the second half of 2009 are consistent with the presence of Montara oil in those areas; and secondly, whether, in its weathered form, Montara oil was toxic to seaweed growing in the Rote/Kupang region at that time. This section of the reasons deals, primarily, with the first inquiry. Despite the topic’s overlap with the second inquiry, I will return to discuss the toxicological effects of weathered Montara oil later in these reasons.

The Leeder Reports

684    I have briefly referred to 64 field-collected samples of spilled Montara taken at various locations during the course of the spill. Leeder Consulting reported on these samples. They were mostly collected between 16 September and 8 October 2009. This represents only 22 days in the middle of the spill. The samples were mostly collected within about 50 km of the wellhead platform. The experts assumed that the samples represented the range of weathering, whilst recognising the possibility that there might be more highly weathered oils/waxes that were not sampled.

685    Based on a number of those reports, Professor Ball provided the following summary of the observed weathering of the oil at the time:

6.8     Evaporation occurs mainly during the first 24-48 hours after release which greatly reduces the number of volatile components. Some crude oils may lose up to 40% of their volume due to evaporation in the first few days after a release. The substance remaining after evaporation is called weathered crude oil. The smell associated with this weathered oil following evaporation will also change from a smell of gasoline to a heavier oil, smell for like engine oil. The composition of any released product remaining in the affected area is likely to be substantially different than the originally released crude oil. Due to the weathering process, the remaining product is generally considered to have less potential for causing adverse health effects.

6.9     The unique composition of Montara crude changes as soon as the oil enters the environment due to weathering. A number of processes including evaporation, dissolution, photochemical oxidation and biodegradation contribute to weathering of the oil, resulting in changes in chemical composition. For example, evaporation of the more volatile chemicals present in Montara crude occurred, resulting in the loss of compounds in the C6-C12 range. Many of these compounds are toxic to biota, as they are quickly able to interfere with cellular processes and membrane integrity. However, in the atmosphere these compounds are greatly diluted and undergo degradation.

6.10     A number of factors influence weathering; time and temperatures are important along with mixing; maximum weathering occurs in warmer temperatures with high winds ensuring mixing of the oil; also the longer the oil is in the environment the greater the weathering. Montara crude oil weathered significantly. Up to 88% loss was reported (Location 12o 41.4 S, 124o 42.7 E). As the process of weathering increased both the appearance and the properties of the oil slick changed. Over time the slick changed colour to a khaki colour containing oil and wax ...

6.11    Further weathering of Montara crude oil results in the appearance of a white solid wax cake ...

6.12     This change in the appearance of the Montara crude oil during weathering was accompanied by changes in the chemical and physical properties of the oil. These changes included:

    An increase in the pour point of the oil from 27°C (Montara fresh crude, Sample 2009021829) through to 51°C in Montara crude exhibiting 88% weathering (Sample 2009020424). This confirms that weathered Montara crude will be solid. The pour point of a crude oil is the lowest temperature that the oil will flow when it is cooled. ASTM D97, Standard Test Method for Pour Point of Crude Oils is used for pour point analysis. Higher pour point indicates that the oil contains more paraffin (Speight 2018). This indicates that as Montara crude oil aged it became more solid.

    An increase in the adhesion properties of Montara crude oil from 4.7 mg/cm² with fresh crude oil to 7.8 mg/cm² in 14% weathered Montara crude oil to 21 mg/cm² in 88% weathered Montara crude oil. Adhesion is a significant property as it assesses the ability of the crude oil to adhere to biological material. In this case as the oil weathered the ability of the oil to bind to biological material increased. Despite the fact that weathering was found to vary among samples, any amount of weathering will increase the adhesion properties of the oil.

    The flash point of the oil increased from <25°C in Montara fresh crude (Sample 2009021829) to consistently >62°C in all Montara crude oil with >30% weathering. The results here suggest that the probability of the oil catching alight decreased significantly during weathering.

    Fresh Montara crude (Sample 2009021829) was found to contain BTEX, not present in weathered Montara crude oil samples. Therefore any acute toxicity associated with these compounds will have been negated through the evaporation process.

    The wax content (%) of Montara crude increased from 11.3% in fresh Montara crude (see Table 1 in report) (Sample 2009021829) to 79% in Montara crude with 88% weathering and generally increased with % weathering. This would have resulted in an increase in the greasy waxy properties of the weathered oil.

    Analysis of the aromatic hydrocarbons of Montara crude during aging revealed a general decrease in their concentration from 269,000 mg/kg (Montara Fresh Crude Sample 2009021829) to 4,200 mg/kg in Montara crude that has been weathered by 88% (Sample 2009020424), representing a 98.4% reduction. In this sample only large aromatic hydrocarbons (C16-35) remained following 88% weathering. Generally, as weathering increased the concentration of aromatic hydrocarbons decreased, thereby reducing the acute toxicity of the weathered Montara crude oil.

    Analysis of the aliphatic hydrocarbons of Montara crude during aging revealed a general increase in their concentration from 58.08% (Montara fresh crude Sample 2009021829) to 74.7% in Montara crude that has been weathered by 88% (Sample 2009020424). This was largely due to an increase in high molecular weight aliphatic hydrocarbons (C16-C35) accompanied by a significant reduction in the lower molecular weight (and more biologically active) aliphatic hydrocarbons (AL-EC 5-16). This will again result in significantly reduced viscosity of the weathered Montara crude oil.

686    Based on the same reports, Professor Ball also provided the following descriptions of the form, properties and behaviour of the Montara oil in the marine environment:

7.4     The results confirm that during the initial stages of weathering of Montara crude, evaporation occurred during the first 24-48 hours after release which greatly reduces the number of volatile components. The smell associated with this weathered oil following evaporation will also change from a smell of gasoline to a heavier oil, smelling like engine oil. Montara crude oil lost many of the lower molecular weight hydrocarbons within the first few days and therefore the wax content of the oil increased making it a solid compound at all environmental temperatures. Whilst this would result in a reduction in the immediate chemical toxicity, the impact of the wax physically coating any biota would be significant. No quantitation of the wax content of weathered Montara crude oil was detailed, however it is known that the original wax content of Montara crude oil was 11.3%. As we know that up to 88% weathering occurred and given the fact that wax will not readily weather, it is my opinion that at this level of weathering almost all the remaining oil was wax. Evidence to support this conclusion comes from a study by Leeder Consulting which assessed the adhesion of various oil to duck feathers. Seven weathered oils sampled with varying weathering losses (11% to 88%) were selected for the study and compared to three reference oils: fresh Montara crude, fresh Light Arabian and a heavy fuel oil. Of the Montara oil samples, fresh Montara crude oil showed the lowest adhesion (4.7 mg/cm2) together with Montara crude oil with 11% and 12% weathering loss (2009020419 and 2009020418). Montara crude oil showing more than 14% weathering loss showed much greater adhesion to duck feathers (20-74 mg/cm2), similar to that observed for heavy crude oil (HFO380). This represents a significant increase in the adhesion properties of Montara crude oil as it weathered. The results indicate that on contact with biological material, significant adhesion of Montara weathered crude oil to the biota would occur, through an increase in the wax content of the weathered oil.

7.5    As temperatures increased some of the remaining non-solid components of the weathered Montara crude oil were released from the solid, forming a sheen on the surface of the sea. This thin film of oil on top of water settled as a thin layer, and the thickness of the layer, due to an optical phenomena called interference, caused the thin layer to shimmer in different colours often referred to as a rainbow ...

7.6    The weathering of Montara crude oil varied from sample to sample. One reason for this was the variability of environmental conditions. For example, if temperatures and wave action were low, less mixing and hence less emulsification would occur even in the presence of chemical dispersants.

687    Dr Stout agreed with Professor Ball’s summary of Leeder Consulting’s findings, subject to one qualification: he disagreed with Professor Ball’s assessment of the changes to the oil’s adhesion. Dr Stout considered adhesion to be a rarely studied property of oil. A literature search conducted by him revealed that none of the reported studies on adhesion were with respect to high wax oils like the Montara oil. Nearly all the oils studied were liquids when the adhesion tests were carried out.

688    The adhesion tests carried out by Leeder Consulting were to determine the potential for Montara oil to adhere to the feathers of marine birds. In simple terms, the tests were carried out by immersing pre-weighed duck feathers into the Montara oil for 30 seconds, allowing the feathers to drain for 30 minutes, and then re-weighing the feathers to determine the mass of the oil that had adhered. The testing was carried out at 28°C. By reference to the pour points of the samples of oil tested for adhesion, Dr Stout inferred that only fresh Montara oil was liquid during the adhesion tests, with the field-controlled samples “increasingly solid”. This raised a question for Dr Stout as to how reliably the “dunk and drain” technique used in the tests could measure the adhesion of the weathered oil. He said that not enough was known about the effects of temperature and pour point on the adhesion measurements that had been made for the “waxy Montara oils”.

689    Dr Stout did not think that Professor Ball’s opinions concerning the increased adhesion of weathered Montara oil were justified by the Leeder Consulting reports. First, he did not accept that the results of adhesion of the Montara oil to duck feathers could be extended to all biological materials, and submerged seaweed in particular. Secondly, Dr Stout argued that the Leeder Consulting adhesion data did not show a clear trend of increased adhesion with weathering. He observed a lack of relationship between adhesion and wax content.

690    Having noted the existence and broad outline of this disagreement, I will return to the question of adhesion when dealing with the topic of toxicity.

691    Dr Stout also noted that Professor Ball’s summary description did not provide any view concerning the rate at which Montara oil weathered after its release to the sea. I do not understand this to be a criticism of Professor Ball’s evidence. Rather, it is an observation about what can be concluded from the Leeder Consulting reports. The exact period of time each field-collected sample spent in the environment prior to collection is not known. Dr Stout said that the effects and rates of the weathering processes were “rigorously quantified” in his laboratory studies, to which I now turn.

Dr Stout’s analyses

692    Dr Stout carried out laboratory weathering studies on Montara-2 oil. Broadly speaking, there were two studies. They were carried out over an 18-month period. The results of his work were only made available in mid-March 2019, shortly before the commencement of the hearing of this proceeding in June 2019.

693    Dr Stout’s studies were designed to measure the effects of weathering on the chemical composition of Montara oil. The experts agreed that Dr Stout’s chemical composition results were directly relevant to the topic of toxicity. They also agreed, however, that the results were not useful in assessing certain physical properties of weathered Montara oil (specifically, visual appearance, wax content, pour point, viscosity, smell, and adhesiveness).

694    The first study was a short-term (evaporation) study carried out over 7 days. The other experts called on this topic, Professor Ball and Dr Fingas, agreed that this study replicated the evaporation that would be experienced from the sea surface by floating Montara oil. The short-term weathering study showed that:

(a)    there was a 20.8% reduction in the mass of the oil due exclusively to evaporation over the 7 day period, including 14% within 24 hours;

(b)    compounds below n–C11 were progressively depleted and ultimately eliminated from the oil over 7 days with some losses extending up to n-C14 (depletion of compounds boiling above n-C14 were not observed);

(c)    benzene, toluene, ethylbenzene, and total xylenes (BTEX) were evaporated within 4, 24, 24, and 48 hours, respectively; and

(d)    there was an 18% loss of total PAHs, including an 89% loss of naphthalene and a 46% loss of ethyl-naphthalene.

695    The experts agreed that the short-term weathering study:

12    … showed evaporation at temperatures akin to Timor Sea surface water during the Montara oil spill removed or markedly reduced BTEX and 2-ring PAHs from the Montara oil within hours to days. Comparable losses due to evaporation are expected to have occurred during the actual Montara oil spill. However, in addition, we expect that the losses evident in the [short-term weathering] study may even be conservative because the study did not consider any losses due to evaporation of the oil before reaching the sea surface (i.e., aerosolization).

696    The second study carried out by Dr Stout was an 84 day, long-term study to measure the long-term effects and rates of biodegradation and photo-oxidation of (what he described as) three common forms of Montara oil that existed during the spill. These forms were a chemically-dispersed Montara-2 oil (pre-evaporated 21 wt%); a wax-depleted oil fraction of the Montara-2 oil (pre-evaporated 21 wt%); and a wax-enriched fraction of the Montara-2 oil (pre-evaporated 7 wt%).

697    Direct observations by AMSA during the spill reported on solid waxy substances (which Dr Stout called wax-rich residues) floating on the sea surface. These wax-rich residues were seen floating among, and separately from, liquid oil. Dr Stout considered that the physical properties of the fresh Montara oil he analysed, particularly its high wax content (13.7 wt%) and high pour point (24°C), were consistent with these observations, especially as the oil weathered.

698    Evaporation causes an oil’s wax content and pour point to increase as volatile compounds evaporate into the air. Dr Stout noted that when fresh Montara oil (here, Montara-2 oil) was evaporated in the laboratory, its wax content increased to 18.7 wt% and its pour point increased to 29°C. He also noted that the average surface water temperature in the Timor Sea during the spill was approximately 29°C (slightly lower at night, and higher during the day). He opined that, when the temperature of the floating Montara oil fell below its pour point (mostly at night-times) wax-rich particles would have crystallised and then, through wave action, separated from the balance of a wax-depleted oil fraction. He also opined that daytime warming was likely to have re-melted the wax-rich particles until their pour point increased above ambient daytime temperatures, after which they remained as floating, solid, wax-rich residues.

699    As solid, wax-rich residues formed and collected during the Montara oil spill were not available to him, Dr Stout prepared his own residues by a chilled centrifugation technique in which the whole Montara-2 oil was chilled below its pour point during high-speed centrifugation, causing wax particles to crystallise and sink (after three such steps), forming a residue he called the wax-enriched oil fraction. Dr Stout considered the wax-enriched oil fraction, so produced, to be “reasonably representative” of the solid, wax-rich residues that were observed floating at sea during the spill—although, based on one sample analysed by Leeder Consulting, he thought that “purer” waxes were formed at that time. Thus, in understanding Dr Stout’s evidence it is important to bear in mind the distinction between the wax-rich residues (the term applied to the wax particles observed floating in the Timor Sea) and the wax-enriched oil fraction (the term applied to the artificially-created oil fraction obtained by Dr Stout through chilled centrifugation).

700    Dr Stout’s long-term weathering study was carried out in the laboratory using microcosms consisting of a 1 L glass jar with 600 ml of Timor Sea seawater and 45 µl of each form of oil. The seawater was maintained at 29°C and subjected to artificial sunlight by an ultraviolet-enhanced lamp. An indigenous bacterial culture obtained from Timor Sea seawater was added to the seawater mixture, as were nutrients. There were 78 microcosms (63 samples and 15 control samples). Dr Stout depicted each microcosm:

Figure 7: Schematic drawing of a microcosm used in the long-term weathering study. Oil (45 μL) was added to Timor Sea seawater (600 mL) maintained at 29°C, amended with indigenous bacteria and nutrients, continuously gently stirred, and irradiated (12 hr/day) for up to 84 days. ...

701    In describing the study, Dr Stout said:

[104]     The long-term weathering study was more complicated as it needed to control multiple conditions – temperature, oxygen availability, water/oil movement, salinity, bacterial abundance, nutrient availability, ultra-violet light exposure, oil chemistry (i.e., the three forms of oil), and time – concurrently. However, this complexity was not unique to the Montara oil spill. Laboratory microcosm studies, such as my long-term weathering study, are the conventional way by which oil spill scientists study the rates and effects of weathering on spilled oil. Furthermore, microcosm studies yield weathered oils with compositions that resemble those of naturally-weathered oils, which indicates that the processes of biodegradation and photo-oxidation that occur in nature can be reproduced in the laboratory. Although the results obtained from my long-term weathering study are limited to the specific conditions of the study, those conditions were set to mimic, as practically as possible within the constraints of the laboratory, those that existed during the actual Montara oil spill. Therefore, it is my opinion that the rates and effects of weathering determined during the long-term weathering study reasonably represent those that existed in the days and weeks following the release of Montara oil into the Timor Sea

702    Dr Stout described the overall results of the study in respect of the three forms of oil as follows:

[112]     ... The results of the study show:

(a)    Weathering through continued evaporation (beyond that of the short-term study), biodegradation, and/or photo-oxidation commenced in all three forms of Montara oil within the first 7 days;

(b)    Over the course of the 12-week study, two phases of long-term weathering were evident, being fastest and most extensive within the "early" phase (0-28 days), and slower and less extensive within the "late" phase (28-84 days);

(c)    Throughout the study the chemically-dispersed Montara crude oil and its (non-dispersed) wax-depleted oil fraction weathered at comparable rates and to comparable extents, indicating dispersant (Slickgone NS) did not obviously increase (or inhibit) weathering, at least not at the sampling resolution of my study;

(d)    The (non-dispersed) wax-enriched oil fraction weathered less and more slowly than the other two forms of oil studied, wherein weathering was likely slowed due to the wax's physical state; i.e., the smaller oil-water interface of the discrete, hydrophobic, floating waxy particles inhibited their weathering relative to the larger interfaces of the chemically- dispersed oil droplets or wax-depleted oil's thin sheen;

(e)    Although evaporation initially contributed, biodegradation played the major role in loss of n-alkanes, while a combination of biodegradation and photo-oxidation led to the losses among the PAHs, mostly within the first 28 days;

(f)    Biodegradation was ultimately so severe that the normally-recalcitrant biomarker compound, 17α(H),21β(H)-hopane, was partially depleted;

(g)    Timor Sea water undoubtedly contains a consortium of indigenous, oil-degrading bacteria capable of biodegrading Montara oil.

703    Dr Stout summarised his conclusions as follows:

[115]    In summary, my long-term weathering study showed that the combined effects of continued evaporation, biodegradation, and photo-oxidation dramatically changed the composition of Montara oil, and its wax-depleted and wax-enriched oil fractions derived from the oil, over the course of days and weeks. PAHs were rapidly and significantly reduced within the first 7 to 28 days, mostly (~80%) the first 7 days, albeit more slowly and to a lesser extent in the wax-rich residues (waxy particles) derived from the oil.

[116]    Collectively, based upon the results of my short- and long-term oil weathering studies, the spilled Montara oil dramatically and rapidly changed its composition and concentrations due to weathering at/near the sea surface. These changes commenced quickly and proceeded rapidly, albeit somewhat more slowly in floating wax-rich residues than in chemically-dispersed oil droplets or floating sheens. The mass of the spilled oil was substantially reduced as aromatic hydrocarbons were eliminated from or markedly reduced (1) within hours in the case of MAHs and (2) mostly within the first 7 days and overwhelmingly within the first 28 days in the case of PAHs.

[117]    Because the conditions of my short- and long-term laboratory weathering studies mimicked, as practically as possible within the constraints of the laboratory, those conditions that existed during the actual Montara oil spill, it is my opinion that the changes in composition and concentrations of the Montara oil determined during these studies reasonably represent those that existed in the hours, days, and weeks following the release of Montara oil into the Timor Sea. This conclusion is further corroborated upon comparison of the laboratory results to those of field-collected oils and waxes...

704    Dr Stout theorised that the slower and overall lower depletions of PAHs in the floating waxy particles were due to the PAHs “inside” the particles being less available to bacteria and less exposed to UV radiation than dispersed oil droplets or floating sheens.

705    As I have already noted, the field-collected samples analysed by Leeder Consulting were of an unknown age when collected. This fact precluded any rigorous quantitative assessment of the rate(s) of weathering of the Montara oil during the spill.

706    Despite this shortcoming, Dr Stout considered the Leeder Consulting data to be important because, in his view, they provided corroboration of the wax-enrichment and weathering processes simulated in his laboratory studies. Dr Stout remarked that chromatograms of the “white waxy particles” collected at sea closely resembled those of his laboratory-produced, wax-enriched oil fractions after weathering in the laboratory for 70 to 84 days. He opined that the white waxy particles observed during the spill may have taken multiple weeks to be produced and that, owing to their semi-solid nature, were likely to be the most persistent form of Montara oil in the environment.

707    With regard to this persistence, Dr Stout said (in Annexure E to his principal report):

[125]     My long-term weathering study showed that small waxy particles (~2 to 7 mm) physically lasted the duration of the study (84 days). Because weathering preferentially occurs on the exterior surface of the waxy particles, it follows that the rate of weathering of the waxy particles is inversely proportional to their size (surface area-to-volume ratio). Specifically, smaller waxy particles (larger surface area-to-volume ratio) would persist for a shorter period of time and vice versa. Physical processes (wave action or abrasion) would tend to reduce the particle size of waxes over time, eventually making smaller and smaller waxy particles, with larger and larger surface area- to-volume ratios, which would eventually weather completely.

708    Dr Stout also observed that Leeder Consulting had measured the concentrations of total aromatic hydrocarbons (>C7 to C35) in 13 field-collected samples and 16 PAHs in 42 field-collected samples. The data showed that, as the Montara oil weathered and/or became enriched in wax, the concentrations of total aromatic hydrocarbons and the measured PAHs decreased substantially. Dr Stout drew particular attention to the only “pure” wax sample measured by Leeder Consulting for aromatics, which retained only 1.6% of the total aromatic hydrocarbons (>C7 to C35) and 0% of the 16 PAHs. Dr Stout argued that this suggested that the loss of total aromatic hydrocarbons and PAHs from the “white waxy particles” observed during the spill likely exceeded the losses achieved in his laboratory weathering study.

709    Dr Stout also measured the concentrations of hydrocarbons, including monocyclic aromatic hydrocarbons or monoaromatic hydrocarbons (MAHs) and PAHs, dissolved in the seawater, or within oil droplets suspended in the seawater (particulate oil), potentially associated with three different forms of Montara oil (the WAF study). The three forms were: Montara-2 oil treated with dispersant (Slickgone NS); evaporated (21 wt%) Montara crude oil (obtained by evaporating fresh Montara-2 oil to the same degree as occurred after 7 days of evaporative weathering under Timor Sea temperature conditions); and an evaporated (7 wt%), wax-enriched oil fraction of Montara-2 oil. Dr Stout considered that these forms represented, respectively: fresh Montara oil sprayed with dispersant in conditions closest to the West Atlas platform; untreated, weathered Montara oil further from the platform; and the floating, solid, waxy residues observed in the field.

710    MAH concentrations were measured in the seawater directly, whereas PAH concentrations were measured in both the dissolved phase and particulate phase. Dr Stout referred to the dissolved and particulate phases as the water soluble fraction (WSF) and particulate oil fraction (POF), respectively. Taken together they comprise the water accommodated fraction (WAF).

711    It is important to understand that the WAF fraction is defined by the equation WAF = WSF + POF. By definition, the WAF fraction does not include the “floating oil”, including the floating wax-rich oil, that remained in the vessel containing the WAF fraction after harvesting to produce the WSF and POF fractions.

712    Dr Stout found that the seawater in contact with fresh Montara oil treated with dispersant contained the highest concentrations of all hydrocarbons in both dissolved and particulate phases. He said that this was due to a combination of this oil’s “freshness” and its dispersion into the water, which would have promoted dissolution of the oil’s more soluble compounds and the formation of suspended oil particles that remained in the water.

713    As this study was conducted at 29°C to mimic the average water temperature in the Timor Sea throughout the spill, the wax-enriched oil fraction remained as a solid, floating wax-rich particle throughout the experiment. Nonetheless, seawater in contact with this particle contained 814µg/L of MAHs, comprising mostly of BTEX (619µg/L), which apparently dissolved into the seawater in contact with the particle. PAHs were also dissolved into the seawater in contact with the particle (348µg/L), albeit in lower concentrations than the other two forms of oil studied. The PAHs in the wax-enriched oil fraction mostly comprised dissolved naphthalenes in the WSF (283µg/L).

714    It is to be noted that MAHs and PAHs are typically associated with toxic effects on marine biota. The objective of this study was to produce data of use to others to assess the potential toxicological impact(s) of the Montara oil spill on aquatic organisms.

Criticisms of Dr Stout’s long-term weathering study

715    Professor Ball expressed concern about the way in which Dr Stout had produced his wax-enriched fraction on which he conducted his long-term weathering study. He observed that spinning the oil at high velocity to produce the wax-enriched oil fraction was not, to his knowledge, a published and validated technique, and was not a process to which the Montara oil was subjected during the spill. He observed that Dr Stout’s opinion that the wax-enriched oil fraction was “reasonably representative” of the solid, wax-rich residues that were observed floating at sea during the spill, was an undefined standard involving a subjective judgment based on his (Dr Stout’s) visualisation of photographic images and the fact that the product he formed was a solid at room temperature. Professor Ball expressed surprise that, given that Dr Stout had 18 months to conduct his analyses, an 18-month weathering experiment had not been performed.

716    In the Joint Report on Chemical Composition, the experts agreed that Dr Stout’s long-term weathering study did not replicate (in the sense of mimic) the dissolution of Montara oil through weathering. This was because the study was conducted using 600 ml of seawater in glass containers, such as depicted above. They said that the glass containers prohibited the diffusion/dispersion of any dissolved chemicals away from the oil in the container. They observed that, in an open ocean, with an “infinite” volume of water, these dissolved chemicals would have been allowed to move away from the oil. As Professor Ball put it in a responding report dated 4 May 2019, Dr Stout:

5.4    … designed a simple experiment, with only a temperature and agitation being considered as environmental factors. Other key factors such as the impact of tides, winds and the presence of macro-flora and -fauna were outside the scope of this study.

717    Also, Professor Ball and Dr Fingas did not consider that Dr Stout’s long-term weathering study replicated the biodegradation and photo-oxidation of Montara oil through weathering. Dr Fingas commented that Dr Stout’s study provided no comparison point with the weathering of fresh Montara oil. Thus, what is not known from the study are the properties of the oil at different weathering stages; anything about unaltered Montara oil as it weathers past 7 days; the oil’s true viscosity at standard temperatures; and whether separation of the oil occurred and, if so, what happened to the separated fractions. Dr Fingas also commented on the fact that, despite Dr Stout’s long-term weathering study, little is known about how the appearance of the oil (for example, whether it is white, green, orange, yellow, or brown) is related to its chemical and physical properties.

718    Like Dr Fingas, Professor Ball commented on the absence of a control using fresh Montara oil to compare results over the 84 days of the experiment, although he noted that other controls were used in the study. Professor Ball described this as a significant flaw in the study. Further, a parallel experiment would have allowed the measurement over time of important parameters such as wax content, pour point, viscosity, dispersability, and toxicity.

719    Professor Ball also commented that Dr Stout’s biodegradation work involved isolation of a specific hydrocarbon-degrading microorganism from a pre-enriched seawater sample, which was then added to the seawater at the beginning of the long-term study. Professor Ball said that this addition would significantly impact on the biodegradation process. Further, after 21 days of incubation a further addition of the microorganism was made. Professor Ball argued that this was a bioaugmentation event aimed at enhancing the biodegradation of the petrogenic hydrocarbons. He pointed out that, as there was no estimation of the actual numbers or activity of the specific isolate in the fresh seawater, it was not possible to assess the impact of these additions on the biodegradation processes. Indeed, in his responding report of 4 May 2019, Professor Ball argued that Dr Stout’s comments on biodegradation were based purely on chemical fingerprinting data.

720    In relation to Dr Stout’s comments concerning “pure” wax samples, Professor Ball noted that Dr Stout had provided no definition of “pure”. He noted further that Dr Stout’s experiments showed that even after 12 weeks incubation, the concentration of PAHs remaining in the wax was 5,320 µg/g, representing 86% degradation (I note that in Table 9 to his report dated 12 March 2019, Dr Stout recorded total remaining PAHs as 6,204 µg/g). Professor Ball said that this wax should not be deemed to be “pure”.

721    Dr Fingas made a similar observation. He said that Dr Stout had repeatedly stated that extensive weathering would result in “pure” wax. Dr Fingas noted that one sample showed a large wax content (I assume this is the sample in the Leeder Consulting reports that Dr Stout relied on), but several other samples did not. He commented that the predominance of the evidence shows that wax concentration simply follows a weathering pattern. Dr Fingas observed that Dr Stout’s results showed that, after 84 days of biodegradation and evaporation, the remaining oil from the wax-enriched oil fraction consisted of 28% saturated hydrocarbons including wax, 1.4% of aromatics, and 71% of other compounds. Thus, the portion containing the waxes only constituted 28% of the mass. Many of the larger aromatics, known for their persistence and toxicity, were retained in that fraction.

Dr Stout’s response

722    Dr Stout defended his long-term weathering study. He contended that laboratory weathering studies are commonly used in oil spill science to understand the weathering that occurs. He said that the conditions of his long-term study mimicked, as practically as possible within the constraints of the laboratory, those conditions that existed during the actual Montara oil spill, by using indigenous bacteria in a concentration that occurred naturally in the Timor Sea, and artificial sunlight. That said, Dr Stout remarked that biodegradation and photo-oxidation did not “wait” to commence only after the spilled oil had evaporated for 7 days, as did his study. Rather, in the real world, bacteria and sunlight would begin acting on the spilled oil simultaneously as it evaporated. In making this observation, Dr Stout was expressing caution that his measurements were likely to have underestimated the extent of weathering on the composition of the spilled oil and the concentrations of hydrocarbons in that oil.

723    Dr Stout also acknowledged that he could not replicate the process of wax-agglomeration and separation during his long-term study. He repeated his earlier contention that one of the Leeder Reports reported on floating wax-rich particles that exhibited a greater loss of PAHs than he was able to achieve in the laboratory.

724    Dr Stout also acknowledged that the rates of weathering of different forms of Montara oil during the spill would have varied and that the rates he observed were specific to the particular fractions he studied. He argued, however, that the specific wax-enriched oil fraction he studied was but one fraction along a “continuum of increasing wax-enriched particles that formed during the actual spill”.

725    Dr Stout accepted that chemically-dispersed droplets would not have been able to diffuse/dilute within the glass vessels he had used, but argued that this meant it was likely that chemically-dispersed Montara oil weathered more quickly during the actual spill than he observed in his long-term study. On the other hand, he said it was likely that thicker floating oil slicks, not thin sheens, weathered more slowly during the oil spill than the sheens he used in his study.

726    With respect to the non-inclusion of fresh Montara oil in his long-term study, Dr Stout argued that, as evaporation quickly altered the chemical composition of spilled Montara oil, it was “completely reasonable” that he should commence his long-term study with evaporated oil, not fresh oil.

727    With respect to the addition of the bacteria to his study, Dr Stout stressed that only indigenous bacteria isolated from the Timor Sea had been added using standard microbiological procedures. He explained that there was nothing special about the bacteria he had used and that such additions were not uncommon in laboratory weathering studies because of the “difficulty-to-impossibility” of keeping bacteria alive during the shipping of seawater samples, in this case from Australia to the United States of America. Thus, he argued, the biodegradation during his long-term study was not significantly affected by the addition of live bacteria at their real-world concentration.

Conclusions on Dr Stout’s long-term weathering study

728    Dr Stout’s long-term weathering study was an attempt to mimic, in the laboratory, the chemical composition of spilled Montara oil that had weathered over an 84 day period. His study was based on the three forms of Montara-2 oil he had prepared. The applicant did not challenge the skill with which this work was carried out or the actual results that were obtained. What the applicant did challenge, through Professor Ball and Dr Fingas, was whether the results obtained could be accepted as reliably representing the actual chemical composition of weathered Montara oil given that the experiments were: (a) benchtop experiments; and (b) carried out, in the case of the wax-enriched oil fraction, on starting materials that had been derived from a process (chilled centrifugation) that was certainly not a process to which the spilled Montara oil had been subjected in the field. The latter point was the gravamen of Professor Ball’s and Dr Fingas’ criticism, that no control using fresh Montara oil had been provided to show that the wax-enriched fraction was, in fact, representative of the spilled oil, or any part of the spilled oil, that had weathered.

729    Dr Stout argued that his wax-enriched oil fraction was one fraction along a continuum of increasing wax-enriched particles that formed during the actual spill. He also argued, based on one of the Leeder Reports, that the loss of PAHs observed from his study was likely to be conservative.

730    Even though laboratory studies have their limitations in mimicking the real world, as Dr Stout himself acknowledged, I certainly would not dismiss his work because it was based on such studies. However, there is no reason to think that, at the time of the spill, the Montara oil separated into the particular wax-depleted and the wax-enriched fractions which Dr Stout used as two of the starting points for his long-term weathering study. It is, of course, possible that the wax-enriched fraction obtained through chilled centrifugation was representative of some part of the spilled oil that had weathered. I could not say otherwise. Indeed, the Leeder chromatograms provide some broad support for that conclusion. But if Dr Stout’s wax-enriched fraction was, as he argued, on the continuum of increasing wax-enriched particles that formed during the actual spill then, absent a control, there is no objective standard by which it can be known where that fraction sat on that continuum or to what extent it might represent the chemical composition of the spilled oil that, on the applicant’s case, reached the shores of Rote/Kupang. This is the principal reservation that I have with the results of the long-term weathering study carried out with respect to the wax-enriched fraction.

731    I accept that, as a general trend, the weathering of the oil spilled from the H1 Well would have resulted in the rapid loss or depletion of MAHs and PAHs. But the three forms of oil studied by Dr Stout do not represent all the weathered states in which the spilled oil might have reached the coastal waters of the Rote/Kupang region. Therefore, his results are not conclusive of the PAHs present in that oil. I note, in any event, that in Dr Stout’s long-term study some PAHs remained in the three forms of oil after 28 days, with the weathering of the remaining PAHs proceeding only incrementally, particularly in the wax-enriched fraction.

732    Another reservation I have with respect to the long-term study is the addition of bacteria, including after 21 days, to the materials that Dr Stout was studying. The reason why the bacteria were added has been explained. I do not for one moment think that they were added to deliberately enhance bioaugmentation. Based on Dr Stout’s undoubted experience, I accept his evidence that the addition of such bacteria is not uncommon in laboratory weathering studies. However, no estimation of the actual numbers of activity of the specific isolate in fresh seawater was carried out. It is not possible to assess the impact of this addition on the biodegradation process. Therefore, the extent to which that addition might have influenced the results that were obtained does introduce a further element of uncertainty as to the extent to which the long-term study did, in fact, mimic the weathering of Montara oil that took place during the course of the spill.

733    The applicant submitted that the likely consequence of Dr Stout adding bacteria is that a greater amount of oil was biodegraded than would have been the case with the oil that was spilled from the H1 Well, thereby reducing the quantities of compounds present in the studied fractions over the entire course of the 84 day long-term weathering experiment. The applicant submitted that to the extent that the Toxic Units Model deployed by Dr Maki, discussed below, included data from the long-term weathering study to determine the acute and chronic toxicity of the Montara oil on particular days, it underestimates the toxicity of the oil.

734    Having regard to the other evidence before me, the reservations I have expressed do not, ultimately, affect the conclusions to which I have come on the principal factual questions for decision in this case—whether Montara oil spilled from the H1 Well blowout reached the Rote/Kupang region and, if so, whether that oil damaged the seaweed crops in that region. This is so even though Dr Stout’s work was directed, in part, to assist in determining the toxicity of fresh and weathered Montara oil, a topic to which I will return.

735    One matter that is not in doubt is that the wax content of Montara oil increases with weathering. As it weathers, Montara oil will result in the formation, and separation from the oil, of wax-rich residues which, in physical appearance, can be described by expressions such as blocks, clumps, balls, and globules. These aggregates will be variously coloured, depending on the extent of the weathering that takes place.

The LEMIGAS analyses

736    At this point, it is convenient to discuss analyses carried out by LEMIGAS. As I have already noted, LEMIGAS is an Indonesian governmental oil and gas research organisation. The analyses were carried out on material (sediments, coral, tar balls, and mangrove root) collected from certain locations in the Rote/Kupang region in December 2017 and April 2019, long after the Montara oil spill. Those undertaking the analyses concluded that four of the samples collected in 2017 and 2019 contained Montara oil.

737    The LEMIGAS analyses were introduced into evidence by an unusual route. The applicant served the LEMIGAS analyses on the respondent. Dr Fingas also prepared a report in which he concluded that even more samples taken by LEMIGAS showed, or possibly showed, the presence of Montara oil. Specifically, Dr Fingas concluded that, when properly analysed, the LEMIGAS data revealed four samples showing the presence of Montara oil and eight samples showing the possible presence of Montara oil. In a responding report, Dr Stout criticised the LEMIGAS analyses and findings, as well as Dr Fingas’ findings. In the end, the applicant made a forensic decision not to rely on the LEMIGAS analyses. This decision was made on the basis that, when all of the evidence on this particular topic was taken into account, the Court was unlikely to conclude, on the balance of probabilities, that the samples taken by LEMIGAS in December 2017 and April 2019 contained Montara oil. I note for completeness that LEMIGAS also analysed three tar balls sampled in 2009 (which had been collected approximately 150 kms north of the Montara wellhead platform) and concluded that they contained Montara oil. There is no dispute about that finding.

738    The respondent took a different view of the forensic utility of this evidence and decided to pursue the topic. It tendered, without objection, certain annexures from a report prepared by LEMIGAS, and an expert report from Dr Stout dated 25 September 2019. The applicant tendered an amended report from Dr Fingas dated 3 December 2019.

739    Dr Stout analysed the LEMIGAS data using a protocol prepared by the Centre for European Norms called CEN 15522 – 2. Dr Stout’s evidence was that this protocol provides the state-of-the-art methodology for oil “fingerprinting”. “Fingerprinting” oil involves comparing the ratios of certain, highly-specific biomarkers present in the oil with the known ratios of those biomarkers in a reference oil (here, Montara oil). One of the ratios used by LEMIGAS, and which featured in the evidence before me, was C24 tetracyclic terpane/ C30-diahopane. In its analyses, LEMIGAS referred to C24 tetracyclic terpane as compound “8” and to C30-diahopane as compound “D”.

740    Using the CEN 15522 – 2 protocol, Dr Stout concluded that none of the samples collected by LEMIGAS in December 2017 and April 2019 contained Montara oil. LEMIGAS did not use this protocol or, it seems, any similarly known protocol.

741    With respect to the LEMIGAS analyses, Dr Stout said that, because of insufficient quality control, the collected chemical fingerprinting data lacked reliability at the outset. The data did not meet typical environmental geochemistry standards. Further, according to Dr Stout, the protocol that was used to compare the samples’ fingerprints with Montara oil’s fingerprint was poorly explained and over-simplistic. He said that it did not meet the scientific rigour of any of the published oil spill identification protocols. Further, according to Dr Stout, those undertaking the LEMIGAS analyses ignored key diagnostic features of Montara oil when concluding that the four samples “matched” Montara oil.

742    Dr Stout also observed that the concentrations of TPHs measured by LEMIGAS were likely to be biased (high) due to the presence of elemental sulphur and/or the presence of naturally-occurring biogenic hydrocarbons. Those undertaking the analyses appeared to have assumed that TPHs represent only petroleum hydrocarbons. As Dr Stout explained:

[81]     The problem of biogenic hydrocarbons is very important when measuring TPH in soils or sediments. This is because some soils or sediments, such as a seagrass bed or mangrove forest, will contain an abundance of decaying plant, algae, and bacteria biomass. This decaying biomass is what makes such soils/sediments dark in color. This decaying biomass contains naturally-occurring hydrocarbon derived from plant waxes, terpenes, fats, and lipids that were part of the plants, algae, and bacteria when they were alive. These hydrocarbons are soluble in n-hexane and therefore contribute to TPH, even though they are not petroleum. This is a well-recognized issue in environmental geochemistry since sometimes “clean” sediments are erroneously considered to be contaminated with petroleum, when they are not.

743    Dr Stout said that simply detecting TPHs in an ad hoc collection of sediments is not a reliable indicator of oil contamination. In any event, Dr Stout said that, given the low concentrations of TPHs measured in most of the collected sediments, the actual concentrations of any oil in each sample was equivocal (i.e., clearly low and perhaps absent).

744    In summary, Dr Stout concluded that:

[12]      the limited ad hoc studies conducted in 2017 and 2019 provide no chemical evidence of any impact to the shorelines of West Timor, Semau Island, and Rote Island by Montara crude oil. The results do show multiple “non-Montara” oils are present at mostly low concentrations throughout the area – likely owing to pervasive natural oil seeps in the region.

745    Dr Stout’s criticism of Dr Fingas’ evaluation of the LEMIGAS data was more trenchant. Dr Fingas employed regression analysis to compare the LEMIGAS biomarker ratios. Dr Stout described this method as “unique, unpublished, single-tiered and arbitrary”. Dr Stout observed that no past or current oil spill protocols rely on Dr Fingas’ method and that Dr Fingas himself had cited no supporting references in the oil spill literature. Dr Stout argued that the “matches” that Dr Fingas achieved when carrying out his evaluation were meaningless. He argued that, by Dr Fingas’ method, false positives were easily obtained. He argued that Dr Fingas had used duplicated, re-scaled and non-diagnostic ratios and that Dr Fingas had also ignored, or diluted the importance of, some key diagnostic features.

746    In response, Dr Fingas maintained his findings. He disputed that he had used duplicated ratios. He also emphasised that one problem with oil “fingerprinting” was that the ratios of the biomarkers used for comparison change with weathering. Dr Fingas said that there are no current statistical methods to correct for the effects of the long-term weathering on these biomarkers. He argued that the biomarkers that Dr Stout had used are good for short-term comparison (meaning “good” for days or maybe weeks). He said that highly-weathered samples of the same oil are hard to identify using these biomarkers. He said that, in the case of the LEMIGAS samples, there may have been extensive biodegradation of the biomarkers that were used for comparison and that, in fact, more of the LEMIGAS samples may have contained Montara oil.

747    For his part, Dr Stout said:

65    I considered the possible effect of weathering of the Montara oil on its biomarker distribution even though most biomarkers are considered extremely resistant to weathering on environmental timescales. Some of the “key” triterpene biomarkers I consider to be particularly diagnostic of the Montara oil are actually known for their resistance to weathering. For example, the C24 tetracyclic terpane … and C30-diahopane … are both reportedly even more resistant to biodegradation than hopanes. Thus, the absence or reduced relative abundances of one or both these two “key” biomarkers relative to the Montara oil cannot be reasonably explained by biodegradation. Therefore, differences in these biomarkers’ relative abundances (versus the Montara oil) indicates the presence of a different type(s) of oil in many samples …

Conclusion

748    Based on Dr Stout’s evidence, I have strong reservations that the LEMIGAS analyses are reliable for the various reasons he advanced. I am certainly not satisfied that it has been shown that the samples collected by LEMIGAS in 2017 and 2019 did contain Montara oil.

749    Further, on this topic I prefer the rigour of Dr Stout’s methodology and analysis over Dr Fingas’ analysis. I am not persuaded that Dr Fingas’ regression analysis possessed the rigour to enable firm conclusions to be drawn. Further, Dr Fingas’ analysis did not account for the deficiencies in sampling to which Dr Stout referred. In other words, Dr Fingas’ analysis could be no better than the sampling that LEMIGAS carried out. In reaching my acceptance of Dr Stout’s analysis, I do not ignore the possibility that the weathering of biomarkers may have influenced the negative findings he made. However, I also think that it would be entirely speculative for me to conclude that, but for the effects of weathering, the samples would have shown the presence of Montara oil.

750    Plainly, it does not follow from these findings that Montara oil did not reach the shores of Rote/Kupang in the latter part of 2009 as a result of the H1 Well blowout. My conclusion is that the LEMIGAS analyses and Dr Fingas’ analysis do not establish the converse proposition.

The Sibert sample

751    Sedeoen is a small village on the western coastline of Rote. It stretches from the coast back (for about 1 km) to the main road that goes to Ba’a, the main town of Rote. The beach at Sedeoen is approximately 2 to 3 km north of Nemberala. It is a west-facing, sandy beach that leads to a tidal flat that extends for about 300 m before it reaches deeper water. There is a reef (called the Bombora) that is about 500 to 600 m from the high tide mark. There is another reef about 300 m further out, where some waves break. To the north and south there is a flat reef that runs all the way along the western end of Rote.

752    In 2009, Adrian Sibert lived at Sedeoen. He had a boat which he moored at the beach. In late September 2009, he was visiting Nemberala. He heard about oil washing up on the beach. He went to investigate. He gave this evidence:

45.     ... I walked to Nemberela beach in front of Anugerah homestay. This was about a five minute walk from Johnny's bar. By the time I got to the beach it was late afternoon, around 5 to 5.30 pm, just before sunset. I saw these oily mustard coloured globules on the beach along the tide line mark for as far as you could see. I could also see the same globules in the water within the lagoon in front of the Anugerah homestay. I walked along the beach for 100 metres observing these globules. I felt the globules washed up on the beach. They were leaving shiny patterns like a rainbow in the light of the setting sun. I had never seen anything like the globules before.

753    The next morning he decided to go fishing at Sedeoen. When he arrived at the beach, where his boat was moored, he saw globules along the high tide mark, the same as he had seen the previous evening at Nemberala. When he ventured out in his boat, about 50 m from the shore, he noticed foamy patches floating on the surface of the water. He said:

51.     I looked at the patches of foam more closely. I could see two different substances and dead fish within the white foam. The foam was floating on top, up above the surface water level. The pale milky-mustardy coloured globules were floating underneath the white foam and within the surface area covered by the foam. The white foam was a brighter colour than the globules. The globules were still floating but they were at the surface level of the water with the fish. There were about six dead fish floating in the patch of foam, which covered an area of about six metres in diameter.

754    Mr Sibert decided to take a sample. He went back to the beach where he picked up an empty 500 ml plastic water bottle. He said it was not unusual to find bottles on the beach at Sedeoen because the seaweed farmers used them as floats. He then returned to his boat and motored offshore. He said:

62.     In between the two reefs out the front of my property I could see a concentration of the foam patches. I could see many dead mackerel around half a kilo in weight and 20 cm in length and some dead bonito and oily globules floating in a large foam patch in that area. I decided to take a sample of the water from that particular patch of foam.

65.     In the middle of this foam patch I used my gaff to try to spread the foam apart to expose the globules underneath. My gaff is just a stick with a big hook on it attached to the end that I made for my fishing. I used my gaff to push the foam away to get a water sample with the globules in it.

67.     I placed the water bottle into the water to get the sample and I pushed three to four of the globules into the bottle for collection. When the globules were in the bottle they floated to the top and joined together.

755