BRINKWORTH & ORS v COPPING & ORS, BRINKWORTH & ORS v HIGGINS & ORS, BRINKWORTH & ORS v JUSTIN & ORS, BRINKWORTH & ORS v TRIMBOLI & ORS [2026] SASCA 50

On Appeal from SUPREME COURT OF SOUTH AUSTRALIA (THE HONOURABLE JUSTICE STEIN) CIV-21-011444, CIV-21-011906, CIV-21-012765, CIV-22-003113 Appellants: BENJAMIN RONALD JOHN BRINKWORTH, JOHN CHARLES FINNIS, ANGAS THOMAS KENNETH BRINKWORTH, PATRICIA ANN BRINKWORTH Counsel: MR A HARRIS KC WITH MR N FLOREANI KC - Solicitor: GILCHRIST CONNELL Respondents: DAVID COPPING & ORS (CIV-25-008515), KEITH CHARLES HIGGINS & ORS (CIV-25- 008542), MICHELLE CATHERINE JUSTIN (CIV-25-008631), TERENA TRIMBOLI (CIV-25-008652) Counsel: MR M WHITTEN KC WITH MS T FLAHERTY - Solicitor: HALL & WILCOX Hearing Date/s: 05/03/2026 File No/s: CIV-25-008515, CIV-25-008542, CIV-25-008631, CIV-25-008652 A SUPREME COURT OF SOUTH AUSTRALIA (Court of Appeal: Civil) DISCLAIMER - Every effort has been made to comply with suppression orders or statutory provisions prohibiting publication that may apply to this judgment. The onus remains on any person using material in the judgment to ensure that the intended use of that material does not breach any such order or provision. Further enquiries may be directed to the Registry of the Court in which it was generated. BRINKWORTH & ORS v COPPING & ORS, BRINKWORTH & ORS v HIGGINS & ORS, BRINKWORTH & ORS v JUSTIN & ORS, BRINKWORTH & ORS v TRIMBOLI & ORS [2026] SASCA 50 Judgment of the Court of Appeal (The Honourable President Livesey, the Honourable Justice Bleby and the Honourable Justice B Doyle) 14 May 2026 APPEAL AND NEW TRIAL - APPEAL - GENERAL PRINCIPLES - INTERFERENCE WITH JUDGE'S FINDINGS OF FACT - PROOF AND EVIDENCE - BURDEN OF PROOF TORTS - NEGLIGENCE - PROCEDURE AND EVIDENCE - EVIDENCE - ONUS OF PROOF AND STANDARD OF PROOF On 11 January 2021, a bushfire ignited on a rural property known as Westlands. It spread in a south- easterly direction destroying more than 14,000 hectares of land, structures, livestock and fencing, and threatening the township of Lucindale before it was contained the next day. Following a trial on liability only, the appellants (Westlands’ owners or their executors) were found liable in negligence and nuisance to the respondents, who owned neighbouring properties affected by the fire. This was on the basis that the fire originated from a vegetation heap which was lit in winter 2020 and was not sufficiently extinguished so that it continued to internally smoulder for several months (the ‘active heap’) until it transitioned into flaming combustion in dry, hot and windy conditions on 11 January 2021. The primary judge found that the active heap was constructed in a way that gave rise to a risk of long-term smouldering combustion and that if it had been constructed in accordance with the Country Fire Service (‘CFS’) Vegetation Pile Burning Code of Practice the risk of smouldering combustion would have been significantly reduced if not eliminated. Fire breaks would also have reduced the -- 1 of 69 -- risk and appropriate monitoring would have identified the active heap as still burning, so that steps could have been taken to properly extinguish it before the bushfire season started. The judge found that the appellants breached their duty of care to neighbouring landholders and that breach caused the harm suffered by the respondents. The appellants submit that: 1. the primary judge erred by treating the question of causation as involving a preference between the ‘active heap theory’ and an alternative cause proposed by the appellants as a plausible explanation for the origin of the fire, namely, that the fire spread from a tree that had been struck by lightning in November 2020 (the ‘lightning tree B theory’). They submitted the judge erred by considering first whether the lightning tree B theory was established and thereafter effectively treating the active heap theory as the likely cause, thereby reversing the onus of proof; 2. the primary judge’s acceptance of the active heap theory was contrary to an incontrovertible fact, namely, the Cheney, Gould & Catchpole backing fire spread rate (‘CG&C rate’) calculations, which were said to have the result that there was an enormous disparity between the northern extent of the fire as observed at 1.18 pm on 11 January 2021 and the extent of the fire that could have been reached over the relevant period applying the backing fire spread rate from the location of the active heap; 3. the primary judge failed to grapple with an irreconcilable conflict between the fire pattern indicators identified by one of the experts who the judge found to be reliable (Mr Woods) and the judge’s reasons for not applying the backing fire spread rate. Held, dismissing the appeal: 1. the primary judge’s acceptance of the active heap theory did not follow from a rejection of the lightning tree B theory as a probable cause of the fire. The judge found the latter not to be a plausible explanation for the fire. The judge was satisfied having regard to all of the evidence that the active heap was the likely cause of the fire’s spread; 2. the judge’s reasons for considering that the active heap was the origin of the fire were not glaringly improbable or contrary to any compelling inference or incontrovertible fact. The application of the CG&C rate was not an incontrovertible fact. Any ‘disparity’ between the observed extent of the fire and that which would be suggested by application of the CG&C rate was not as pronounced as contended for by the appellants. The judge’s reasons closely engaged with the potential application and implications of the CG&C rate before finding, on all the evidence, that it did not stand in the way of accepting other evidence that pointed to the active heap as the origin of the fire; 3. the primary judge considered the appellants’ criticisms of Mr Woods’ evidence in detail and her reasons did not fail to engage with those criticisms. The primary judge’s treatment of this issue was not inconsistent with her analysis of the CG&C rate; 4. the appellants have not demonstrated that the finding that the active heap was the origin of the fire was wrong. Civil Liability Act 1936 (SA) s 35; Uniform Civil Rules 2020 (SA) rr 218.17(2) and 218.18(1), referred to. AA v The Trustees of the Roman Catholic Church for the Diocese of Maitland-Newcastle (2026) 100 ALJR 170, [2026] HCA 2; Briginshaw v Briginshaw (1937) 60 CLR 336; Devries v Australian National Railways Commission (1997) 177 CLR 472; Fox v Percy (2003) 214 CLR 118; Higgins v Brinkworth [2025] SASC 104; Hutchinson v Van Den Berg [2024] SASCA 117; Kuligowski v Metrobus (2004) 220 CLR 363; Lee v Lee (2019) 266 CLR 129; Murray v Kickmaier [1979] 1 NSWLR 414; Rhesa Shipping Co SA v Edmunds [1985] 1 WLR 948; Singh v Bains [2026] EWCA Civ 408; Suttor v Gundowda Pty Ltd (1950) 81 CLR 418, discussed. -- 2 of 69 -- BRINKWORTH & ORS v COPPING & ORS, BRINKWORTH & ORS v HIGGINS & ORS, BRINKWORTH & ORS v JUSTIN & ORS, BRINKWORTH & ORS v TRIMBOLI & ORS [2026] SASCA 50 Court of Appeal – Civil: Livesey P, Bleby and B Doyle JJA 1 THE COURT: On 11 January 2021, a bushfire ignited on a rural property known as Westlands. The fire spread in a south-easterly direction destroying more than 14,000 hectares of land, structures, livestock and fencing, and threatening the township of Lucindale before it was contained the next day. 2 Following a trial on liability only, the appellants (Westlands’ owners or their executors) were found liable in negligence and nuisance to the respondents, who owned neighbouring properties affected by the fire. This was on the basis that the fire originated from a vegetation heap which was lit in winter 2020 and was not sufficiently extinguished so that it continued to internally smoulder for several months (the ‘active heap’) until it transitioned into flaming combustion in dry, hot and windy conditions on 11 January 2021. 3 The primary judge found that the active heap was constructed in a way that gave rise to a risk of long-term smouldering combustion and that if it had been constructed in accordance with the Country Fire Service (‘CFS’) Vegetation Pile Burning Code of Practice the risk of smouldering combustion would have been significantly reduced if not eliminated. Fire breaks would also have reduced the risk and appropriate monitoring would have identified the active heap as still burning, so that steps could have been taken to properly extinguish it before the bushfire season started. The judge found that the appellants breached their duty of care to neighbouring landholders and that breach caused the harm suffered by the respondents. 4 The appeal concerns only the judge’s reasoning that led her to conclude that the active heap was the cause of the bushfire. There are two main parts to the appellants’ submissions on appeal. 5 First, the appellants contend that the primary judge erred by treating the question of causation of the fire as one that involved a preference between the ‘active heap theory’ and a theory proposed by the appellants at trial that the fire had spread from a tree (‘tree B’) that had been earlier struck by lightning (the ‘lightning tree B theory’). The appellants submit that the judge erred by considering first whether the lightning tree B theory was established and thereafter effectively treating the active heap theory as the likely cause, thereby reversing the onus of proof. 6 Secondly, the appellants criticise the primary judge’s treatment of aspects of the evidence relevant to the likely origin of the fire. In relation to this part of the appeal, the appellants accepted that the finding about the origins of the fire were based on a combination of primary findings likely to have been affected by -- 3 of 69 -- [2026] SASCA 50 The Court 2 impressions about the credibility and reliability of witnesses formed by the trial judge as a result of seeing and hearing them give evidence. The judge had the opportunity to consider the lay and expert witnesses’ evidence in the context of a relatively long trial as it unfolded, and having regard to the ‘feeling of [the] case’.1 7 It follows that although the appeal court must undertake a ‘real review’ to determine whether the trial judge erred in fact, appellate restraint applies with respect to interference with the findings concerning the origin of the fire unless they were shown to be ‘glaringly improbable’ or ‘contrary to compelling inferences’.2 Findings of the relevant kind must stand unless it can be shown that the trial judge has failed to use or has palpably misused his or her advantage or has acted on evidence which was inconsistent with facts incontrovertibly established by the evidence or which were glaringly improbable.3 8 The appellants contended that the judge’s finding that the active heap was the origin of the fire was contrary to an incontrovertible fact, namely, the Cheney, Gould & Catchpole backing fire spread rate equation (‘CG&C rate’),4 or rather, the disparity between the extent of spread suggested by it and observations made by a first responder (Mr England) as to the extent of the northern spread of the fire ground a little while after the fire started. The appellants also contend that the judge failed to deal with what was said to be an irreconcilable conflict between the backing fire pattern indicators identified by one of the experts who the judge found to be reliable and the judge’s reasons for not applying the backing fire spread rate. 9 For the reasons that follow, the grounds of appeal are not established. Read as a whole, the judge’s conclusion that the active heap was the origin of the fire did not follow from a rejection of the lightning tree B theory as a probable cause of the fire. On a fair reading of the reasons, when the judge found that the lightning tree B theory was neither probable nor plausible, this conveyed that the theory was, having regard to all of the evidence, no more than a theoretical possibility. The judge was satisfied, having regard to all of the evidence, but particularly the evidence positively supporting the active heap theory, that the active heap was the likely cause of the fire. That was not on the basis that it was the more likely of the two hypotheses that were the focus of the trial. It was on the basis that the evidence led the judge to an actual sense of persuasion that the active heap was the cause of the fire. There is no significance in the fact that, in presenting her conclusions about the two live hypotheses, the judge referred first to the lightning tree B theory. The sequence in which the reasons are presented does not in this case suggest compartmentalised or sequential reasoning of a problematic kind. On the contrary, 1 Fox v Percy (2003) 214 CLR 118 at [23] (Gleeson CJ, Gummow and Kirby JJ). The benefit of ‘watching a trial unfold in real time’ was recently emphasised in AA v The Trustees of the Roman Catholic Church for the Diocese of Maitland-Newcastle (2026) 100 ALJR 170; [2026] HCA 2 at [63] (Gageler CJ, Jagot and Beech-Jones JJ). 2 Lee v Lee (2019) 266 CLR 129 at [55] (Bell, Gageler, Nettle and Edelman JJ). 3 Devries v Australian National Railways Commission (1993) 177 CLR 472 at 479 (Brennan, Gaudron and McHugh JJ). 4 Cheney, Gould & Catchpole, ‘Prediction of Fire Spread in Grasslands’ (1998) 8 International Journal of Wildland Fire 1 (referred to in these reasons as ‘Cheney, Gould & Catchpole’) at 5. -- 4 of 69 -- [2026] SASCA 50 The Court 3 the detailed reasons of the judge reflect an iterative approach, and leave no room for any suggestion that, in forming (as distinct from expressing) conclusions about a topic, the judge did not bear in mind the evidence relating to other matters potentially bearing on the topic under consideration. 10 The judge’s reasons for considering that the active heap was the origin of the fire were not glaringly improbable or contrary to any compelling inference or incontrovertible fact. The CG&C rate, or the spread of the fire suggested by its application to the fire, was not an incontrovertible fact. 11 The application of the rate was something contended for by the appellants and two of the experts. Its application, and in the alternative, the probative weight that could be afforded to it, was contested by the respondents and other witnesses, and it was in any event dependent upon the confidence that the court could have about the integers required for its application (critically, the period of time elapsing between ignition and a particular observation of the extent of the rear perimeter of the fire). 12 The judge’s reasons closely engaged with the CG&C rate and the judge ultimately decided for a combination of reasons that it did not stand in the way of her concluding, based on all of the evidence, that the active heap was the origin of the fire. That conclusion was not glaringly improbable. Nor was there any necessary inconsistency arising from the judge’s reliance upon evidence about fire pattern indicators and her rejection of the argument that the CG&C rate precluded the active heap being the origin of the fire. 13 The question of the origin of the fire required weighing a number of matters apart from those that arose by a comparison between the fire’s apparent progression, or its observed aftermath, and the competing points of origin. It also required consideration of the inherent likelihood of the proposed point of origin being a source of flaming combustion on 11 January 2021. The judge’s ultimate conclusions involved an assessment of a range of matters including a preference for experts who gave lengthy evidence. The reasons given by the trial judge (‘Reasons’)5 are cogent and thorough. The appellants have not shown that the judge’s conclusion on causation was wrong. 14 More detailed reasons for these conclusions follow. Structure of the judge’s reasons 15 The judge summarised her conclusions in the ‘Overview’ appearing at the outset of the Reasons in these terms:6 For the reasons I set out below, the applicants have proved on the balance of probabilities that: 5 Higgins v Brinkworth [2025] SASC 104. 6 Reasons [10] (footnotes omitted). -- 5 of 69 -- [2026] SASCA 50 The Court 4 • the active heap was lit in mid-2020; • the active heap was constructed in a way which did not meet the requirements of the CFS Vegetation Pile Burning Code of Practice (‘CFS Code’); • the active heap was constructed in a way which facilitated smouldering combustion within the core of the active heap; • steps were not taken to ensure the active heap was fully extinguished after it was lit or before the fire danger season commenced; • there were no fire breaks around the active heap; • the weather conditions on 11 January 2021 were hot, dry and windy; • those conditions resulted in material which had been smouldering internally in the active heap during the intervening period since it was lit becoming exposed, thereby coming into contact with oxygen; • the smouldering combustion converted into flaming combustion; • the flaming combustion ignited surrounding dry grass; • the flaming combustion spread in the hot, dry windy conditions causing the ensuing fire; • the respondents owed a duty of care to the applicants; • the respondents breached their duty of care to the applicants; • the respondents’ breaches of their duty of care were the cause of the fire; • the respondents created a nuisance which was the cause of the fire; • the applicants thereby suffered harm; • the respondents are liable for losses caused by the fire. 16 The balance of the Reasons was structured as follows: A Background B Principles of fire behaviour and investigation C Investigation of the fire D Experts and their qualifications E Expert reports F Methodology of fire investigation G Evidence relevant to competing theories H Assessment of witnesses -- 6 of 69 -- [2026] SASCA 50 The Court 5 I Causation J Negligence and nuisance – remaining issues Conclusion and orders 17 The fact that the judge’s analysis of causation preceded consideration of questions of negligence and nuisance reflected the way the parties presented their submissions and invited the court to proceed. It was acknowledged on all sides that the most critical issue was whether the active heap caused the fire. If it did not, that was the end of the claims. If it did, that went a significant way towards determining the claims because, on that hypothesis, a smouldering heap had been allowed to subsist into the dangerous fire conditions of January. 18 It was ‘Part I: Causation’ that was the focus of attention on appeal. However, much of the groundwork for the judge’s discussion in that part appears in the earlier parts of the Reasons. The judge’s summary and treatment of the evidence at a granular level was, by and large, unchallenged. 19 It is not necessary to lay out all of the detail of that groundwork, but it is helpful to explain more about the background facts, the course of the investigation and some generally accepted propositions about the behaviour of grassland bushfires, before turning to the expert and lay evidence relevant to causation. Background 20 The fire commenced on 11 January 2021 on Westlands. The first report to triple zero was at 12.30 pm and the Country Fire Service (‘CFS’) was alerted at 12.37 pm by several calls reporting smoke in three separate locations near the Mount Scott Conservation Park (‘Conservation Park’). 21 The attempts made by the CFS to control and contain the fire included the use of aerial bombers. The fire became very fast moving and more than 250 firefighters were engaged in attempts to slow the fire and protect Lucindale. The fire burned in a south to south-east direction towards Lucindale from the area of origin and ultimately destroyed 14,074 hectares of land, 27 structures, about 7,000 livestock and many kilometres of fencing. On 12 January 2021 at 12.30 pm the fire was classified as contained. 22 The fire scar showing the final extent of the fire’s spread is seen in the following map.7 7 Reasons [12]. -- 7 of 69 -- [2026] SASCA 50 The Court 6 23 Evidence about the weather conditions prior to and on 11 January 2021 was contained in Bureau of Meteorology (‘BOM’) records. There were BOM automatic weather stations at Cape Jaffa, Coonawarra, Naracoorte, Padthaway, Keith and Robe. Each recorded temperature, relative humidity, wind direction and so on. The first responder to the fire, Mr England, also gave evidence about the local weather conditions experienced on the fire ground.8 24 There was negligible rain in the area of the fire between 1 January 2021 and the date of the fire. The conditions on the day of the fire were hot and dry. BOM records, including from the Naracoorte and Keith automatic weather stations, recorded temperatures ranging between about 35.3 and 38.9 degrees Celsius between about 12.00 pm and 4.30 pm on 11 January 2021. The Naracoorte weather station recorded low relative humidity, between about 10 to 13 per cent, and winds generally from a north-westerly direction between 35 to 52 kilometres per hour. The Keith weather station recorded temperatures from about 36.7 to 39.4 degrees Celsius; low relative humidity of between about 10 to 12 per cent, and winds generally from a north to north-westerly direction to a west to north- westerly direction, between 31 to 35 kilometres per hour.9 The grassland fire danger index on 11 January 2021 peaked at over 70 at 4.00 pm.10 25 As will be explained, the weather conditions in November 2020 were also relevant because a storm on that day was posited as a source of a possible lightning 8 Reasons [21]-[22]. 9 Reasons [27]. 10 Reasons [28]. -- 8 of 69 -- [2026] SASCA 50 The Court 7 strike on tree B. The volume of rain associated with lightning strikes on that occasion was relevant to the potential for ignition and the prospect of no visible damage being caused.11 Daily observations at 9.00 am on 11 November 2020, the day after the storm and lightning strikes, recorded rainfall between 2.6 millimetres and 6 millimetres for the four closest weather stations near Lucindale. Rain up to 4 millimetres was recorded in Naracoorte on 12 November 2020.12 26 Westlands is used for broad acre sheep and cattle farming operations. Its total area is 3,103 hectares and, at the relevant time, it carried thousands of sheep and hundreds of cattle.13 27 The general area of origin of the fire was in a paddock located in a slight valley between two ridges approximately to the east and west and running uphill to the northern side. Westlands is oriented approximately north-south, with most of its area located south of the Conservation Park. The northern tip of Westlands is adjacent to the Conservation Park on the park’s eastern side. The general area of origin is depicted (by the diagonally striped semi-circle) in the following map.14 11 Reasons [23]-[24]. 12 Reasons [25]. 13 Reasons [19]. 14 Reasons [15]-[16]. -- 9 of 69 -- [2026] SASCA 50 The Court 8 28 An aerial image also assists in understanding the terrain and the approximate area of origin of the fire relative to the fire scar. This image represents a view of the fireground from the north.15 29 The handwritten notation ‘AOO’ represents the approximate area of origin. The Conservation Park can be seen in the bottom right (north west) part of the photograph. The photograph depicts what was described by one of the witnesses as an irregularly shaped ‘heel’. 30 The first responder, Mr England, responded to reports of smoke received by the CFS and went to the scene. Mr England was called by the appellants. He became an important witness. The judge’s acceptance of him as an honest, reliable and credible witness is not challenged on appeal.16 Mr England took photographs commencing at about 1.18 pm.17 The CFS engaged in suppression activities including at the ‘heel’ of the fire to prevent it travelling into the Conservation Park.18 31 The paddock in which the fire started was largely grassland with a number of trees, primarily eucalypt on sandy soil, on a limestone base.19 There were a number of vegetation piles in various locations across the paddock.20 15 Reasons [188]. 16 Reasons [768]-[769]. 17 Reasons [13]. 18 Reasons [14]. 19 Reasons [18]. 20 Reasons [17]. -- 10 of 69 -- [2026] SASCA 50 The Court 9 32 In the course of the trial, there were a number of locations in the area of the origin of the fire upon which attention was focused. In the end, the most significant were the locations of the active heap and tree B. However, there was also reference to the ‘geolocated tree’, ‘origin A’ and the ‘cool pile’. The relative location of these sites, albeit as noted on a map that pre-dated the fire, is shown below.21 33 The relative locations of tree B and the active heap are also shown in a blown- up image from 2019 that is related (in the image below) to the fire scar image reproduced earlier.22 21 Reasons [20], [208]. 22 Reasons [929]. -- 11 of 69 -- [2026] SASCA 50 The Court 10 34 Tree B is located to the north of the active heap. On appeal, the parties agreed that the distance between the two locations was in the order of 100 metres. 35 This difference in location between the two possible ignition points was of significance because, on the appellants’ case at trial, it was not feasible that the fire could have burned in a northerly direction (into the prevailing wind, as a ‘backing fire’) from the active heap to the point to which Mr England observed it had progressed and as depicted in photographs he took at 1.18 pm. 36 That distance was estimated by one of the experts to be about 160 to 170 metres.23 Based on allowing a lag time of between 15 and 45 minutes between the fire igniting and it first being reported (at around 12.30 pm), the fire had been active for between around 56 and 86 minutes at the time of Mr England’s observation. 37 In the appellants’ submission, Cheney, Gould & Catchpole’s paper on predicting fire spread in grassfires suggests a maximum spread rate of a backing fire of about one metre per minute, and a more likely rate of about 0.5 metres per minute (the CG&C rate). It will be necessary to return to this proposition, but the scatter diagram upon which the CG&C rate is based is reproduced here for context.24 23 Reasons [343]. 24 Cheney, Gould & Catchpole at 5 (Figure 2). -- 12 of 69 -- [2026] SASCA 50 The Court 11 38 The highest rate of spread depicted in that scattergram is about 0.06 kilometres per hour, which equates to one metre per minute. At higher fuel moisture contents, the data generally suggested a lower rate of spread. The approximate mean of the data points suggests a spread rate of around 0.03 kilometres per hour, which translates to 0.5 metres per minute. 39 If the fire was only active for between 56 and 86 minutes at the time of Mr England’s observations, it should only have been able to spread (as a backing fire) between 56 and 86 metres. If a spread rate of 0.5 metres per minute is used, it should only have been able to spread between 28 and 43 metres. The appellants highlighted the disparity between these figures and the actual distance of around 160 to 170 metres that would need to have been covered, if indeed the active heap was the origin point. 40 As will be explained, one of the respondents’ experts, Mr Marsden-Smedley, proposed a different spread rate for a fire that was partly a backing fire but partly a flank fire, namely 4.5 metres per minute. This ‘average’ was criticised by the appellants for various reasons and the judge did not directly rely upon it.25 That said, the judge found that a simple application of the CG&C rate might fail to adequately accommodate Mr England’s evidence about the zephyring of the wind in the immediate vicinity of the fire ground and its impact on fire direction and movement.26 There were other reasons, to which it will be necessary to return, why the judge did not consider that the CG&C rate required rejection of the active heap 25 Reasons [989], [996]. 26 Reasons [1020]. -- 13 of 69 -- [2026] SASCA 50 The Court 12 as the origin of the fire. These included that the precise time of the fire’s initiation was not actually known. Whilst the experts were content to work from common assumptions, namely, that the fire was detected and reported within 15 – 45 minutes of its ignition, the fire could well have ignited earlier.27 41 In order to engage in more detail with the evidence relating to the origin of the fire, it is necessary to summarise some of the principles of fire behaviour and investigation, the history of the fire investigation and the arguments advanced throughout the process of the obtaining and exchange of expert reports. Fire behaviour and investigation 42 Frequent reference was made by the experts and the parties to Cheney and Sullivan’s 2008 text, Grassfires: Fuel, weather and fire behaviour28 (‘Cheney & Sullivan’). The judge summarised some general propositions emerging from Cheney & Sullivan in terms which were not challenged on appeal:29 [62] The physical structure of the fuel bed determines fire behaviour. In grassland fires, the continuity of the fuel bed is the main characteristic that influences the spread of fire. Flame height is mostly influenced by the height of the grass and fire intensity is predominantly dependent on fuel load. The main ingredient in vegetation fuel is cellulose. When heat is applied, the fuel first dries out, then breaks down. If heating is slow and the temperature remains below about 250 degrees Celsius, the material will dehydrate and char, producing charcoal. If heating is fast and the temperature is higher than about 250 degrees Celsius, the cellulose produces a flammable compound which can in turn decompose into a range of flammable gasses. These gasses ignite and combust to produce flames which release heat. The heat produced will raise the temperature of adjacent unburnt fuel. After flaming is completed, the charcoal compounds or residual carbon left over from incomplete pyrolysis may combine with oxygen, resulting in smouldering combustion that can produce significant heat but little flame. [63] A fire that starts from a point under a steady wind will form a roughly elliptical shape. The front of the fire is referred to as the head fire. The flames are blown towards the fuel which ignites at the top and progressively burns down into the lower layers. The back of the fire is referred to as the backing fire. The backing fire moves into the wind with the flames leaning over the burnt ground. Backing fires ignite the fuel bed at or near its base and burn slowly but efficiently. The sides of the fire are referred to as the flanks of the fire. The edge of the flanking fire is generally parallel to the wind direction, meaning the flames lean generally along the flank. A major characteristic of a flanking fire in grass fuels is that it can become a heading fire and a backing fire in response to changes in wind direction and will have attributes of each. [64] In moving fires, different fire types occur at different positions along the perimeter. At different times, heading fires, flanking fires and backing fires may occur at any location around the fire perimeter, depending on fluctuations in wind direction. [65] All fires increase their rate of spread after ignition until they reach a quasi-steady rate or average rate of spread for the prevailing weather conditions. The rate of spread will remain the same while the average wind speed remains constant. The rate of spread will 27 Reasons [1016]. 28 Cheney and Sullivan, Grassfires: Fuel, weather and fire behaviour (CSIRO Publishing, 2nd ed, 2017). 29 Reasons [62]-[67], [69]-[73] (footnotes omitted). -- 14 of 69 -- [2026] SASCA 50 The Court 13 depend on wind speed and the width of the head fire at a constant fuel moisture content. If a fire is not affected by changes in wind direction it will maintain a narrow head fire. The head fire width required for a fire to attain its potential maximum rate of spread increases with increasing wind speed. The time a fire may take to reach its quasi-steady rate of spread can vary greatly. The time taken to reach the quasi-steady spread depends on fluctuations in wind direction. The growth of a fire over time cannot be accurately predicted. [66] The shape of a free burning fire is largely determined by wind speed, especially in the early stages of the growth of the fire. The fire will be more elongated and narrow in shape when the wind is stronger. Conversely, the fire shape will be wider if the wind speed is lower. A fire burning in no wind will have a circular perimeter. If wind direction varies, the length to breadth ratio of the fire will be lower. Fuel is discontinuous if there are bare earth areas between clumps of grass. Fuel continuity is the major fuel characteristic which influences fire spread. The cured state of grassland also has a major effect on fire spread. When grasslands are more than 95 per cent cured, they have almost reached their full potential for fire spread. [67] The most dynamic variable influencing grassfire behaviour is wind. According to the Cheney and Sullivan text, wind speed fluctuates widely over short periods and varies with height above the ground. It is important to use the same height for the measurement of wind speed above ground when relating wind speed to fire spread. Wind speed has a dramatic effect on heading fires but little effect on fires backing into the wind. The authors state the backing rate of spread is influenced by fuel moisture and strongly affected by minor changes in fuel continuity which have no influence on the forward rate of spread. … [69] Fires burning uphill burn faster than fires on a level ground. There is a complex interaction between wind and slope. Hills interfere with the flow of the wind field changing the speed, direction and turbulence of the wind. Wind speed increases with height above the ground. [70] The Cheney and Sullivan text states that grassfires can appear highly erratic with behaviour that is difficult to predict and is variable. Fires respond to changes in fuel and weather in different ways and those factors are highly variable across the landscape and over time. Rapid changes in wind direction and wind speed can immediately change grassfire behaviour. Wind can vary greatly, both in time and location, and sometimes at locations that are close to each other. The variation is a result of turbulence embedded in the wind flow that forms lulls and gusts. Grassfires respond almost immediately to changes in wind speed and direction. Fires can show a rapid increase in rate of spread in response to frequent changes in wind directions. [71] According to the Cheney and Sullivan text, an understanding of how grassfires burn can help an investigator recognise fire patterns and point them to the origin of a fire. Eyewitness observations and physical evidence alone can make it difficult to reconstruct fire spread. When investigating only the cause and origin of the fire, the Cheney and Sullivan text recommends considering information over the fire’s life to help ensure conclusions are consistent with observations made elsewhere and with known patterns of fire behaviour. The authors recommend the investigator draw together as much information as possible to explain how a fire starting at a particular location could logically develop under the prevailing wind conditions to burn out the area impacted. [72] The Cheney and Sullivan text suggests the first thing for an investigator to establish is the direction of the prevailing wind and any significant changes in wind direction with information should be sought from as many wind recording stations in the district as -- 15 of 69 -- [2026] SASCA 50 The Court 14 possible. Such observations should be matched to physical evidence left by the fire. Patterns of burning can provide specific information on local wind directions. For example, the pattern of charring can show the direction the wind was blowing when a tree was burned. [73] The authors observe that it can be straight forward to determine a fire’s origin if fire fighters arrive quickly and suppress the rear perimeter of the fire soon after ignition. However, if suppression at the back of the fire was delayed or burning out operations were conducted upwind of the suspected point of ignition, it may be very difficult. Single ignition points from back burning operations may leave a pattern identical to that of the original ignition. The Cheney and Sullivan text recommends that investigators use all indicators to establish the approximate location of a change from a backing fire to a heading fire in the direction of the prevailing wind. In a passage that featured significantly in cross- examination, the authors state as follows:30 If the time when the very back of the fire was suppressed is known, and the probable ignition time is also known, it is possible to use the fact that fires backing directly into the wind spread at a constant rate … to locate the probable region of ignition. The elliptical shape of the back flanks can help focus the search for the ignition source, but the timing of lateral spread abreast or ahead of the ignition point is difficult to determine because of alternating heading and backing behaviour of the flank fire. 43 As has been mentioned, an important part of the appellants’ arguments on appeal concerns the Cheney, Gould & Catchpole paper on predicting fire spread in grasslands. We return to this in more detail under the relevant grounds of appeal. 44 Two other texts that were extensively canvassed at trial and in the Reasons were the National Wildfire Coordinating Group’s Guide to Wildland Fire Origin and Cause Determination (the ‘NWCG Guide’) and the National Fire Protection Association Guide for Fire and Explosion Investigations (the ‘NFPA Guide’). 45 The NWCG Guide contains chapters on fire behaviour and fire patterns. As is explained in the Reasons: [87] Fire pattern indicator vectors are described as a group of individual fire pattern indicators located near each other which as a group reflect the fire spread vector within that area showing the direction of fire progression at that point. Fire pattern indicator vectors identify transition zones which can be subtle. The NWCG Guide describes the backing fire vector as characterised by slower spread, lower intensity, less damage relative to advancing and lateral areas and microscale fire pattern indicators. Lateral fire vectors are characterised by a rate of spread and intensity between that of the backing and advancing fire and indicators can have characteristics of backing or advancing, depending on fire behaviour circumstances. Higher intensity flanks may leave indicators consistent with advancing fire spread, with a more defined and narrower transition zone, but a lower intensity flank may leave backing type indicators exhibiting a more subtle and wider transition zone. [88] The NWCG Guide refers to a transition zone which is described as an area of directional change which can change from a heading fire to a flanking fire, a heading fire 30 Cheney and Sullivan, Grassfires: Fuel, weather and fire behaviour (CSIRO Publishing, 2nd ed, 2017) at 112. -- 16 of 69 -- [2026] SASCA 50 The Court 15 to a backing fire, or a backing fire to a flanking fire. According to the NWCG Guide, identifying transition zones is key to accurately interpreting fire patterns. Changes in wind direction will affect the spread direction of the head fire and transition zones. [89] The NWCG Guide explains that there are 11 categories of fire pattern indicators, the physical appearance of which will differ with the direction of fire progression (the vector). Fire pattern indicators are divided into microscale and macroscale fire pattern indicators. Macroscale fire pattern indicators are usually associated with larger objects or areas. [90] In considering fire pattern interpretation, the NWCG Guide sets out the following principles. • Single fire pattern indicators reflect the fire’s direction at a precise point and may be unreliable in the context of overall fire progression. Accordingly, the interpretation should be based on the majority of fire pattern indicators within an indicator category. • Using as many categories of fire pattern indicators as possible increases reliability and consequently interpretation should be based on a variety of categories. • As fire does not burn in perfectly straight lines, radical but brief directional changes may occur. Fire pattern indicators will align with the progression at the point of each indicator. A single fire pattern indicator may therefore be accurate within a 180 degree arc. • Fire pattern indicators should be interpreted within the context of fire behaviour principles. The fire behaviour context should be determined through weather observations, topography, reliable witness information and reconstruction of probable fuel conditions. • Fire pattern indicators will usually become less pronounced when approaching the ignition area. Most fires start with low intensity and will progress outwards from the ignition area with intensity usually increasing as the fire progresses outwards. In the initial area of combustion, most of the fire pattern indicators will be subtle. • Fire pattern indicators should be documented during the investigation including with photographs and directional flags. • One should work from the area of more intense burning to follow the fire’s progression back to the area of less intense burning to the ignition area. Transition zones, being areas of directional change based on variations in intensity, may outline the specific origin area and can be identified though the fire pattern indicators. [91] The NWCG Guide recommends avoiding attempts to prematurely locate the ignition area. This is in part because indicators become increasingly subtle closer to the ignition area. [92] The NWCG Guide explains that a fire pattern indicator’s vector can usually be determined by examining the appearance of a fire pattern indicator. While indicators accurately reflect fire behaviour at a particular point, the individual vector may not be consistent with general fire progression. The fire behaviour context is therefore important. [93] Fire pattern indicators include, relevantly, protection and angle of char. Fuel and non- combustible objects will be unburnt or will exhibit less damage on the side which is not exposed to the advancing fire. Comparing the charring, staining, ash and so on, on the -- 17 of 69 -- [2026] SASCA 50 The Court 16 exposed and non-exposed sides of fuels and non-combustible objects can demonstrate the fire direction. Protection is said to be most reliable in low to moderate intensity fires and can accurately show fire direction. [94] The NWCG Guide notes that due to the lower intensity of backing fires, microscale protection fire pattern indicators are generally located on more protected and smaller fuels and objects. [95] According to the NWCG Guide, the charred remains of grass stems will have different appearances depending on the direction of the fire’s travel and intensity. In advancing fires, the flames will attack the stem from the top and burn them to ground level, completely consuming all but the very base of the stem. The NWCG Guide states that a backing fire will burn stalks off at the base and leave unburned stalks on the ground with their heads pointing towards the ignition area. The flank fire may leave some stalks burnt clean and others on the ground. Grass stems are described as a usually very reliable indicator, segregating backing from advancing areas well and defining lateral areas. There are some exceptions. Grass stems that have fallen in the same direction due to high winds may not be reliable as a directional indicator but the amount of grass stem remains can be used to assist determining transition areas and fire intensity. [96] Angle of char fire pattern indicators are formed when fire burns up to, past and beyond standing fuel such as a tree. The flame height and angle corresponding with head, flank and backing fire directions and intensities will char the fuel at an angle compared to both the unburnt portion of the object and the slope. … [102] The NWCG Guide recommends that the investigator first identify an advancing area of the fire consistent with witness statements and the context. Once that area has been identified, fire behaviour context and macroscale fire pattern indicators should be used to follow the advancing fire backwards to its source. The guide cautions against starting the area of search for the general origin too close to the transition of advancing and backing vectors. The NWCG Guide recommends a process for working through the general origin area, including identifying and marking the location of each fire pattern indicator with a visible marker. Generally, a head fire indicator is marked in red, a flanking fire indicator in yellow and a backing fire indicator in blue. 46 The NFPA Guide covers similar topics, and with respect to fire pattern indicators, the judge summarised its contents in these terms: [111] The NFPA Guide refers to various fire pattern indicators. In a section on grass stems, it states that heading fire areas are generally characterised by an absence of residual grass stems while grass growing in clumps may not be fully consumed and may show protection patterns. In backing fire spread areas, and sometimes in flanking areas, flames will attack the stalk at the base, toppling the rest into the burned area with the remaining grass heads pointing generally in the direction from which the fire came. The NFPA Guide refers to char patterns which can assist in determining fire direction. The NFPA Guide states that ash is often deposited on the windward side of objects and can be used to reconstruct probable fuel volumes. However, ash indicators can degrade quickly when exposed to high winds or moisture. The NFPA Guide describes non-combustible objects or fuel shielding the unexposed side of a fuel from heat damage and thus exhibiting less damage on the side which is shielded from the advancing fire. Objects on the ground will protect the fuels on the unexposed side and surface fuels on the exposed side will show a clean burn line as opposed to the protected side where surface fuels will appear uneven. -- 18 of 69 -- [2026] SASCA 50 The Court 17 [112] The NFPA Guide indicates that information from first arriving fire fighters can determine the initial area of investigation. Those people can provide valuable information relevant to determining origin and cause. [113] A section of the NFPA Guide addresses general principles of burn pattern interpretation. It recommends basing interpretation on the majority of the indicators within an indicator category, the totality of indicators and fire behaviour principles. It states that a single indicator may only be accurate within a 180 degree arc and indicators will usually be less pronounced near the origin. It recommends that fire investigation work from the area of most intense burning following the fire’s advancing spread back to origin. 47 Three further academic works, summaries of which can be found in the Reasons, may be briefly mentioned.31 First, Steensland’s paper titled ‘Long-Term Thermal Residency in Woody Debris Piles’ (the ‘Steensland paper’) addressed the capacity for woody debris piles to maintain smouldering combustion for days, weeks or even months, before escaping. Secondly, the 2009 paper of Mäkelä titled ‘Attachment of Natural Lightning Flashes to Trees: Preliminary Statistical Characteristics’ addressed the effects of lightning on trees in different weather conditions. Thirdly, Parker and Babrauskas’ article titled ‘Validation of NWCG Wildfire Directional Indicators in Test Burns in Coastal California’ (the ‘Parker and Babrauskas paper’) involved a validation exercise of the reliability of fire pattern indicators. The investigation 48 South Australian Police (‘SAPOL’) officers attended to investigate the potential cause of the fire on 12 January 2021 and CFS investigators were also tasked to conduct an investigation. Those investigators gave evidence and their records and photographs were tendered as well as considered by the experts.32 49 On the evening of 11 January 2021, a SAPOL investigator (Brevet Sergeant Downs) met the initial CFS incident controller (Mr McLaren) at the property and they travelled to a valley near the Conservation Park which McLaren suspected to be the area of origin. He observed two piles of pushed up vegetation which he thought might have been used as burn-off piles. One was glowing and on fire and was considered an area of interest. 50 The following day Brevet Sergeant Downs met with Mr England whose focus as one of the first responders had been to control the edge of the fire to prevent it getting into the Conservation Park.33 Downs returned to the valley and the active heap was still smouldering and giving off significant heat, whereas another pile was cool. The active heap that had been glowing the previous night contained large burnt trees which had been piled on top of each other and heaped with sandy soil.34 He was confident the fire had started in the valley area and he thought the 31 Reasons [115]-[123]. 32 Reasons [124]. 33 Reasons [127]. 34 Reasons [128]. -- 19 of 69 -- [2026] SASCA 50 The Court 18 active heap was the most probable cause of the fire but was not 100 per cent certain. He recorded the cause of the fire as ‘undetermined’.35 51 The cool heap was removed from consideration as a potential source reasonably quickly because it had not been burned.36 The focus of the investigation was the active heap. Downs checked an online webpage to find any records of any active lightning strikes. Because this suggested none had been recorded, he removed lightning from his consideration. He was therefore not looking for evidence of lightning strikes.37 52 Brevet Sergeant Downs’ notes referred to a discussion with the owner of the property to the effect that the two heaps on the property were burnt in July 2020.38 That information came from Mr McLaren. Mr McLaren’s evidence was that he had been told this by the property manager, Mr Degoumois, but there was some contest about that at trial.39 Detective Brevet Sergeant McManus also gave evidence that Mr Degoumois had made a statement to that effect.40 Brevet Sergeant Lisa McGregor also gave evidence to that effect.41 53 Ms Roberta Rice was the primary CFS investigator. She attended the area of origin of the fire late on 11 January 2021 and again on 12 and 13 January 2021, and she prepared a ‘Bushfire Investigator’s Worksheet Bushfire Investigation’.42 Whilst the report described the fire cause as ‘undetermined’, it ascribed the probable cause of the bushfire as ‘rekindle of burn off heap(s), underground burning root system burnt to surface in grasses +/- combination of both’. It was noted that fire indicators (grass, char marks, leaf freeze) pointed towards an actively burning burn off heap near a damaged burnt gum tree, which was hot and ‘venting’ on the north side.43 The investigation report included a number of annotated photographs, including the following which showed the active heap and the nearby gum tree.44 35 Reasons [129]. 36 Reasons [135]. 37 Reasons [135]-[136]. 38 Reasons [130]. 39 Reasons [149]. 40 Reasons [154]. 41 Reasons [168], [174]. 42 Reasons [177]-[181], [201]. 43 Reasons [184]-[185]. 44 Reasons [189]. -- 20 of 69 -- [2026] SASCA 50 The Court 19 54 In her evidence, Ms Rice said that she reported the cause as being ‘undetermined’ because she was not 100 per cent certain.45 She accepted as a possibility, but was not convinced by, the suggestion that the fire started at a point further to the north of the active heap and burned under the influence of the wind to where she recorded a head fire at waypoint 148.46 She gave a detailed explanation in re-examination for why she was not convinced of that possibility.47 45 Reasons [261]. 46 Reasons [260]. 47 Reasons [262]. -- 21 of 69 -- [2026] SASCA 50 The Court 20 55 The report eliminated powerlines, electric stock fences, machinery, spontaneous combustion, and lightning (recorded as not present, based on incident management team and BOM checks) as possible causes of the fire.48 56 Many of Ms Rice’s observations formed part of the basis for Mr Woods’ expert opinions, and there was extensive consideration of them at trial, as well as in the Reasons.49 She identified particular locations as ‘waypoints’. Dr Marsden- Smedley later identified these on an aerial photograph reproduced earlier in these reasons.50 The experts 57 The respondents (the applicants at trial) relied upon expert evidence of Mr Richard Woods and Dr Jonathan Marsden-Smedley. Mr Woods had experience in bushfire management and investigation. Dr Marsden-Smedley’s expertise was in fire behaviour analysis. He was also a qualified fire investigator. 58 The appellants (the respondents at trial) relied upon expert evidence of Mr Timothy Cousins and Mr Roger Fenwick. Mr Cousins was an engineering systems failure analyst and disaster recovery consultant with a Graduate Certificate in Fire Investigation and a certification as a fire and explosion investigator and instructor. Mr Fenwick held a Bachelor of Science (Forestry) and had, amongst other qualifications, a Graduate Certificate in Bushfire Protection. 59 The appellants submitted at trial that the way in which the expert reports developed over time was of some significance. The judge summarised the reports in a broadly chronological fashion to demonstrate their development. For present purposes, a brief survey suffices. A more detailed analysis of the reports is contained in the Reasons. Woods’ first report 60 Mr Woods conducted a site visit on 5 April 2022 and examined the scene using the NWCG methodology. Whilst the scene had been subject to grazing impact, weathering and regrowth, he believed there was a sufficient number of fire pattern indicators to interpret the progression of the fire.51 61 He located, marked and examined macro scale fire pattern indicators (angle of char and angle of scorch in eucalyptus trees) moving back to the area identified as the specific origin area (the active heap) where he located lateral fire pattern indicators to the east and west of the pile on tree roots and limbs on the ground. He identified backing fire pattern indicators on the northern side of the active heap, mainly protection, on surface tree roots and small limbs on the ground.52 Mr Woods’ examination of the active heap was considered to be consistent with it 48 Reasons [186]. 49 Reasons [199]-[232]. 50 See above at [32]. 51 Reasons [286]. 52 Reasons [291]. -- 22 of 69 -- [2026] SASCA 50 The Court 21 supporting long term smouldering combustion and, ultimately, causing the wildfire in the conditions reported on 11 January 2021.53 He considered the circumstances identified in the Steensland paper were highly likely to have occurred at the active heap.54 Marsden-Smedley’s first report 62 Dr Marsden-Smedley conducted a site visit on 5 April 2022. The only indicators he felt comfortable to use to show spread direction were charcoal scars on trees. These supported that the fire spread to the south-east, away from the active heap.55 63 Dr Marsden-Smedley examined photographs of the active heap taken within 24 hours of the fire and considered it was not plausible that such heat and volume of ash could have formed within 24 hours of ignition. He considered the active heap must have been burning for several weeks, and almost certainly for months.56 64 In his view the only plausible cause of the fire was the active heap flaring up during the severe fire danger on 11 January 2021.57 Fenwick’s first report 65 Mr Fenwick was asked to assume various facts including that the active heap was not burned in mid 2020.58 Mr Fenwick observed that fire spread rate prediction is not an exact science, and that calculations are intended to assist fire control practitioners fighting an advancing fire. In his view, predictions depending on great precision from the applied model were likely to be unreliable.59 66 Mr Fenwick considered that the appearance and characteristics of the active heap were consistent with ignition in the previous several days and totally inconsistent with ignition six months previously.60 67 Mr Fenwick considered many of the usual fire indicators were unreliable at the site. He criticised Mr Woods’ report and disagreed with Mr Woods’ interpretation of particular fire indicators shown in photographs in his report.61 68 Mr Fenwick considered that the relative positions of the final rear perimeter of the fire and the active heap looked ‘wrong’. Using a backing rate spread of 0.7 metres per minute, he did not consider the active heap was the likely source of 53 Reasons [292]-[294]. 54 Reasons [294]. 55 Reasons [295]. 56 Reasons [297]. 57 Reasons [299]. 58 Reasons [301]. 59 Reasons [303]. 60 Reasons [305]. 61 Reasons [306]-[307]. -- 23 of 69 -- [2026] SASCA 50 The Court 22 ignition. He considered an ignition point further to the north was more likely, but he also thought a single ignition point was inconsistent with the shape of the fire.62 Cousins’ first report 69 Mr Cousins first inspected the site on 19 February 2022. He was critical of the reliability of key fire indicators and Mr Woods’ approach. Because of the passage of time, he thought only tree char lines were reliable and that material on the ground was not reliable because it could have been moved.63 Whilst generally he said tree char lines appeared to support a fire origin in the region of the active heap,64 for various reasons he considered that the active heap theory was flawed. The reasons included his analysis of satellite records of thermal anomalies which detected burning on 11 and 17 June 2020, but not subsequently, until 11 January 2021.65 He considered it unlikely the active heap could have burnt for the intervening period,66 and that satellite images from December 2020 suggested the active heap was unburnt at the time.67 He also considered that, using the CG&C rate, the fire origin must have been further to the north in order to explain the rear perimeter of the fire.68 70 He identified various trees he referred to as the geolocated tree and plausible origin trees A, B and C. He considered that tree B was located in the relevant range for a backing fire to have progressed to the location photographed by Mr England. The others were too far from or too close to the active heap.69 71 Mr Cousins reviewed lightning location data from ‘Weatherzone’ from November 2020 to January 2021. He concluded there were at least three lightning strikes where the 95 per cent confidence ellipses overlapped the active heap and accordingly concluded lightning could not be excluded as an ignition source.70 In his view, fires could survive for 70 to 90 days after a lightning strike.71 72 Mr Cousins considered it quite likely that tree B fell in high winds on the day of the fire allowing high winds to ventilate the smouldering in its roots, facilitating transition to flaming.72 73 In a supplementary report he was given further (different) information about the timing of Mr England’s photographs. Assuming a maximum detection delay of 45 minutes, he concluded that even if the fire spread from the active heap at a 62 Reasons [314]. 63 Reasons [318]. 64 Reasons [319]. 65 Reasons [322]. 66 Reasons [317]. 67 Reasons [323]. 68 Reasons [328]. 69 Reasons [329]. 70 Reasons [322]. 71 Reasons [327]. 72 Reasons [330]. -- 24 of 69 -- [2026] SASCA 50 The Court 23 backing fire rate of 0.83 metres per minute, it would not have reached the area shown in Mr England’s photograph by 1.18 pm.73 74 He adhered to the hypothesis that weather conditions likely caused one of the trunks of tree B to fall, exposing the root mass, effectively fanning smouldering and facilitating its transition to flaming.74 Marsden-Smedley’s reply report 75 In his reply report, Dr Marsden-Smedley responded to the contention of Mr Cousins and Mr Fenwick that there was insufficient time for a backing fire to have travelled to the location pictured in Mr England’s photographs at 1.18 pm. 76 He used weather data to determine an approximate angle between a direct line from the active heap and the location pictured by Mr England and the direction of the wind, concluding it varied from about 35 to 40 degrees. He considered the fire did not spread as a ‘pure’ backing fire but was about halfway between a flank and a backing fire. He pointed in Mr England’s photographs to smoke from the fire blowing in from the fire’s edge, consistent with an intermediate flank-back fire.75 Adopting a time frame of 56 to 86 minutes, Dr Marsden-Smedley’s estimates suggested that an intermediate fire spread of that kind could have travelled between 250 and 385 metres, meaning the lesser distance travelled of 160 to 170 metres was therefore possible, particularly if there were fuel discontinuities.76 77 It is convenient to repeat the primary judge’s summary of Dr Marsden- Smedley’s opinions on this topic:77 [346] Dr Marsden-Smedley referred to a book by Cheney and Sullivan which estimated the length to breadth ratio of grassland head and flank fires from wind speed. Dr Marsden- Smedley also referred to the Cheney, Gould and Catchpole paper which included an equation for predicting back fire spread rate from fuel moisture content. However, Dr Marsden-Smedley considered that back fire prediction equation had a poor fit to the data used to generate it. In his view, the data in the graph depicted in the Cheney, Gould and Catchpole paper contained a large degree of scatter in the relationship between fuel moisture and back spread rate. Accordingly, Dr Marsden-Smedley considered the equation in the Cheney, Gould and Catchpole paper should be used with caution as its predictions were associated with a high degree of uncertainty. [347] Given the angle between the wind direction and the line between the active heap and geolocated tree (about 40 degrees), Dr Marsden-Smedley considered the spread rate in that area was most likely to have been slightly less than the average spread rate of a back and flank fire. Using the formulae in the Cheney and Sullivan text and the Cheney, Gould and Catchpole paper, he estimated that between 12:00 pm and 1:00 pm on 11 January 2021, the flank and back fire spread rates would have been about 8.5 metres per minute and 0.5 metres per minute respectively which he considered produced an average spread rate of 73 Reasons [339]. 74 Reasons [340]. 75 Reasons [342]. 76 Reasons [343]. 77 Reasons [346]-[347] (footnotes omitted). -- 25 of 69 -- [2026] SASCA 50 The Court 24 about 4.5 metres per minute. Using that figure, Dr Marsden-Smedley estimated the fire would have spread between about 250 and 385 metres in the intervening time. Dr Marsden- Smedley considered those estimates were probably slightly high, with actual spread distances likely to be between 180 and 330 metres as a consequence of the average angle between the fire spread direction and the wind direction. He accepted the rate may have been further reduced by fuel discontinuities from animal tracks. Dr Marsden-Smedley considered it was highly likely the fire from the active heap would have burned to the location photographed by Mr England. Using those spread rates, Dr Marsden-Smedley estimated a fire starting at tree B would have travelled well beyond the location in Mr England’s photographs. 78 Dr Marsden-Smedley emphasised that the large amount of soil, log and branch material piled above ground level gave the active heap a high potential to sustain a smouldering burn for an extended period. He had observed smouldering for more than six months in similar situations.78 79 He also noted that tree B had live foliage on 12 January 2021. In his view, this was inconsistent with a tree burning for two months.79 Woods’ supplementary report 80 Mr Woods’ supplementary report responded to the assertion that the backing fire could not have covered the requisite distance. In Mr Woods’ experience, localised weather and fuel conditions can vary greatly from records of automatic weather stations. Having regard to fuel loads, conditions and weather, the distance was within the bounds of influence on fire behaviour posed by erratic localised weather and fuel conditions.80 81 Mr Woods also said that fully cured grassland fuels can emit limited volumes of light smoke initially which reduces the potential of observation of a significant smoke column, and local wind changes could inhibit a smoke plume or disperse smoke at different rates.81 It was feasible that the fire was burning undetected for some time prior to it being reported.82 82 Mr Woods addressed criticisms of his interpretation of fire pattern indicators. In relation to the backing fire spread rate calculations, he said there was limited global research for determining an accurate rate in grassland fuel. In his experience, the use of estimates to assess the timing of rate of spread of backing fire in grassland fuels was to be treated with great caution.83 83 Mr Woods said it was highly unlikely a lightning strike to a tree the size of tree B would hold over for 62 days without prior detection. He also considered 78 Reasons [348]. 79 Reasons [349]. 80 Reasons [350]. 81 Reasons [352]. 82 Reasons [360]. 83 Reasons [358]. -- 26 of 69 -- [2026] SASCA 50 The Court 25 that the fire pattern indicators did not support an advancing fire originating from tree B. Rather they supported the fire advancing towards tree B.84 84 Furthermore, Mr Woods disagreed with the theory that the fire might have been ignited by the collapse of a trunk of tree B. The presence of green foliage and unburnt bark on parts of the tree seen on the ground instead suggested that the limb fell after the fire had burned through the area.85 Reports in anticipation of and following conclave 85 Mr Cousins prepared further materials in advance of the experts’ conclave and a joint report prepared by the experts following the conclave contained a helpful body of common ground in relation to the mechanics of a smouldering heap and lightning strikes upon trees. These were comprehensively summarised by the primary judge.86 So were the subsequent reports prepared by Mr Fenwick, Mr Woods and Dr Marsden-Smedley.87 Evidence relevant to competing theories 86 After surveying these expert reports and canvassing the appellants’ criticisms of the CFS investigation methodology and the respondents’ experts, in Part G of the Reasons, the judge marshalled, on a topic-by-topic basis, the evidence bearing on: whether the active heap was lit in mid-2020; whether lightning may have struck tree B in November 2020 and, if so, whether it could have been the cause of ignition in January 2021 (the lightning tree B hypothesis); and the active heap hypothesis. 87 In respect of the lightning tree B hypothesis, as the Reasons recount, Mr Cousins’ hypothesis that the collapse of part of tree B may have led to the ignition was challenged in cross-examination by reference to a photograph that appeared to show that part of the tree had not fallen over at a point when the fire was well in progress. Mr Cousins then proposed alternative mechanisms, commenting ‘I put one up, but I’m happy to let that one go. We might modify it by the wind swaying a branch, cracking the ground, soil cracking, but I need a bit of extra oxygen or air into the smoulder area to assist, in order for that to transition to flaming’. As the judge observed, those mechanisms were not referred to in his reports.88 88 In respect of the active heap hypothesis, the judge described in some detail the observations of Mr England. The detail of this evidence had not been available for the experts to consider in their reports. The important features of that evidence included that when he first came over the hill and saw the fire, it was running up the gully towards the Conservation Park, but as he got closer to it, the wind swung 84 Reasons [378]. 85 Reasons [379]. 86 Reasons [380]-[417]. 87 Reasons [418]-[446]. 88 Reasons [561]. -- 27 of 69 -- [2026] SASCA 50 The Court 26 from a north-easterly around to a westerly to north-westerly direction. This stopped the run up the hill and that portion of the fire became a backing fire. This was occurring regularly.89 89 Mr England described seeing fingers of burnt grass where fire ran uphill as a heading fire, before the wind changed and turned the fire into a backing fire. He said the wind changed the fire direction about 10 times while he was in the gully. His observations included that he observed the fire ‘backing very strongly’ in some areas and ‘making quite good movement back against the wind’.90 90 Over the course of over 35 pages, the judge summarised in close detail the evidence given in chief and under cross-examination in relation to the active heap hypothesis (including the CG&C rate) given by each of the expert witnesses, and, in particular, in light of Mr England’s observations.91 Assessment of witnesses 91 Having comprehensively canvassed the evidence, in Part H, the judge made observations about her assessment of the witnesses. Relevantly, the judge accepted Ms Rice and Mr England as honest, reliable and credible.92 The judge found Mr Degoumois (who had given evidence that he had not lit the active heap) to be an unsatisfactory witness.93 92 In respect of the expert witnesses, the judge canvassed the parties’ submissions about the witnesses and concluded, with supporting reasoning, that: • whilst some aspects of Dr Marsden-Smedley’s evidence may have conveyed a sense of a lack of objectivity, the appellants’ characterisation of him as an unimpressive witness intent on serving the respondents’ interests should be rejected. Indeed, the judge had ‘no doubt’ that he had ‘significant and impressive experience which he brought to bear in reaching his opinions and giving his evidence’;94 • Mr Woods had appropriate qualifications and expertise and was clear and direct in his evidence, making concessions where appropriate and giving clear and cogent explanations for why he maintained the positions he took. He appeared impartial, objective and reasonable, and impressed the judge as knowledgeable, experienced, competent and reliable;95 • Mr Cousins had a tendency to be argumentative and slightly defensive, and left the judge with the impression (based on his posture and demeanour) that 89 Reasons [588]. 90 Reasons [589]-[590], [603]. 91 Reasons [605]-[747]. 92 Reasons [757], [768]-[770]. 93 Reasons [770]-[775]. 94 Reasons [796]. 95 Reasons [797]-[809]. -- 28 of 69 -- [2026] SASCA 50 The Court 27 he was avoiding eye contact with counsel and focusing upon the judge in order to reinforce his credibility. In some respects his evidence appeared to be searching for a theory to fit his hypothesis. Overall, the judge had significant concerns about relying on aspects of Mr Cousins’ evidence given his stated areas of qualification. The judge considered Mr Cousins lacked expertise in bushfires and bushfire management and she expressed some concerns about some of his methodology and ‘the extent to which he was objective and independent’;96 and • at times it was unclear whether Mr Fenwick was seeking to protect the appellants’ position rather than engage with material put to him that may give rise to challenges to foundations for his opinions. The judge was unimpressed by a number of indiscreet and bombastic written communications authored by Mr Fenwick which gave the very strong impression that he was aligning himself with the position of the appellants. Whilst the judge did not doubt Mr Fenwick’s significant experience and expertise, unfortunately, the correspondence gave rise to a suggestion of a lack of impartiality and objectivity.97 93 The judge indicated that whilst the assessments she made of the witnesses bore on the weight she accorded to their opinions, what ultimately mattered most when assessing the expert opinions was the extent to which they were based on relevant expertise and a reliable foundation.98 Conclusions on causation 94 In Part I (‘Causation’), the judge addressed the two live hypotheses as to the origin of the fire’s spread. Lightning tree B theory 95 The judge considered that, with reference to the evidence given about the probabilities of a lightning strike having hit the active heap, that the probability of a strike having hit tree B on 10 November 2020 was low.99 96 Next, the judge found that the investigators and experts did not identify evidence of damage to tree B consistent with a lightning strike.100 The judge did not, however, treat that as conclusive; it was possible for a strike to have occurred and left minimal or no evidence.101 96 Reasons [810]-[840]. 97 Reasons [841]-[858]. 98 Reasons [859]. 99 Reasons [869]-[870]. 100 Reasons [872]. 101 Reasons [873]. -- 29 of 69 -- [2026] SASCA 50 The Court 28 97 The judge found that the lightning strikes on 10 November 2020 were accompanied by rain. Those weather conditions were unlikely to be conducive to tree B igniting if it had been struck.102 98 There was no challenge made on appeal to any of these intermediate findings. 99 By reference to a number of photographs and the expert evidence given about what could be seen in and inferred from them, the judge found that: • when Mr England attended the fire ground in the afternoon, tree B was not visibly burning and none of its leaders had collapsed;103 • in the evening of 11 January 2021, there was a fire in the base of tree B;104 • one of the leaders of tree B collapsed after 8.30 pm on 11 January 2021 onto ground that was already burnt;105 • immediately prior to the start of the fire, tree B was living with live, green foliage, and it was intact and standing when the fire started;106 • prior to the fire, the tree was not showing signs of injury or structural deficiency which may have indicated the tree was burning internally;107 • if the tree had been burning within, it is likely to have shown some sign of distress such as the greying of foliage;108 and • tree B was not large enough to have contained sufficient material to sustain smouldering combustion for 62 days, whether within its root system or its lignotuber or both.109 100 These findings were not challenged in the grounds of appeal. That being said, in submissions, the appellants directed attention to a photograph depicting tree B after the fire and contended that the judge was wrong to find that prior to the fire the tree was not showing signs of injury or structural deficiency. They argued that the right hand leader can be seen to be in an advanced state of structural collapse and that the foliage associated with that leader appears less healthy than the adjacent leaders, suggesting compromised vitality consistent with prior trauma.110 As the respondents pointed out, those submissions were not anchored 102 Reasons [874]-[876]. 103 Reasons [880]. 104 Reasons [883]. 105 Reasons [886]. 106 Reasons [886]. 107 Reasons [887]. 108 Reasons [887]. 109 Reasons [891]. 110 Appellants’ written submissions [25]-[26]. -- 30 of 69 -- [2026] SASCA 50 The Court 29 in a ground of appeal.111 Further, the photograph in question was not put to the witnesses so that they could comment on that proposition and it was not a submission distinctly made at trial. The suggestions made for the first time at the hearing of the appeal should be rejected as being beyond the proper scope of the appeal.112 101 Turning to the mechanism of escape, the judge said: [893] The [appellants] pointed to a crack in the trunk or branch of tree B or the tree rocking in the wind as possible mechanisms for fire escape and contended that while it was not possible to say which occurred, the same could be said for the active heap hypothesis. [894] Neither Mr Cousins nor Mr Fenwick considered the mechanism of escape of the fire important. However, I consider the factual circumstances in which smouldering combustion may have escaped from within the tree to start the fire to be an important element of the tree B theory. Once Mr Cousins’ initial theory that the wind caused a limb to fall prior to the fire was disproven by the applicants demonstrating that the leader only fell after the fire commenced, Mr Cousins proffered only the most general explanation, that is, that somehow there was some form of opening by which internal smouldering combustion received a source of oxygen which enabled sparks or embers to escape. I found it most surprising that Mr Cousins considered the mechanism of escape to be irrelevant and that he took the view that whether he was changing his hypothesis in the witness box was only semantics. During cross-examination, Mr Cousins’ views of the mechanism changed; he “needed” the fire to start in the roots and get to the grass and there needed to be enough fuel to keep the holdover fire smouldering for 62 days; he was happy to “let go” aspects of the theory which simply represented a mechanism by which oxygen may have reached the root structure; he then hypothesised that swaying limbs could have cracked the ground, which was not previously referred to in his reports; and he took the view the mechanism did not concern him. I was concerned by his evidence that any change in opinion in mechanism did not warrant a supplementary report and that he was happy for it to be addressed in cross-examination. Together with the other matters to which I have referred, I consider this suggested a lack of appreciation of his obligations to the Court as an expert witness. [895] Mr Fenwick accepted there was an unsubstantiated leap on his part to rely on tree B having collapsed as the mechanism giving rise to the escape of fire. In cross-examination he said he did not consider root exposure a likely mechanism at all. Mr Fenwick referred to the photograph which showed a glowing centre in the bole of tree B as demonstrating there was a hole which would have given a source of oxygen and allowed embers to escape. This position differed from his previous acceptance of the mechanism as the collapse of tree B exposing a hole to the inside of the tree which allowed smouldering combustion to transition to flaming combustion and escape to surrounding vegetation. I do not consider that the change in mechanism constituted a “trivial variation” to that theory, contrary to Mr Fenwick’s position. Mr Fenwick was not able to explain why a mechanism reliant on a pre-existing hole in tree B would not have resulted in a bushfire in the period between November 2020 and January 2021, during which there were a number of high fire danger days. His explanation that it was just “good luck” was unconvincing. Similarly, Mr Fenwick was unable to explain how such a hole could have limited oxygen inflow but 111 The only part of the grounds directed to the substantive merits of the lightning tree B theory was the proposition in ground [3.4], that ‘[t]here was another possible source of ignition of the fire, namely a lightning strike, which had been wrongly excluded in the initial investigation and never properly investigated at the time or since’. 112 Suttor v Gundowda Pty Ltd (1950) 81 CLR 418 at 438 (Latham CJ, Williams and Fullagar JJ). -- 31 of 69 -- [2026] SASCA 50 The Court 30 then opened up sufficiently to increase oxygen inflow on 11 January 2021. The inability to provide any reasonable explanation for the mechanism by which smouldering combustion within the tree, whether its roots or bole, escaped in light of the state of the tree before, during and after the fire remained a significant flaw in the tree B theory. [896] In cross-examination Mr Cousins candidly explained that in hypothesising the lightning tree B theory as plausible, he did not equate that to probable, but rather that it could be anywhere on a spectrum from “possible” to “absolute”. He could not assign a figure or probability to the hypothesis. Mr Cousins’ willingness to proffer the hypothesis absent any associated, general estimate of the associated likelihood of that theory detracted significantly from the tree B theory. 102 The judge then considered the fire pattern indicators, noting, amongst other things, that none of the experts observed any advancing pattern indicators that would support the idea that the fire progressed from tree B towards the active heap.113 103 The judge assessed the evidence relevant to whether ‘venting’ (a potential sign of long-term burning) had been witnessed in connection with tree B. There was some contention about this topic arising from the way in which Ms Rice’s notes compared with her spreadsheet. The judge concluded that the evidence as a whole did not support the conclusion there was venting in the vicinity of tree B and she declined to make that finding. However, she indicated that even if she was wrong in that conclusion, it would not have been sufficient to persuade her that smouldering combustion in the roots of tree B was the cause of the fire in light of all of the other evidence.114 The judge then addressed the rival positions concerning the significance of the apparent charring of some of the roots found on the fire ground and concluded that the state of tree B and its roots were consistent with the tree being burned during the fire as opposed to being the cause of the fire.115 104 The judge expressed her conclusion about the lightning tree B theory in these terms: [908] In my view, in light of the combined circumstances of: • the low likelihood of lightning striking the tree in light of the Weatherzone data; • the lack of any physical indication of tree B being struck by lightning; • the low prospect of wet lightning causing ignition; • the low likelihood of ignition occurring in a living tree; • the state of the tree, including its green canopy, as observed before and after the fire, justifying the inference the tree was alive with green foliage prior to the fire; 113 Reasons [897]. 114 Reasons [902]. 115 Reasons [907]. -- 32 of 69 -- [2026] SASCA 50 The Court 31 • the fact the canopy of the fallen leader had green leaves which had not been burnt and were not showing signs of distress which would be expected if the tree was burning within; • the lack of sufficient fuel in the roots and lignotuber to sustain smouldering combustion for 62 days; • the fact that tree B collapsed after the fire and accordingly the collapse could not have given rise to the mechanism for escape of smouldering to flaming combustion; • the absence of a cogent and clear explanation for the mechanism for escape of any smouldering combustion in light of the evidence the leader fell after the fire; and • the absence of evidence of advancing fire pattern indicators to the south of tree B; [909] I am not persuaded that a lightning strike to tree B was a plausible or probable cause of the Lucindale fire. Active heap hypothesis 105 The judge addressed in some detail the question whether the active heap had been lit in June or July 2020 and ultimately found that it had been established, on the balance of probabilities, that the active heap was lit at that time.116 That reasoning (which, importantly, appears not to have depended upon any of the evidence suggesting the active heap was the origin point of the fire) was not challenged on appeal. 106 The judge’s reasons then proceeded to consider: • whether the investigation methodology was flawed – as to which the judge concluded it was not;117 • whether the criticisms by Mr Cousins and Mr Fenwick of the methodology of Dr Marsden-Smedley and Mr Woods were well-founded;118 • the contention that the fire could not have moved the requisite distance from the active heap to the location observed by Mr England at 1.18 pm – as to which the judge concluded that specific findings about fire ignition location could not be made by the simple application of the CG&C rate or other models, including Dr Marsden-Smedley’s average spread rate figure, because of the extent of variability of local conditions and the impact of those conditions on the fire behaviour and movement;119 and • the reliability of fire pattern indicators – as to which the judge ultimately concluded that they indicated an advancing fire vector in an approximately southerly vicinity of the active heap, lateral fire vectors to the side of it, and 116 Reasons [910]-[966]. 117 Reasons [968]-[975]. 118 Reasons [976]-[981]. 119 Reasons [982]-[1024]. -- 33 of 69 -- [2026] SASCA 50 The Court 32 a backing fire vector to the north of the active heap between it and tree B, consistent with a backing fire moving north from the active heap and an advancing fire moving south of the active heap.120 The fire pattern indicators were inconsistent with the fire having originated in or in the vicinity of tree B, and were consistent with the fire having originated in the vicinity of the active heap.121 107 The judge’s more detailed reasoning in respect of the CG&C rate issue and the fire pattern indicators will be considered under grounds 2 and 3 below. 108 The judge concluded that the fire originated from the active heap,122 before recording her findings: • as to the composition and size of the heap;123 • that its construction gave rise to the potential for smouldering combustion within the active heap;124 • that it was established on the balance of probabilities that there was long- term smouldering combustion within the active heap;125 • that it was established on the balance of probabilities that that long-term smouldering combustion continued within the active heap for approximately six months;126 and • that the smouldering materials within the heap were exposed to increased oxygen and transitioned to flaming combustion as a consequence of weather conditions on 11 January 2021.127 109 This reasoning was substantially independent of the issues of the backing rate and the fire pattern indicators, and there was no direct challenge made to it on appeal. 110 The judge then concluded her analysis of causation by saying:128 I therefore find on the balance of probabilities that: • smouldering combustion was continuing within the active heap on 11 January 2021; 120 Reasons [1026]-[1090]. 121 Reasons [1090]. 122 Reasons [1091]. 123 Reasons [1093]-[1098]. 124 Reasons [1109]. 125 Reasons [1110]-[1122]. 126 Reasons [1126]. 127 Reasons [1131]. 128 Reasons [1132]. -- 34 of 69 -- [2026] SASCA 50 The Court 33 • smouldering materials within the active heap were exposed to the surface of the active heap including through the effect of the weather conditions on the day; • those smouldering materials transitioned to flaming combustion when exposed to oxygen; • the flames spread to the surrounding vegetation in the hot, dry conditions and spread under the influence of wind to become the Lucindale fire; • the fire moved in the directions I have found above; • the resulting fire ultimately burnt the area depicted in the fire scar map at the commencement of my reasons. Appeal ground 1 111 This ground of appeal contends that the primary judge erroneously approached the question of the cause of the fire as if it were a binary choice by firstly rejecting the probability or plausibility of the lightning tree B theory (for which the appellants bore an evidentiary onus only), rather than determining whether, on the balance of probabilities, the active heap caused the fire, and thereby ‘reversed the onus of proof’. 112 There are several dimensions to this complaint. The first is that the judge wrongly approached the question of the fire’s origin by confining attention to two possible sites or causes. The second is that the judge considered that because she was not persuaded the lightning tree B theory was probable or plausible, this proved or tended to establish that the active heap was the cause. The third was that even if the lightning tree B theory was not shown to be probable or plausible, the judge was required to bring to bear that it remained a rational possible alternative in deciding whether the active heap theory was established. A related complaint was that the judge made a sequential evaluation of the competing hypotheses leading to an isolated treatment of each, where the lightning tree B theory was considered by itself, rather than as part of the assessment of the coherence and completeness of the evidence relevant to the active heap hypothesis. According to the appellants, by prematurely dismissing the lightning tree B theory, instead of considering it as part of the assessment of the active heap hypothesis, it resulted in the exclusion of the third possibility – that neither of the proposed causes was convincing. Principles of proof 113 At the level of principle, the following propositions may be identified as relevant to this case. (1) The respondents (the applicants at trial) carried the onus of demonstrating that the active heap was the cause of the fire.129 129 Civil Liability Act 1936 (SA), s 35. -- 35 of 69 -- [2026] SASCA 50 The Court 34 (2) This required the court to feel an actual sense of persuasion that the fire spread from the active heap.130 (3) It was not sufficient only to find that amongst the hypothesised causes, the active heap was the more likely, or the least unlikely.131 (4) A judge is not bound always to make a finding one way or the other on an ultimate issue in a civil case; a third alternative is to say that the party on whom the burden of proof lies in relation any averment made by them has failed to discharge that burden.132 (5) The appellants (the respondents at trial) were not required to demonstrate that the lightning tree B theory was the probable cause of the fire. Whilst it remained a realistic possibility, that possibility was potentially relevant to whether it was established, on the balance of probabilities, that the active heap was the cause of the fire. This is a reflection of the truism that no single item of evidence can have its proper significance considered if it is assessed disengaged from the circumstances of time, place and behaviour supplied by all the evidence, and that a judge’s fact-finding should not be approached in a segmented way.133 (6) However, that is not to say that if, once all the evidence was weighed, the lightning tree B theory remained a possible cause of the fire, this precluded the court being satisfied that the active heap theory was, on the balance of probabilities, the cause.134 114 There can be no real controversy about these basic principles. However, their application has to be understood in light of the parties’ forensic conduct of the case. Moreover, the question whether the judge departed from these principles is to be approached with a realistic appreciation that whilst all evidence is to be weighed relative to any other evidence bearing on the issue, written reasons for judgment have to be addressed in a particular sequence, and for convenience will usually deal with the merits or demerits of a body of evidence in one place, even though a final decision whether to act on the body of evidence will require consideration to be given to any competing bodies of evidence.135 As was recently observed:136 130 Briginshaw v Briginshaw (1938) 60 CLR 336 at 361 (Dixon J). 131 J D Heydon, Cross on Evidence at [9015]. 132 See, eg, Rhesa Shipping Co SA v Edmunds [1985] 1 WLR 948 at 955 (Lord Brandon of Oakbrook), referred to with approval in Kuligowski v Metrobus (2004) 220 CLR 363 at [60] (Gleeson CJ, McHugh, Gummow, Kirby, Hayne, Callinan and Heydon JJ). 133 Hutchinson v Van Den Berg [2024] SASCA 117 at [45] (Kourakis CJ, Lovell and S Doyle JJA). 134 See, eg, Murray v Kickmaier [1979] 1 NSWLR 414. 135 Hutchinson v Van Den Berg [2024] SASCA 117 at [59] (Kourakis CJ, Lovell and S Doyle JJA). 136 Singh v Bains [2026] EWCA Civ 408 at [136] (Miles LJ, Peter Jackson and Jeremy Baker LJJ agreeing). -- 36 of 69 -- [2026] SASCA 50 The Court 35 Any fact-finding exercise requires the court to marshal all of the evidence, including the witness testimony, the documents and the uncontested events, and to test each strand of evidence against the others, and against the inherent probabilities and motives of the actors. 115 In the end, however, conclusions have to be expressed or presented to the reader in some sequence even where each of them has been reached by an iterative review of the inputs relevant to each of them. 116 The extent to which a separate treatment of conflicting bodies of evidence may reflect or reveal segmented reasoning of a problematic kind depends on the extent to which the competing bodies of evidence engage with one another in a sense above and beyond that acceptance of one body of evidence involves a finding rationally inconsistent with acceptance of the other.137 The forensic context 117 The respondents (applicants at trial) alleged in their statements of claim that the fire escaped into surrounding vegetation from the active heap. Their case depended upon establishing that fact. 118 By their defences, the appellants specifically denied that the active heap was the cause or point of ignition of the fire and pleaded further that the fire ‘ignited by other causes, with the plausible cause being a lightning strike to a tree on the Blackford Property on or about 10 November 2020’ (the lightning tree B theory). That pleading gave fair notice to the respondents that a positive alternative cause would be promoted, but it did not relieve the respondents of their onus to establish that the active heap was the cause of the fire. Proof that the active heap was the cause of the fire would not necessarily be achieved by a rejection of any positive alternative cause. Having said that, there could be no actual sense of persuasion that the fire spread from the active heap unless any alternative cause was addressed and, in all likelihood, rejected. 119 The way in which the defence was expressed likely reflected the conclusion that had earlier been expressed by Mr Cousins in his report of 29 August 2022. He concluded that a plausible alternative fire cause and origin was a lightning strike to the tree in question. 120 In their joint report, the experts relied upon by the parties agreed that all other plausible fire ignition sources had been eliminated and that the only two possible sources were the active heap or a lightning strike.138 121 In his trial evidence, Mr Cousins did not go so far as to say that a lightning strike to tree B was the probable cause of the fire but he considered it was plausible, which he used to mean more than merely possible and to express probabilities above possible and through to certain. 137 cf. Hutchinson v Van Den Berg [2024] SASCA 117 at [65]-[67] (Kourakis CJ, Lovell and S Doyle JJA). 138 Reasons [9]. -- 37 of 69 -- [2026] SASCA 50 The Court 36 122 The respondents’ closing submissions noted that all other potential causes of ignition of the fire were eliminated by all experts and investigators. They contended that the lightning tree B theory did not withstand scrutiny and was no more than a theoretical construct consisting of a series of interdependent speculative improbabilities. They contended that the active heap was not only the most probable, but the only plausible cause of the fire. 123 The appellants’ closing submissions urged the court not to find on the balance of probabilities that the active heap was the cause of the fire. Those submissions went on to say that although it was not necessary for the court to do so, it could find on the balance of probabilities that tree B was a suitable host tree within which smouldering combustion could have harboured for 62 days following a lightning strike, and that tree B was a plausible origin of the fire. 124 The appellants emphasised that the court was not bound to select one of two causes and that another possibility was that the court would not be satisfied as to the cause of the fire. The appellants supported that possibility with reference to authority. The judge’s approach 125 In the ‘Overview’, the primary judge noted that from the perspective of the experts, all other plausible fire ignition sources apart from tree B and the active heap had been eliminated.139 The primary judge observed that the parties’ evidence and submissions thus centred on only two potential competing causes of the fire. However, the judge qualified that observation by saying:140 Nevertheless, determination of the success or otherwise of the applicants’ case and the respondents’ potential liability does not reduce to assessing which of the active heap hypothesis and the tree B hypothesis is more likely. The applicants must prove on the balance of probabilities that the active heap was the source of the ignition, and, if so, the respondents were relevantly negligent or caused a nuisance which resulted in the applicants’ loss. 126 The appellants rightly accept that this was a correct approach. Their complaint appears substantially to depend upon the following propositions: • the judge addressed the lightning tree B theory first; • the judge concluded her consideration of the lightning tree B theory by stating that she was ‘not persuaded that a lightning strike to tree B was a plausible or probable cause of the Lucindale fire’141 (emphasis added); and • the judge then considered the active heap theory without explicitly weighing the possibility of the alternative lightning tree B theory as part of the reasons 139 Reasons [9]. 140 Reasons [9]. 141 Reasons [909]. -- 38 of 69 -- [2026] SASCA 50 The Court 37 why the active heap theory might not be established on the balance of probabilities. 127 As has been explained, within Part I (‘Causation’), the judge expressed her conclusions about the lightning tree B theory before addressing the active heap theory. However, she prefaced her analysis of the lightning tree B theory with the following introductory comments:142 [861] In order for the applicants to prove that the respondents’ negligence caused the fire, the applicants must prove that the negligence was a necessary condition of the occurrence of harm (factual causation); and that it is appropriate for the scope of the negligent person’s liability to extend to the harm so caused (scope of liability). [862] The applicants bear the burden of proving on the balance of probabilities any fact relevant to the issue of causation. Determination that negligence was a necessary condition of the occurrence of harm is factual, requiring proof by the applicants of relevant facts on the balance of probabilities. A determination that it is appropriate for the scope of the negligent person’s liability to extend to the harm caused is normative, turning on consideration of, among other matters, whether or not, and if so why, responsibility for the harm should be imposed on the negligent party. [863] The proof of facts requires the fact finder to have an actual persuasion of the existence of those facts before they can be found. An applicant can discharge the burden of proof by adducing evidence of a fact the existence of which, absent further evidence, is sufficient to justify drawing an inference that it is more likely than not that the event occurred or the state of affairs existed. A fact finder does not pass from conjecture into inference unless a fact is found which provides a reason for concluding it is likely that a specific event happened or a specific state of affairs existed. Proof of facts on the balance of probabilities can be established by circumstantial evidence by proof of primary or intermediate facts from which a further fact can be inferred. It is not necessary that all reasonable hypotheses consistent with the non-existence of a fact or inconsistent with the existence of the fact be excluded before the fact can be found. [864] The inference of causation may be drawn from all of the evidence, including expert evidence. Where competing causes are advanced by the applicant and respondent, the Court must be satisfied on objective and reasonable grounds that the case for accepting the suggested means of causation occurred is stronger than the contrary and the applicant must do more than prove circumstances which give rise to alternative inferences of equal probability such that the choice between them amounts to conjecture.. It is not sufficient to simply conclude that if an alternative postulated cause is less likely, the remaining postulated cause becomes the probable cause. [865] In this case there is no direct evidence of cause. Accordingly, I must make findings of fact and I must consider what inferences I can reasonably draw from the facts as found. [866] I turn now to address the questions which arise for consideration in determining whether the applicants have established on the balance of probabilities that the active heap was the cause of the Lucindale fire. 142 Reasons [861]-[866] (footnotes omitted). -- 39 of 69 -- [2026] SASCA 50 The Court 38 128 There was and could be no complaint about that articulation of the required approach. 129 The judge’s reasons then addressed the lightning tree B theory before they addressed the active heap theory. Again, however, the opening paragraph of the judge’s reasons in relation to the former serves to emphasise the approach which the judge considered was applicable. She said:143 The respondents proffered the lightning tree B theory, and thus bore an evidential burden of adducing evidence to support that theory which was posited as contrary to the applicants’ active heap theory. However, the applicants at all times bore the onus of establishing on the balance of probabilities their case that the cause of the fire was the active heap. 130 If it was an acceptable technique to separately express findings in relation to the two theories, one of them had to be set out first. In circumstances where the judge repeatedly emphasised that any consideration of the lightning tree B theory could not relieve the respondents of their burden as claimants, the fact that the lightning tree B theory findings were expressed first is not significant. The judge’s statement, at the end of her consideration of the lightning tree B theory, that she was ‘not persuaded that a lightning strike to tree B was a plausible or probable cause of the Lucindale fire’, cannot be understood as reflecting a departure from the approach explicitly adopted elsewhere. 131 It is important to note that the conclusion expressed was not simply that the lightning tree B theory was not ‘probable’. It was found not to be ‘plausible’. In context, that conveyed that the judge did not think the lightning tree B theory was a realistic possibility. It was untenable. The context includes the way in which the parties used the word ‘plausible’. But the surest indication of that meaning comes from the cumulation of intermediate propositions, set out earlier in paragraph [104] above, which together contributed to the conclusion. The judge had found that standing alone, each of several propositions, all of which needed to be true for the lightning tree B theory to have been correct, was unlikely. There were, on the judge’s assessment, compounding improbabilities. 132 There is no suggestion, in the analysis relating to the active heap theory, that the judge treated her findings against the lightning tree B theory being probable or plausible as having any particular weight in supporting her finding that the active heap was the origin of the fire. That is to say, whilst, on one view, as a matter of logic, the rejection of one of two potentially plausible origins as a plausible origin might have supported the likelihood of the remaining origin, the Reasons do not suggest that such an approach was adopted, or at least, that reasoning of that kind exerted any material significance on the judge’s reasoning or conclusion. The appellants’ submission that the existence of a low probability did not relieve the respondents of their burden nor ‘entitle the Court to treat the active heap theory as proven by default’,144 is undoubtedly correct. But the implicit suggestion that the 143 Reasons [867]. 144 Appellants’ written submissions [24]. -- 40 of 69 -- [2026] SASCA 50 The Court 39 judge treated the active heap theory as proven by default is, with respect, unsustainable. 133 The judge’s reasons for being satisfied of the active heap hypothesis involved consideration of some bodies of evidence that were relevant to both theories, such as the fire pattern indications, but were to a significant extent independent of the rejection of the lightning tree B theory. 134 Finally, it is necessary to consider the possibility that the judge’s satisfaction as to the active heap theory wrongly excluded the possibility that the lightning tree B theory might be correct. For the reasons given earlier, in concluding that it was not even ‘plausible’, the judge appears to have concluded that the lightning tree B theory was not a realistic possibility. It may be doubted whether, if that was all that it was, it could play any real role in tending against satisfaction on the balance of probabilities of the active heap theory. Certainly, a mere possibility that is not a realistic possibility does not preclude being persuaded to the requisite standard of something inconsistent with that possibility. In any event, there is no reason to think that, when concluding that the active heap theory was to be accepted, the judge had wrongly excluded from consideration the material that identified the lightning tree B theory as something that was, at least, a possibility. 135 We would reject any suggestion that, in the final analysis, the lightning tree B theory was found to be implausible without bearing in mind any weaknesses in the rival (active heap) theory, or that, in the final analysis, the active heap theory was found to be the cause without bearing in mind the lightning tree B theory. The structure of the Reasons shows that the evidence, witnesses and issues that bore on each theory were considered in detail, as were the parties’ submissions about them, before the theories were considered largely by reference to their own cogency, in Part I. The sequence in which the intermediate and final conclusions were expressed cannot be taken to reflect that the judge’s actual reasoning was in fact segmented or sequential in any problematic way. Indeed, the structure of the Reasons very much reflects an iterative process where the evidence and arguments about all issues relevant to causation were comprehensively canvassed (in Parts B, C, D, E, F, G and H) before conclusions were expressed about the two live theories (in Part I). 136 Finally, and in any event, as the respondents submitted, if it were to be accepted that the judge’s approach to these questions was methodologically flawed, establishing that proposition would not result in a grant of the order sought on appeal, which was that this Court enter judgment in favour of the appellants. Unless it were also shown that the ultimate outcome was wrong, the best that the appellants could achieve would be an order for remittal or re-trial. Such an order was not sought. Whilst that might not necessarily preclude relief of that kind being granted,145 we are not persuaded that there is a basis for it. 145 Uniform Civil Rules 2020 (SA), rr 218.17(2) and 218.18(1). -- 41 of 69 -- [2026] SASCA 50 The Court 40 Appeal grounds 2 and 3 137 Appeal grounds 2 and 3 were formulated in these terms: 2. The [judge] erred in rejecting, contrary to the incontrovertible facts, the Cheney Gould & Catchpole backing fire spread rate equation on the basis that it ‘may fail to adequately factor in consequences arising from the localised weather on fire behaviour as occurred in this case’, despite: 2.1 The [judge] rejected the Marsden-Smedley fire spread rate in reaching her ultimate decision that the point of ignition was the active heap (Reasons [996]); 2.2 The [judge] did not address the enormous disparity between the distance that the backing fire would have covered on the application of the Cheney Gould & Catchpole spread rate equations (namely 56 – 86 metres) and the distance actually covered as evidenced by the England photograph at 1.18 pm (namely 250 – 385 metres146). 2.3 More was required in order to reject the applicability of the Cheney Gould & Catchpole equation as a valid means of testing the active heap hypothesis, such as how the alleged localised weather effect could have brought about an increase in backing fire spread rate from the published equations by a factor of eight. The [judge] failed to address this fundamental question. 3. The [judge] erred in failing to conclude that she was unable to find on the balance of probabilities that the fire was ignited by smouldering combustion in the active heap because: 3.1 She was unable to reconcile the eight-fold increase in the distance of the backing fire from the authoritative publications on backing fire spread, even allowing for the effect of local weather conditions observed by England, and given that the fire reached its quasi-steady state within five to ten minutes of ignition; and 3.2 She was unable to reconcile the alleged impact of local weather conditions described by England on fire behaviour at the back of the fire, which was inconsistent with the prevailing wind and behaviour of the head fire, which had reached quasi steady state in five to ten minutes; 3.3 She was unable to explain the irreconcilable conflict between the backing fire pattern indicators identified by Woods and the backing fire spread rates from authoritative publications such as Cheney Gould & Catchpole which should have caused her to reject or materially doubt the reliability of Woods’ evidence of fire pattern indicators given the contemporaneous evidence in the CFS investigation that backing fire indicators were ‘confused’, the failure of Woods to obtain a representative sample of fire pattern indicators given indicators in multiple categories contrary to NFPA Guidelines and the fact that Marsden-Smedley did not use micro fire pattern indicators and prevaricated about the inappropriateness of their use so as to avoid criticisms [of] Woods; 146 This distance is in error. The relevant distance was 160 to 170 metres. -- 42 of 69 -- [2026] SASCA 50 The Court 41 3.4 There was another possible source of ignition of the fire, namely a lightning strike, which had been wrongly excluded in the initial investigation and never properly investigated at the time or since. Introductory observations 138 The appellants contend that an application of the CG&C rate demonstrates that the distance that may have been covered by the fire between its ignition and Mr England’s observations at 1.18 pm was no more than about 56 – 86 metres. They say that the more likely distance, assuming a rate of 0.5 metres per minute, was more like 28 – 43 metres. The distance actually covered, if the active heap was the ignition point, was about 160 to 170 metres. The appellants contend that the judge’s reasons failed properly to engage with what was described in their submissions as an ‘enormous’ and ‘eightfold’ disparity. 139 Before descending into the detail of the submissions, it may be noted that if the time that elapsed between ignition and 1.18 pm was at the upper end of the assumed range (86 minutes), and if the spread rate at the higher end of the range of one metre per minute is used, this in fact represents an approximate ‘twofold’ disparity. If the rate of 0.5 metres is adopted, it involves a little less than a ‘fourfold’ disparity. It is only if the shorter time period and the 0.5 metre rate are adopted that something approaching an ‘eightfold’ disparity could be suggested.147 140 Those differences aside, describing a disparity in terms of multiples may tend to obscure that what was in issue was a relatively short distance in absolute terms. A ‘fourfold’ disparity equates in practical terms to a distance between the predicted and apparently observed extremity of the fire front of a little over 100 metres. For context, the experts agreed that the spread rate of a head fire under prevailing winds of the speed experienced around the time of the fire was well over 100 metres per minute. Whilst no-one suggests that after reaching its quasi-steady spread rate the fire was travelling in a northerly direction as a heading fire, or at anything like that rate, it was common ground that the spread rate of a fire flank (whilst complex precisely to calculate) can be significant. 141 Accepting the criticisms that were made of Dr Marsden-Smedley’s 4.5 metre per minute rate, if there were short periods of time in which, in the area near the origin point, the wind direction was such that the fire’s northern front represented a flanking fire, the spread rate in a northerly direction would likely have been faster than one metre per minute. If the judge was not wrong to allow for the possibility that the fire had started more than 45 minutes before the 12.30 pm call was placed, there would have been a longer period of time over which the posited disparity of 100 metres may have been covered. 142 There were other reasons why the judge did not consider that the CG&C rate undermined the active heap theory, to which it will be necessary to return, but on 147 The ratio of 28 metres to the midpoint between 160 and 170 metres (165 metres) is less than 1: 6. It is only if it is assumed that the fire would not have moved to the north at all during a 15 minute period before the fire achieved its quasi-steady state that an eightfold disparity would be in play. -- 43 of 69 -- [2026] SASCA 50 The Court 42 appeal, the respondents made a more fundamental contention about this issue. It was that whereas the appellants’ approach assumed that the CG&C rate (or rather, its asserted application in this case) had the status of an incontrovertible fact that could not be satisfactorily reconciled with a finding that the active heap was the fire’s origin, the anterior question was whether it could reliably be applied as a tool for determining the origin of a fire, or at least the origin of this fire. 143 Another aspect to the appellants’ submissions under grounds 2 and 3 was that there was an inconsistency, not reconciled by the judge, between accepting Mr England’s evidence of localised wind conditions being erratic (so as to provide a possible explanation for the rate of spread from the active heap to the northern reach of the fire at 1.18 pm), and accepting Dr Marsden-Smedley’s analysis that the observed extent of the fire front and fire ground at 1.40 pm (a distance of about 7.5 kilometres from the hypothesised ignition point, with about 1100 hectares burned) roughly accorded with what he expected based on the prevailing wind speed (about 32.4 kilometres per hour) in a south or south easterly direction (that is a north or north-westerly wind). 144 In other words, if the fire front was behaving as though it was subject to the prevailing wind suggested by the weather station reports, it was incongruous to accept that, back near the hypothesised ignition point, the fire was spreading north other than as a backing fire. In the appellants’ submission, the fire could not be ‘schizophrenic’. 145 In the appellants’ submission, minute-to-minute variations are factored into the use of an overall prevailing wind, and could not explain markedly variant behaviour at the heel of the fire. The appellants drew attention, in particular, to a figure depicting the results of an experimental fire (the ‘Gunn Point fire’) contained in Cheney & Sullivan. The figure is reproduced below: -- 44 of 69 -- [2026] SASCA 50 The Court 43 146 Changes in wind direction are signified by the three different blue lines or arrows. The series of expanding isochrones represent the extent of the fire ground over time. The appellants’ essential submission was that whilst the wind direction changes cause the front and the flank of the fire to move, the isochrones to the rear or ‘heel’ of the fire are very tightly bunched and not materially affected by these variations. 147 It needs to be borne in mind, however, that the prevailing wind direction changes shown in this experiment are all within a band of approximately 90 degrees. Moreover, the total forward spread of the fire covered by the experiment appears to be in the order of about 500 – 600 metres. It might be thought unlikely that the wind conditions at the front of the fire, as it spread from the ignition point to a distance of around 500 – 600 metres away, would have varied markedly (or more than momentarily) from the wind conditions at the ignition point. Whether the same is true whilst the fire travelled in the order of seven kilometres, however, is a different question. 148 The possibility of different weather conditions applying at different locations in the same fire at the same time was referred to in the evidence as ‘decoupling’. Indeed, Cheney & Sullivan contains a chapter titled ‘Local Variation and Erratic Fire Behaviour’ which commences with this text: -- 45 of 69 -- [2026] SASCA 50 The Court 44 As described earlier, grassfires can appear highly erratic, with behaviour that is very variable and difficult to predict. Much of this variation can be taken into account by taking average values of fuel and wind conditions and predicting average rates of spread over periods of 15 minutes or more. However, it is important for firefighters to understand that, over much shorter periods, grassfires will respond to many factors that cannot be included in a fire spread prediction system and are, for all intents and purposes, unpredictable. Not only does a fire respond to changes in fuel and weather in different ways but these factors themselves are rarely constant; they are highly variable in time and across the landscape. Wind As we have seen, wind plays an important role in determining the behaviour of a grassfire. Rapid changes in wind direction and strength can cause immediate changes in a grassfire’s behaviour. Wind varies greatly in strength and direction, both in time and in location. Wind at one point in the landscape will be different at another time but will also be different somewhere else at the same time – some-times even quite near. This variation is a result of turbulence embedded in the wind flow that forms gusts (increases in wind speed over the average) and lulls (decreases in wind speed over the average). Meteorologists generally describe the surface wind and in statistical terms (ie. average wind speed and direction) and use values that are averaged over defined intervals, e.g. 10-20 minutes, in order to overcome the variations caused by the gusts and lulls. The turbulence generally takes the form of eddies, that result from a number of mechanisms. 149 The text goes on to explain ‘eddy’ formation and the potential impact of the fire itself on localised wind conditions. 150 In the present case there is the further complication that the approximate origin of the fire was located in a valley (albeit a relatively shallow one), whereas at least some of the distance covered by the fire traversed territory outside the valley. 151 It must be true, as the appellants submitted, that when the fire progressed in a generally southern direction from its origin, this was under a prevailing wind at ground level in the valley. It is also likely that by the time the fire front left the valley and proceeded in a more easterly direction, this was after Mr England’s initial observations. The real question, however, is whether, before, or perhaps in the period shortly after, the fire commenced to move in a generally southerly direction from the origin point (or both), the wind at the ground level around the origin point may have caused the fire to spread in a northerly direction other than as a pure backing fire. 152 Put another way, and bearing in mind Mr England’s observations of the erratic wind conditions, was it plausible that the fire front was generally subject to a generally northerly prevailing wind, but that for periods before this or after the fire front was some distance away, the prevailing wind back at the origin point was gusting in different directions, so as to allow for the fire to spread to the north as a kind of hybrid between a flanking fire and a backing fire? -- 46 of 69 -- [2026] SASCA 50 The Court 45 153 In that context, Mr England’s observations, the acceptance of which by the judge the appellants explicitly did not challenge, assume significance. He described the fire as having swung around in a westerly direction when he first saw it, albeit under a relatively light wind that he estimated at 3 – 5 knots, and maybe 6 knots later. The appellants submitted that those kinds of wind speeds are at the lower baseline of winds that can be meaningfully modelled, and that at lesser wind speeds a fire spreads very slowly in any event. In the appellants’ submission, winds of that speed would not produce a significant flank fire spread rate. 154 Another important aspect of the appellants’ challenge to the judge’s reasoning under appeal grounds 2 and 3 relates to fire pattern indicators. The judge found Mr Woods’ analysis of fire pattern indicators (summarised in the diagram below)148 supported the active heap theory. 148 Reasons [1037]. -- 47 of 69 -- [2026] SASCA 50 The Court 46 155 In essence, the appellants contend that Mr Woods’ analysis, which involved reliance on micro fire pattern indicators observed some time after the fire, was unreliable. They pointed to the tension between his analysis (which argued in favour of the fire heading north from the active heap as a backing fire) and the judge’s treatment of the CG&C rate. The CG&C rate 156 In their 1998 paper titled ‘Prediction of Fire Spread in Grasslands’, Cheney, Gould and Catchpole described a model to predict fire spread in grasslands from -- 48 of 69 -- [2026] SASCA 50 The Court 47 windspeed at 10 m, dead fuel moisture, and degree of grass curing in three defined pasture types. The authors stated: The aim of this work was to develop a model for fire spread in continuous grasslands and compare the predictions of this model with the spread rates of a number of wildfires. The model was based on data from experimental fires but required a number of logical assumptions to predict fire spread beyond the range of the data. 157 As the paper goes on to explain, the authors built upon field data from the ‘Annabarroo experiments’ and other qualitative and quantitative field observations. The authors discussed how, subject to the size of fire, fuel load and conditions, linear or power relationships might exist between rate of spread and wind speed. For the purposes of the model, it was necessary to define a rate of spread when the wind speed was zero. In that context, the authors observed: Completely calm conditions rarely exist in the field. When they do, fires are restrained to some degree by the convection behind the leading edge and are circular backing-fires. Therefore, we have adopted the backing rate of spread for grass fires (see Figure 2) as the rate of spread at zero wind speed (R0). In this data set wind speed at 2 m (U2) was not a significant variable although the range of U2 was from 0.5 to 4 ms-1. 158 Figure 2 is the scattergram that was set out earlier in these reasons and is reproduced again for ease of reference. 159 The authors went on to observe: In the field, the transition from a backing fire to a heading fire is unstable and complicated by thermal activity. When wind speeds at 10 m (U10) are less than 5 km h-1 they are aptly described as light and variable and fires progress erratically both in speed and direction as -- 49 of 69 -- [2026] SASCA 50 The Court 48 they respond to gusts and lulls caused by thermal activity. Under these conditions we could not obtain reliable correlations between head fire spread and wind speed at 2 m anemometers located near the fire. Often the wind affecting the head fires was quite different from that recorded at the anemometer depending on passage of gusts and lulls over the site. At times a head fire was observed to progress quite rapidly under the influence of a local puff while the wind recorded at the anemometer site was calm or even in the opposite direction to the direction of fire spread. During our experiments we obtained consistent head fire spread only when U10 > 5 km h-1. 160 As will be apparent from these extracts, the purpose of the paper is to predict the forward spread of fire. The data relating to backing fire spread rate was used to assist in formulating a forward spread rate at low wind speeds. The authors expressed reservations about the utility of the model at these low speeds. Indeed, they expressed reservations about the predictive accuracy of the model(s) generally. The concluding paragraphs of the paper stated: The distribution of the data on the scattergram (Figure 4) illustrate the practical reality of predicting fire spread: there is likely to be a high degree of variability associated with measures of both fuel and weather variables. … … Fire suppression agencies may prefer to plain their logistics on the basis that actual fire spreads are likely to exceed predictions on only 15% of occasions. 161 The paper did not purport to be a basis for a retrospective analysis of the origin of a fire even for forward spread, although, of course, to the extent that it is reliable in a predictive sense, it must have some utility in a retrospective context. What is clear, however, is that the paper was not purporting to provide a definitive basis for predicting or retrospectively analysing the backing fire spread rate. The authors had reference to data relevant to backing fire spread rates (Figure 2), but it was not a focus of the model. Expert evidence relating to the CG&C rate and backing fire spread rates 162 The CG&C rate and the publication from which it is derived was the subject of consideration by the experts. Without being comprehensive, it is useful to collect some of the discussion of it. 163 Mr Fenwick raised the broad question of the fire spread north of the active heap in his first report. He said that in the absence of suppression, fire spread to the north-east and south-west would be expected to take 30 minutes to travel about 60 metres, with a range of 45-90 metres, but that fire spread rate prediction was not an exact science, and that calculations were intended to assist fire control practitioners fighting an advancing fire and that predictions depending on great precision from the applied model were likely to be unreliable.149 He thought that a fire spreading from the active heap to the upwind most end of the burnt area would travel at a backing rate of spread of 0.7 metres per minute.150 149 Reasons [303]. 150 Reasons [314]. -- 50 of 69 -- [2026] SASCA 50 The Court 49 164 Mr Cousins in his first report referred to the CG&C rate and, applying that rate, concluded that the distance from the active heap to the rear perimeter of the fire was too great for a backing fire to have moved that far at the time Mr England took his photographs.151 In his amended supplementary report he adopted a backing fire spread rate of 0.55 metres per minute and opined that even using a rate of 0.83 metres per minute and a maximum detection delay of 45 minutes, the fire would not have reached the location photographed by Mr England by 1.18 pm.152 165 In Dr Marsden-Smedley’s reply report, he referred to the CG&C rate and said that the prediction equation adopted in that paper (Rate = 0.054 – [0.032 x dead fuel moisture percentage]) had a poor fit to the data used to generate it. Because of the degree of scatter in the data, he considered the equation in the paper should be used with caution as its predictions were associated with a high degree of uncertainty.153 He adopted a somewhat rudimentary average of a backing rate of 0.5 metres per minute and a flank spread rate of 8.5 metres per minute, namely 4.5 metres per minute, before observing that this was probably excessive for various reasons.154 166 In Mr Woods’ supplementary report, he said since 1998 when the Cheney, Gould & Catchpole paper had been published, understanding of wildfire behaviour had evolved significantly, particularly after the Canberra and Victorian bushfires. He said fire behaviour calculations had been updated following research that found spread data inaccurately reflected the actual speed of fire travel.155 He said there was very limited global research determining an accurate rate of spread of a backing fire in grassland fuel. In his experience, the use of estimates to assess the timing of rate of spread of backing fire in grassland fuels was to be treated with great caution. Back fire spread rates were challenging to accurately predict, particularly in early stages of fire development.156 167 In the course of his evidence at trial, Dr Marsden-Smedley said, inter alia:157 • whilst a backing rate of spread is largely unaffected by increasing wind speed, it is affected by fuel moisture and wind direction variation; • in the early stages of a fire when the flank and backing fire are close to each other, they interact and influence each other, making his averaging approach appropriate; 151 Reasons [328]. 152 Reasons [339]. 153 Reasons [346]. 154 Reasons [342], [347]. 155 Reasons [357]. 156 Reasons [358]. 157 Reasons [608], [618], [619], [625], [634]. -- 51 of 69 -- [2026] SASCA 50 The Court 50 • in any event, he considered the spread rate to be unlikely to have been as low as one metre per minute; • he was not aware of any specific literature advocating the use of backing fire spread rate modelling to identify the ignition location of a bushfire; and • the prediction of back fire spread rate modelling was more uncertain than head fire modelling and more effort had gone into the latter. 168 In the course of Mr Woods’ oral evidence at trial, he said, inter alia:158 • in internationally adopted wildfire investigation methodology, backing fire rates of spread derived from estimates based on software or algorithms for fire prediction modelling were not used because they were considered unreliable; • constant wind speed or wind direction does not occur and consequently variations result in the fire moving in different directions at different times with differing rates of spread, and the equations in the Cheney, Gould & Catchpole paper were not considered reliable for estimating the general origin area, the specific origin area or the ignition area; • a decrease or change in the prevailing wind direction could make a backing fire more of a lateral fire in certain circumstances with a higher spread rate, and there were transition zones within the backing fire area where the fire moved from a back to a somewhat lateral fire and back again; and • it was not a nonsense to say that the fire could have covered the distance of around 160 metres in 86 minutes. 169 In the course of Mr Cousins’ oral evidence, he:159 • acknowledged he had not used backing fire spread rate analysis or modelling from the Cheney, Gould & Catchpole paper equations for the purposes of identifying a point of origin in a bushfire or a grassfire; and • agreed that the Cheney, Gould & Catchpole equations and the scattergram were based only on fuel moisture content and did not take into account the effects of wind direction changes during the early stages of a fire. 170 Mr Fenwick, in his oral evidence, adhered to the view that backing fire spread calculations were appropriate to be used in the present context. When in his report he had talked about the unreliability of models, he was referring to forward spread rates. His view that backing fire rates of spread were reliable was based on the 158 Reasons [646], [647], [687]. 159 Reasons [704], [706]. -- 52 of 69 -- [2026] SASCA 50 The Court 51 Cheney, Gould & Catchpole paper and on Cheney & Sullivan.160 He agreed, however, that the CG&C rate analysis did not accommodate changes in wind direction.161 The judge’s treatment of the CG&C rate 171 It will be evident from the following brief summary of the evidence and arguments, drawn as it is from the Reasons, that the judge engaged extensively with the evidence given, and arguments made, about the CG&C rate. 172 In Part I of the Reasons, the judge addressed in some detail the appellants’ criticisms of the averaged rate adopted by Dr Marsden-Smedley. As the judge’s analysis shows, whilst there may have been difficulties with averaging in this context, a similar criticism was also potentially available of the analysis of some of the other experts. At all events, the judge determined not to rely on Dr Marsden- Smedley’s figure of 4.5 metres per minute in reaching her ultimate conclusions.162 173 The judge then considered the appellants’ more general contentions to the effect that an application of the CG&C rate seriously undermined the active heap theory. Whilst lengthy, it is important to reproduce the reasons on this topic in order to demonstrate the various layers to the reasoning.163 Backing fire spread rate prediction [998] On the applicants’ case, use of backing fire spread rate modelling to disprove a source of origin is unsupported by any authoritative work. The respondents’ position, however, is that no authoritative work calls into question the backing fire spread rate equation in the Cheney, Gould and Catchpole paper. [999] The Cheney, Gould and Catchpole paper created equations for predicting forward spread rates based in part on laboratory studies. The backing fire spread rate was identified for the purpose of incorporation into equations for predicting the rate of spread of an advancing fire under different wind speeds. This was necessary because it was recognised that there will be some fire spread even in the absence of any wind. The definition of the rate of spread at zero wind for the purpose of forward spread predictions was derived from extrapolation of data for backing fires shown in figure 2 in the Cheney, Gould and Catchpole paper. The maximum rate of spread for backing fire depicted in figure 2 was about .54 metres per minute. The scattergram in figure 2 was derived by the authors from a number of research fires conducted before 1998, the details of which are not identified in the paper. The graph from which the backing fire spread rate was devised plotted spread rate and fuel moisture content. There was no substantive discussion in the paper about the potential impact of different variables, such as erratic wind speed and direction, on the accuracy or reliability of the backing fire spread rate equation derived from the research fires. This is explicable on the basis that the Cheney, Gould and Catchpole paper does not purport to be a study specifically addressing the prediction of backing fire spread rates nor as addressing the use of backing fire spread rates for identifying the location of fire ignition. 160 Reasons [725]. 161 Reasons [729]. 162 Reasons [996]. 163 Reasons [998]-[1025] (footnotes omitted). References to the ‘applicants’ are to the respondents on appeal, and references to the ‘respondents’ are to the appellants. -- 53 of 69 -- [2026] SASCA 50 The Court 52 [1000] The Cheney, Gould and Catchpole paper recognised that laboratory data and field data cannot be directly compared. The authors observed that rate of spread at zero wind speed may not be an appropriate scaling variable for developing a relationship between wind speed and rate of spread because there may be a change in the mechanism of fire spread above a critical wind speed such as changes from a backing to an advancing fire. The authors stated: The concept of a critical wind speed is more important for our understanding of fire spread mechanism than it is for practical prediction of rate of spread in continuous grasses. Equations for wind speeds of less than 5 km per hour are “purely conceptual models” to predict rate of spread at low wind speeds as fires burning under light wind conditions often spread erratically as they respond to gusts and lulls caused by localised thermal activity. The models could under predict the rate of spread in a short localised gut and over predict during lulls. [1001] The respondents’ experts sought to limit such qualifications to the prediction of forward rates of spread only, while maintaining reliance on the backing fire spread rates in the Cheney, Gould and Catchpole paper. However, the purpose of that paper was to create a model to predict forward rates of spread based on laboratory experimental data, factoring in wildfire data (where available and relevant), and recognising inherent limitations in the models including the potential for variability in the field. While the observations above were made in the context of forward spread rate calculations, nevertheless they expressly acknowledged the limitations of the study and the potential for variability in real world conditions. I understand the different characteristics of backing fire, flank fire and advancing fire movement resulting from the thermodynamic mechanisms underpinning them. However, it is still unclear why the expressed limitations and qualifications would solely be limited to forward spread rate and to exclude any possibility that variability in real world conditions may potentially impact upon the backing fire spread rate equation derived from the data underpinning the paper. [1002] Mr Woods was unshakeable in his opinion that the backing fire rate of spread could not be used to identify the general or specific location of ignition or to estimate how far a fire moved as a backing fire because of the unreliability of rate of spread predictions. Dr Marsden-Smedley also rejected the proposition. Mr Woods referred to the fact that there is limited global research determining an accurate backing fire rate of spread and the use of estimates to assess the timing of the rate of backing fire spread is treated with great caution. This is particularly so in the early stages of fire development and backing fire spread rates are considered challenging to accurately predict. Research has focussed on tools to predict advancing fire spread rates from the perspective of safety of firefighters and at risk communities. [1003] In his first report, Mr Fenwick stated that fire spread date prediction is not an exact science and predictions depending on great precision from the applied model are likely to be unreliable. Mr Cousins accepted that bushfire modelling is not reliable for identifying the area of origin of a fire. Mr Cousins did not think there was an authoritative text recommending backing fire spread rate analysis as a reliable tool for proving or disproving a source of ignition. [1004] I accept I was not referred to any publication expressly challenging the backing fire spread rate equation in the Cheney, Gould and Catchpole paper. However, I was also not taken to any publication endorsing its use apart from the specific section in the Cheney and Sullivan text which recognises its potential utility in the limited circumstances described in the text and which I address below. -- 54 of 69 -- [2026] SASCA 50 The Court 53 [1005] The equations in the Cheney, Gould and Catchpole paper were based on winds measured at 10 metres, that is, 10 metres above ground level as usually measured by automatic weather stations. In this case, the localised wind conditions were those felt close to the ground by Mr England. Mr Fenwick’s evidence was that wind changes such as those described by Mr England will not impact the backing fire rate of spread. However, I was not taken to any research papers or other material to suggest that erratic changes are factored into the Cheney, Gould and Catchpole equation for backing fire spread rates. The Cheney, Gould and Catchpole paper does not suggest that the backing fire spread rate equation was intended to take into account all potential impacts of localised topography, fuel and erratic winds so as to be capable of consistently and accurately predicting backing fire spread rate in differing, variable conditions. [1006] Mr Fenwick relied on an extract of the Cheney and Sullivan text, which refers to the use of backing fire spread in particular circumstances, as an authoritative text recommending the use of backing fire spread rates to identify the source of origin of a fire. The Cheney and Sullivan text states that if there is clear information about the time of the probable ignition and the time the very back of the fire is suppressed is known, the fact that fires back directly into the wind spread at a constant rate can be used to locate the “probable region of ignition”. The extract relied upon by Mr Fenwick cross referred to particular diagram in that text which plots the relationship between backing rate of spread and dead fuel moisture content for fires in continuous fuels of sorghum and other tropical grasses in experimental wind speeds of between three to 20 kilometres per hour. There is no further information within the text about the experimental data or testing which formed the basis for the diagram, however, that diagram bears similarity to the scattergram in the Cheney, Gould and Catchpole paper. [1007] Mr England gave evidence about some of the fire suppression activities which would have impacted the shape of the heel of the fire. However, there was insufficient evidence about all of the precise fire suppression activities taken at the heel of the fire to enable me to determine specifically how those activities impacted the heel of the fire, including its shape. [1008] Mr England’s evidence supports the conclusion that when he arrived, and for some time after, the back of the fire was not backing constantly into the wind. Mr England gave evidence that the wind changed frequently, shifting from north-east across to west-north- west and he saw the fire move between backing, flanking and heading about 10 times while he was in the gully with the fire running in fingers up the hill. [1009] In the extract to which Mr Fenwick pointed in the Cheney and Sullivan text the authors observed that lateral spread abreast or ahead of the ignition point is difficult to determine because of alternating backing and heading behaviour of lateral fires. Mr England’s evidence indicates the fire was changing characteristics constantly. It is thus not possible to draw firm conclusions by using the Cheney, Gould and Catchpole backing fire spread rate or the matters referred to in the Cheney and Sullivan text to locate the region of ignition because of the lack of clear information about relevant factors including the precise fire perimeter at the time Mr England arrived and the time at which different parts of the very back of the fire were suppressed. Mr Fenwick accepted in cross-examination that the various requirements for the use of backing fire spread rates to locate the origin of the fire, as referred to in the Cheney and Sullivan text, were not satisfied in this case. [1010] For the reasons above, I am persuaded that the circumstances in which the Cheney and Sullivan text indicates backing fire spread rates may be used to determine the region of ignition are not apt for use in the context of this fire and, further, are not suitably applied to disprove the active heap as the source of the fire. -- 55 of 69 -- [2026] SASCA 50 The Court 54 [1011] Mr Fenwick maintained firmly that both wind speed and wind direction would not impact on the backing fire spread rate, based on the backing fire vector being about 45 degrees either side of the back of the fire and the Gunn Point fire diagrams depicted in the Cheney and Sullivan text. The Gunn Point fire diagrams in the text depict isochrones at the back of the fire which, with wind direction changes, generally remain more closer spaced than isochrones representing the advancing fire. Not all of the diagrams to which Mr Fenwick referred were put to Dr Marsden-Smedley and Mr Woods, however, Dr Marsden-Smedley was cross-examined on figure 6.7. Figure 6.7 is contained in the chapter of the text addressing local variation and erratic fire behaviour in a section addressing wind. The authors state that fires respond to lulls and gusts in the wind, grassfires can respond almost immediately to changes in wind speed and direction and there can be differences between the impacts of up draughts and down draughts of wind. Figure 6.7 is described as illustrating the effect of successive down draughts which cause a fire to change direction frequently during its initial spread, also causing it to build up rapidly. The figure suggests the back of the experimental fire depicted in the diagram was less responsive to wind changes than the front of the fire. Dr Marsden-Smedley accepted wind speed had a minor effect but did not accept the diagrams depicting the particular experimental fires necessarily meant that the fire at Lucindale followed the same pattern or that the diagrams precluded a conclusion that the fire spread from the active heap to the location depicted in the photographs in the erratic weather of the nature described by Mr England. Mr Woods considered wind speed did not strongly affect backing fire spread rate but variability in wind direction could result in a backing fire causing a higher intensity more lateral fire which would account for the distance travelled. [1012] As set out above, Dr Marsden-Smedley and Mr Wood maintained that backing fires are sensitive to changes in wind direction which can result in changes from backing fire to flank fire. In addition to the matters which I have addressed above, the Cheney and Sullivan text recognises that rapid shifts in localised wind, including lulls and gusts of wind, will impact on fire spread; wind near the ground is variable and direction of fire travel will be erratic with erratic wind direction and rapid changes in wind direction and strength can cause immediate changes in fire behaviour. As the wind lulls and gusts in different directions at a local level, the fire will spread erratically and as the wind erratically pushes flames in different directions fire spread will not be constant in rate or direction. Mr England’s evidence indicates this was the case in the gully when he arrived and for about three hours after. [1013] Neither the Cheney, Gould and Catchpole paper nor the Sullivan and Cheney text expressly exclude the possibility that backing fire spread rates may exceed the rates of spread contained in the data from which the backing fire spread rate equation was derived in the Cheney, Gould and Catchpole paper or the data on which the diagram in the Cheney and Sullivan text was based. Dr Marsden-Smedley and Mr Woods both gave evidence that they had observed backing fire spread rates which exceeded those in the Cheney, Gould and Catchpole paper. Mr Woods referred to his experience with backing fires in grassland fuels and said he had seen an increase in the rate of spread of backing fire with a reduction in wind speed. Dr Marsden-Smedley was criticised on the basis that his PhD work was in different vegetation and the number of fires referenced in his PhD were not a statistically valid basis for drawing conclusions. Dr Marsden-Smedley nevertheless firmly maintained his position and provided an explanation for the data set out in the appendix to his PhD thesis. Mr Woods’ maintained he had experienced greater backing fire spread rates under a reduction in wind speed. I accept the evidence of Dr Marsden-Smedley and Mr Woods that in their experience, it is possible for backing fire spread rates to exceed those set out in the Cheney, Gould and Catchpole paper in some circumstances in the field. I take into account the experience and expertise of Dr Marsden-Smedley and Mr Woods in accepting their evidence on this issue. -- 56 of 69 -- [2026] SASCA 50 The Court 55 [1014] The NWCG Guide and the NFPA Guide recommend fire progression indicators as the mechanism for locating the general area of origin, the specific area of origin and then the search for the ignition point and cause. The CFS investigators and Mr Woods adopted that process. Those Guides do not expressly endorse using backing fire spread to identify the origin of a fire. Despite Mr Cousins and Mr Fenwick’s criticisms of the NWCG and NFPA Guides, both parties relied on them as authoritative texts. I accept that fire pattern indicators are the recognised authoritative method currently employed by wildfire investigators to identify the location of the ignition of the fire. I address the Parker and Babrauskas paper separately below. [1015] There were a number of variables which had the potential to impact on the extent to which the fire had spread to the point observed at 1:18 pm. [1016] First, the precise time of ignition is unknown. Of course, the precise start time is important in assessing the contention the fire could not have moved as far as it did in the relevant time. While counsel for the respondents submitted there was no great uncertainty in the ignition time of the fire, there was in fact no direct evidence about when the fire started. The start time was postulated entirely based on assumptions made by Mr Cousins with which Dr Marsden-Smedley agreed concerning the likely lag time between the fire start and when smoke was first seen. The reliability of the detection delay assumption was important in the respondents’ position that the fire could not have commenced at the active heap and moved as a backing fire to the point at which Mr England photographed it at 1:18 pm. Mr Cousins assumed a detection delay of between 15 and 45 minutes. Mr Cousins’ basis for that assumption was that for experienced fire tower operators, the median detection delay was 15 minutes with 75 per cent of all fires reported in 44 minutes. This was based on a 2012 paper included in Mr Cousins’ bibliography entitled “Field evaluation of two image-based wildland fire detention systems” by Matthews et al. Dr Marsden-Smedley agreed with the assumed time lag without expressing the basis upon which he did so. The paper on which Mr Cousins relied was not tendered. This fire was not detected by experienced fire tower operators. There was no evidence about the reliability of applying rates of detection by experienced fire tower operators to contexts such as applied in this case. The foundation upon which I could assess the reliability of the assumption was thus lacking. Mr Woods said that with the degree of curing of material, the initial fire would have burnt with little smoke and it is possible the fire ignited earlier. Mr Fenwick suggested the start time of the fire was pure speculation. I cannot make specific findings about when the fire started. It is conceivable that the fire started earlier than assumed. [1017] Second, the precise fire boundary at 1:18 pm is not entirely clear. In the photograph taken by Mr England, which has the geolocated tree at the left hand side, it is difficult to see exactly where the fire had spread. The precise boundary is unclear. The precise boundary of the heel of the fire at that time remains unclear. [1018] Third, there is no evidence about the state of the fire or fire ground prior to Mr England arriving on the scene. [1019] Fourth, Mr England gave evidence that the local weather conditions when he arrived on the scene were constantly changing. Mr England’s evidence about the weather constituted evidence of fact of matters observed by him. Mr England was clear and confident in expressing his recollections and his extensive experience as a CFS volunteer justifies the conclusion that Mr England’s observations about matters such as the wind direction and speed and the fire movement were reliable. [1020] Mr England’s evidence justifies the findings that when he arrived, the wind was at low speed in the valley, flicking around in different directions, with consequential impact -- 57 of 69 -- [2026] SASCA 50 The Court 56 on the fire movement. Flame angles and smoke direction in the photographs, as explained by Mr England and the experts, show transitions of movement as described by Mr England. The precise impact such localised conditions had on the movement of the fire, including direction and rate of fire spread, cannot be scientifically modelled. Mr Woods said he found transition zones even in the backing fire vector consistent with Mr England’s evidence. It is not possible for me to make specific findings about how and where the fire moved across a period of time from ignition to the point when Mr England arrived. However, I accept Mr England’s evidence about the local fire ground weather, that the zephyring of the wind locally on the fire ground impacted on the movement and direction of travel of the fire and that the local wind in the gully was different from the wind over the hill. Mr England’s evidence supports findings that there were rapid changes in local wind direction resulting in rapid shifts of fire movement, the fire behaviour on the ground was erratic and localised wind shifts impacted the spread of the fire and I find accordingly. It also supports the conclusion that a simple application of the Cheney, Gould and Catchpole backing fire spread rate equation may fail to adequately factor in consequences arising from the localised weather on the fire behaviour as occurred in this case. Mr Fenwick’s weighted average tends to support that conclusion. [1021] The respondents contended that Mr England’s evidence that the wind did not settle down for three hours is wrong because, if so, the fire would not have advanced as it did. Mr England’s evidence was limited to the wind conditions he experienced and observed in the localised area and he contrasted the wind in the gully with the wind over the hill. His evidence was not evidence about the prevailing wind over the hill. This criticism does not cause me to doubt the reliability of Mr England’s evidence about the wind or the impact it had on the localised fire behaviour. In reaching my conclusions I have not overlooked the fact the BOM automated weather stations demonstrated prevailing winds across the area as set out above. Nevertheless, I accept the evidence of Mr England about the localised wind conditions and the opinions of Mr Woods and Dr Marsden-Smedley based on that evidence. [1022] Fifth, the experts agreed that the local topography, that is, the valley in which the fire was located, would have some effect. I accept and find that the inclined sides of the valley would have impacted the rate of spread of fire moving up the slope. Again, I cannot make precise findings about how the fire moved from the point of origin over time. [1023] In reaching my conclusions I have not overlooked Mr Fenwick’s evidence about the strange shape of the heel of the fire. It does not cause me to doubt my conclusions in light of the variables impacting the fire heel including Mr England’s evidence about the frequent wind change, topography and suppression activities around the heel of the fire. I also have not overlooked the fact the BOM automated weather stations, some 50 kilometre distance from the valley and measuring wind at 30 minute intervals, demonstrated prevailing winds across the area as set out above. Nevertheless, I accept the evidence of Mr England about the localised wind conditions including that the wind in the gully was different from that over the hill and the opinions of Mr Woods and Dr Marsden-Smedley on this topic. [1024] It follows that I cannot make specific findings about the fire ignition location, fire behaviour and movement over time from the start of the fire by simple application of the Cheney, Gould and Catchpole backing fire spread rate equation or other models, including Dr Marsden-Smedley’s average spread rate figure. Reaching firm conclusions by simple application of models is precluded by the extent of variability of local conditions and the impact of those conditions on the fire behaviour and movement. [1025] Given all these matters, in my view, the most reliable evidence upon which to form conclusions concerning the potential cause of the fire is the evidence of the localised weather conditions and the evidence found on the ground after the fire in the form of fire -- 58 of 69 -- [2026] SASCA 50 The Court 57 pattern indicators demonstrating advancing, backing and lateral vectors. I address that evidence in detail below. Evaluation of criticisms of judge’s approach 174 The appellants’ written submissions contended that whilst the judge had acknowledged the difficulties with Dr Marsden-Smedley’s 4.5 metre spread rate and did not rely upon it, ‘[w]hat the Judge did not do … was address the enormous disparity between the distance which the backing fire would have covered on the application of the Cheney Gould and Catchpole spread rate equations’ and the distance suggested by Mr England’s observations.164 175 Putting the adjective ‘enormous’ to one side,165 in our view, the disparity contended for was addressed by the judge. However, the judge was not required to explain away the asserted disparity. The question was whether, having regard to the evidence about the CG&C rate and its applicability, and the confidence that the Court could have about the required integers for its application, the observed extent of the rear of the fire at 1.18 pm contraindicated the active heap being the origin. 176 As appears from the passages that have been extracted, the experts expressed differing opinions on the real-world applicability of the CG&C rate and its robustness as a forensic tool. Dr Marsden-Smedley and Mr Woods both gave evidence of having witnessed real-world backing fires at rates exceeding that provided for by the theoretical equation, and the judge accepted that evidence. The appellants may have characterised that (and Mr England’s evidence) as ‘anecdotal’, but it was evidence they gave that was accepted. Mr England was a very experienced firefighter and was found to have ‘excellent recall’.166 The judge accepted there was no consensus that the CG&C rate was reliably applicable in a forensic context. There was no evidence of the rate having been used in other bushfire investigations to determine origin. 177 There was a rational reason to accept, generally, that Cheney & Sullivan and Cheney, Gould & Catchpole were highly regarded publications, whilst remaining cautious about the applicability of the CG&C rate as a forensic tool. The publications were mainly focused on fire prediction with an emphasis on the forward spread of a fire. Cheney, Gould & Catchpole’s paper does not contain much if any explanation for the data reflecting spread rates in zero wind in the scattergram comprising Figure 2, from which the CG&C rate was derived, but 164 Appellants’ written submissions [49]. 165 In the Appellants’ written submissions (at [49]) and in the Appellants’ reply submissions (at [10]), the disparity was identified as being between 56 – 86 metres (if the CG&C rate was applied) and 250-385 metres (based on Mr England’s observations). This reflected the error in the grounds of appeal. In fact, if a comparison were to be engaged in, it was between the distance suggested by the CG&C rate and the distance of 160-170 metres. As noted earlier in these reasons, the references in the appellants’ submissions to an ‘eightfold disparity’ appear to reflect this error. 166 Reasons [768]-[769]. -- 59 of 69 -- [2026] SASCA 50 The Court 58 there was expert evidence accepting the proposition that it did not accommodate local wind variations. 178 One of the appellants’ contentions on appeal was that ‘given that the equations were the product of extensive field experimentation, it must be assumed that local wind variability was encountered and accounted for to some degree in their formulation. To dismiss them on the basis of such variability is to overlook the very conditions under which they were developed’.167 They further contended that it would be ‘naïve to think that [localised wind changes were] not factored into consideration by Cheney and Sullivan and Cheney, Gould and Catchpole’.168 179 The respondents submitted that this contention was not made at trial. That may not be entirely fair. It is clear from the Reasons that the judge treated the appellants’ reliance upon what might be inferred from the figure depicting the Gunn Point fire as a contention that backing rates were not materially affected by local wind variations. However, as the judge explained, whilst that figure suggests that the back of the experimental fire was less responsive to wind changes than the front of the fire, Dr Marsden-Smedley’s evidence was to the effect that this did not justify a conclusion that the fire at Lucindale followed the same pattern or that the diagram precluded a conclusion that the fire spread from the active heap to the location depicted in the photographs in the erratic weather described by Mr England. As earlier observed, the Gunn Point fire diagram depicts changes within a band of the direction of wind but it is not obvious that it can be extrapolated to more complex variations including not only more radical temporary changes in direction but lulls and gusts. 180 Another feature of the facts which counts against the proposition that the applicability of the CG&C rate was akin to an incontrovertible fact (or gave rise to a relevantly compelling contrary inference) is the topography of the fire ground in question. There was expert evidence accepting that this could affect spread rates. 181 Furthermore, a critical integer in any application of the CG&C rate as a forensic tool in this case was the period of time between ignition and 1.18 pm. As the judge observed, the only basis for the assumed period of delay that was used in the posited application of the CG&C rate was a 2012 paper (not separately tendered) that apparently suggested that for experienced fire tower operators, the median detection delay was 15 minutes with 75 per cent of all fires reported in 44 minutes. This fire was not detected by a fire tower operator. On appeal the appellants submitted that on days of high fire index, there is nothing that farmers do more of than look out the window with their eyes, and stick their nose out the door to see if they can smell smoke. Despite the intuitive appeal of that submission, at most it is supportive of the idea that a long delay in detection was unlikely. Against this, the judge had evidence from Mr Woods that with the degree of curing of material, the initial fire would have burnt with little smoke. The judge 167 Appellants’ written submissions [56]. 168 Appellants’ written submissions [60]. -- 60 of 69 -- [2026] SASCA 50 The Court 59 was not bound to accept the assumed delay detection period adopted by both Mr Cousins and Dr Marsden-Smedley. 182 Standing back, and remembering that the total distance the fire was required to have travelled from the active heap to the approximate location observed by Mr England was in the order of 160 metres, it has not been shown that the CG&C rate (or its implication) had the status of an incontrovertible fact or a compelling inference that was inconsistent with the active heap theory. Fire pattern indicators 183 The appellants contended that if, as Mr Woods asserted, fire pattern indicators reliably identified fire behaviour, the backing fire indicators identified north of the active heap established that it was a backing fire, with the result that the CG&C rate could not be side-stepped on the basis that the fire had a hybrid nature. 184 The appellants contend that there were good reasons to doubt the reliability of the fire pattern indicators relied upon by Mr Woods. These included micro indicators (such as ‘protection’). Dr Marsden-Smedley, who had attended the fire ground on the same day as Mr Woods, had only considered it appropriate to look at macro fire pattern indicators such as charcoal scars on tree trunks. He had agreed that micro fire patterns were much less reliable over time, and that was why he did not have regard to them. 185 In this context, the appellants were critical of the judge’s treatment of Dr Marsden-Smedley’s reluctance to criticise Mr Woods (characterised by them as prevarication on this issue). The judge’s treatment of this issue appears in the course of her assessment of Dr Marsden-Smedley in Part H of the Reasons. She said that:169 Dr Marsden-Smedley was willing to make some concessions. At times, his answers went beyond the questions asked and reinforced his perspective in a manner which had a tendency to appear as a lack of willingness to contemplate any alternative position. When asked questions about what a competent fire investigator would do when looking for fire progression indicators a year after a fire, he appeared to be trying to avoid directly answering in a way which may impact adversely on Mr Woods who had used fire progression indicators which Dr Marsden-Smedley said he would not have used. Dr Marsden-Smedley continued to reiterate what he would do, thus avoiding directly criticising Mr Woods’ approach. That said, his approach may not have been consistent only with a lack of independence, as asserted by the respondents, and may have been influenced by professional courtesy, particularly in circumstances where Mr Woods was at all times present during trial. 186 That approach is likely to have been informed by the judge’s opportunity to see and hear the witnesses over an extended period. However, even if there were a basis for this Court to consider that Dr Marsden-Smedley ought, in discharge of his duty to the court, to have called a spade a spade and simply disagreed with 169 Reasons [790]. -- 61 of 69 -- [2026] SASCA 50 The Court 60 Mr Woods, the question is where that criticism of the judge’s approach goes in the scheme of the case. More particularly, the question is whether it would found a conclusion that the judge was wrong in her ultimate analysis of the issue of causation. 187 So far as concerns the question of fire pattern indicators, it is appropriate to summarise the judge’s reasoning on this topic in Part I. The judge’s conclusions about fire pattern indicators 188 First, the judge rejected a criticism of Mr Woods based on his failure to draw attention to a particular paper expressing concerns about the reliability of fire pattern indicators (the Parker and Babrauskas paper). The judge accepted the explanation he gave including that it had not been peer reviewed. She placed reliance on the fact that the NWCG and NFPA Guides supported the use of fire indicators.170 189 The judge considered that fire pattern indicators were a reliable basis for use by experienced fire investigators in identifying fire vectors provided they were considered in the overall context of the fire behaviour.171 190 Dr Marsden-Smedley had used macro fire pattern indicators and concluded that charcoal scars on tree trunks supported the fire spread direction being towards the south-east from the location of the active heap and the cool heap. The judge accepted this assessment was undertaken consistently with the NWCG Guide.172 191 Turning to Mr Woods’ evidence, the judge said that the details shown in photographs and explained by him during cross-examination assisted her to understand his opinions and to accept them.173 She related the photographs to the diagram prepared by Mr Woods, set out earlier at paragraph [32] of these reasons. The judge addressed a number of them in some detail in the Reasons, explaining the basis upon which they were said by Mr Woods to support the active heap as the origin, and pointing out where the particular indicators were found on large objects or objects that appeared to be partially embedded in the ground. 192 In Mr Woods’ second supplementary report, he included a plot of all images he captured in his scene examination overlayed on a Google Earth image. Advancing fire pattern indicators were shown in red, flanking indicators in yellow and backing indicators in blue.174 170 Reasons [1026]. 171 Reasons [1028]. 172 Reasons [1032]-[1034]. 173 Reasons [1035]. 174 Reasons [1043]-[1044]. -- 62 of 69 -- [2026] SASCA 50 The Court 61 193 The judge addressed a criticism of Mr Woods’ report that he had not included inconsistent fire indications, accepting that there would always be some contrary indicators, that Mr Woods had taken them into account, and that the NWCG Guide confirmed it was not necessary for investigators to document fire pattern indicators considered inconsistent with the overall fire direction.175 The judge also addressed the evidence of Mr Woods about CFS photographs taken shortly after the fire.176 The judge accepted Mr Woods’ interpretations of the fire pattern indicators given his experience, expertise and the clear and cogent evidence he gave.177 She rejected Mr Cousins’ criticism that Mr Woods’ analysis was ‘too good to be true’, and the criticism that he had failed to search for indicators indicative of an ignition source by way of lightning in the vicinity of tree B or further north.178 194 Against this background, the judge then addressed the more fundamental criticism of the reliability of micro fire pattern indicators, and expressed her conclusion on fire vectors, in these terms:179 Criticism of unreliability of micro fire pattern indicators [1069] The respondents criticised the use of protection micro fire pattern indicators on the basis of unreliability, given the potential for movement or environmental degradation. Some protection micro fire pattern indicators were identified on larger objects embedded in the ground which consequently would not have moved. Accordingly, the criticism could not be applied to at least some of the protection fire pattern indicators such as the tree stump identified in the photograph in Mr Woods’ first report. 175 Reasons [1046]. 176 Reasons [1055]-[1059]. 177 Reasons [1063]. 178 Reasons [1064]-[1068]. 179 Reasons [1069]-[1082] (footnotes omitted). Again, in these passages, references to the respondents correspond to the appellants. -- 63 of 69 -- [2026] SASCA 50 The Court 62 [1070] The respondents characterised protection fire pattern indicators as the most vulnerable to misinterpretation including by reason of the potential for a wind shift after the fire to result in the burning of unburnt material and thus leaving misleading indicators. As set out above, Mr Woods was cross-examined on a photograph said to demonstrate the potential for misinterpretation of this nature. Mr Woods’ response to the cross-examination persuasively demonstrated his understanding of such a risk and the need to take it into account in his process. [1071] Mr Woods provided convincing and persuasive explanations for why the grass stems seen in the photographs taken by the CFS and to which he was taken in cross- examination were consistent with, and did not undermine, his interpretation of the fire pattern indicators in the area between the active heap and the cool heap and dead tree. Mr Woods maintained that was a backing area and, accordingly, inconsistent with the lightning tree B theory which would have advancing fire vectors moving from tree B in a generally south-easterly direction. [1072] Some of Mr Woods’ photographs identified lateral fire pattern indicators to the sides of the vicinity of the active heap, some in relatively close proximity to the active heap. The respondents did not address whether there would have been lateral fire pattern indicators in close proximity to the active heap if the fire had moved as an advancing fire from tree B in a southerly direction toward the active heap. [1073] Each of Dr Marsden-Smedley, Mr Cousins and Mr Fenwick would not rely on micro fire pattern indicators, including protection indicators, to draw conclusions a year after the fire. Mr Woods was the only expert who did so. However, Mr Woods was clearly alert to the potential issues and qualifications associated with using protection indicators. These included the potential for items to have moved in the intervening period or for change in the topography, such as soil erosion or plant regrowth. Mr Woods emphasised the importance of viewing the fire pattern indicators in the context of the whole of the fire scene. While Dr Marsden-Smedley would not use micro indicators, the macro fire pattern indicators used by Dr Marsden-Smedley and the macro and micro indicators used by Mr Woods and the CFS investigators immediately after the fire pointed to the same conclusions. [1074] I accept that it is desirable for a fire investigator to use as many available fire pattern indicators and to observe clusters of fire pattern indicators. I also accept that micro fire pattern indicators may be susceptible to environmental factors such as movement or degradation that may reduce their reliability. While no doubt micro pattern indicators could move and be subject to weathering, growth of vegetation, soil erosion and so on, Mr Woods gave evidence he took those factors into account and I accept he did so. Further, it is inherently unlikely that the majority of objects displaying micro fire pattern indicators would have been consistently moved in a certain direction only. Some photographs related to larger objects embedded in the ground, such as the tree stump set out above. Mr Woods was unshakable in his confidence in his identification and interpretation of what he was able to see on the ground, which he put together with the CFS investigation notes and photographs. The different, more cautious approach adopted by Dr Marsden-Smedley did not cause me to doubt Mr Woods’ ability to identify and interpret micro fire pattern indicators after the fire. Mr Woods was not shaken in his cross-examination on the fire pattern indicators. He remained confident in his identification and interpretation of the fire pattern indicators he personally identified and those which appeared in the CFS photographs. I did not consider his confidence excessive or unwarranted. [1075] Mr Woods was an impressive and reliable witness with significant relevant expertise and I have no hesitation in accepting his opinions. Mr Woods presented as a measured, careful and objective expert witness. His evidence was based on what he viewed on the -- 64 of 69 -- [2026] SASCA 50 The Court 63 ground. I accept his evidence that he relied on clusters of fire pattern indicators. I was persuaded that he accurately and reliably identified and interpreted the fire pattern indicators. I accept Mr Woods’ evidence that he had the competence and ability to identify and interpret fire pattern indicators including protection fire pattern indicators at the scene when he attended. Conclusion on fire vectors [1076] I accept Mr Woods’ evidence that the fire pattern indicators he identified and interpreted displayed an advancing fire vector generally to the south of the active heap, lateral fire vectors to the sides and a backing vector generally to the north of the active heap. [1077] Mr Woods’ observations were supported by the CFS photographs and CFS observations. Both Dr Marsden-Smedley and Mr Woods reviewed the CFS reports and photographs which captured at least some micro fire pattern indicators at the time of the fire. The CFS investigators and each of Mr Woods and Dr Marsden-Smedley concluded the fire pattern indicators demonstrated a head fire advancing from the vicinity of the active heap. The CFS investigators and Mr Woods concluded the fire pattern indicators demonstrated a backing fire backing from the vicinity of the active heap. Dr Marsden- Smedley concluded the fire moved back from the vicinity of the active heap, in his reply report characterising the fire movement as a part flank part backing fire. [1078] I accept Mr Cousins possesses expertise in fire ignition and fire behaviour. However, he lacks expertise in bushfire spread and bushfire investigation as set out above. While Mr Fenwick has bushfire behaviour experience based on his many years of practical experience, he was not asked to investigate origin and cause. His instructions were initially limited to asking whether or not he agreed with Dr Marsden-Smedley and Mr Woods and later, whether he agreed with Mr Cousins. While I accept that he may have expressed himself in a casual manner, the tone of his correspondence suggested lack of objectivity. Mr Fenwick’s limited role and the number of matters which he did not take into account told against accepting his opinions over those of Dr Marsden-Smedley and Mr Woods. [1079] I prefer the evidence of Dr Marsden-Smedley and Mr Woods over that of Mr Cousins and Mr Fenwick. I address separately below the respondents’ criticisms of asserted inconsistencies as between Dr Marsden-Smedley’s and Mr Woods’ approaches. [1080] Consistency in identification and interpretation of fire pattern indicators does not necessarily imply a form of assumption or confirmation bias. It is also consistent with the use of an appropriate methodology and with different investigators reaching the same or similar conclusions. I do not accept that Mr Woods’ methodology suffered from a failure to comply with the NWCG methodology or was tainted by assumption or confirmation bias. [1081] While the CFS report notes refer to confusing backing fire indicators, Mr Woods’ analysis was comprehensive. His evidence supports findings of clear backing fire indicators north of the active heap and the cool heap. I have taken into account Ms Rice’s evidence explaining her reference to confusing indicators. She did not accept that the reference to confusing indicators meant there were indicators suggesting both heading and backing fire were present and rather that the confusing indicators were more of a challenge than one would usually find. I do not consider Ms Rice’s view that the indicators were more confusing than usual sufficient to disregard her evidence or the conclusions concerning the fire vectors she identified or to cast significant doubt on Mr Woods’ opinions. Erratic wind changes as described by Mr England contributing to localised -- 65 of 69 -- [2026] SASCA 50 The Court 64 changes in fire movement direction could conceivably have contributed to the confusion in indicators. [1082] There was no evidence of any advancing fire pattern indicators travelling from the vicinity of tree B to the cool heap and the active heap. I accept that no expert examined the area roughly between tree B and the cool heap for fire pattern indicators. The CFS investigators located backing fire indicators in that region. The macro fire pattern indicators identified by Dr Marsden-Smedley and the macro and micro fire pattern indicators identified by Mr Woods, SAPOL and CFS support the movement of the fire as an advancing vector from the active heap and as a backing vector from the active heap. As set out above, I do not accept the criticisms of the CFS methodology. Dr Marsden- Smedley’s identification of macro fire pattern indicators was not challenged. I do not accept the criticisms of Mr Woods’ methodology or his identification or interpretation of fire pattern indicators. 195 Respectfully, the appellants contentions on appeal do not sufficiently engage with the layered reasoning of the judge, and essentially invite the appellate court to disagree on the basis of several relatively high-level contentions. We have considered these contentions but they do not persuade us that the judge’s analysis was wrong. 196 Amongst those propositions is that the failure by Dr Marsden-Smedley and Mr Woods to reconsider the lightning tree B hypothesis once it became apparent the CFS was wrong to assume no lightning strikes might have occurred constituted ‘a failure by them of their obligation as experts to make all proper inquiries’.180 The appellants then contend that the absence of ‘any substantive investigation into this alternative hypothesis ought to have been explicitly addressed in the [judge’s] evaluation of the evidentiary foundation underlying the active heap hypothesis’.181 Putting the adjective ‘substantive’ to one side, the answer to the submission is that the issue was addressed by the judge, in several places.182 197 The appellants also submitted in writing that the judge ‘misse[d] the point’ of the significance of the investigators Downs and Rice having reported the fire’s cause as ‘undetermined’. They submitted that:183 The cause was listed as ‘undetermined’ because the investigators were not satisfied with the cause. This is a matter which the Judge ought to have considered when she was considering whether the respondents had proved on the balance of probabilities that the cause of the fire was the ignition of the active heap. She did not do so. 198 In our view this submission does not advance the appellants’ case. The judge accepted the honesty and reliability of both witnesses.184 They expressed views that were consistent with the active heap theory, though of course not conclusive. Even assuming they had nothing but the slightest suspicion that the active heap was the cause of the fire, this was not a matter that could have played any 180 Appellants’ written submissions [80]. 181 Appellants’ written submissions [80]. 182 See, eg, Reasons [664], [668], [678], [1067]-[1068], [1082]. 183 Appellants’ written submissions [79]. 184 Reasons [754]. -- 66 of 69 -- [2026] SASCA 50 The Court 65 significant role in the view the judge might form having heard extended expert evidence and with the benefit of detailed submissions. By contrast, Ms Rice had no background information at the time about the constitution of the active heap. 199 As has been mentioned, a further and important part of the appellants’ submissions was that there was a difficulty in, on the one hand, finding that the CG&C rate was not applicable because the fire proceeded partly as a flanking fire and partly as a backing fire and, on the other hand, accepting the purport of Mr Woods’ evidence regarding fire pattern indicators as supporting the active heap hypothesis. The asserted inconsistency is that Mr Woods identified backing fire patterns to the north of the active heap. 200 The judge dealt with this issue immediately following the passages extracted above, and followed her analysis with her ultimate findings on fire pattern indicators and fire pattern movement:185 Asserted inconsistencies between Dr Marsden-Smedley and Mr Woods [1083] The respondents relied on asserted inconsistency between Dr Marsden-Smedley’s half-flank, half-backing fire and Mr Woods identification of a backing fire from the vicinity of the active heap. This was based on Mr Woods identifying fire pattern indicators to demonstrate the fire progressed north against the direction of the prevailing wind. The respondents criticised Mr Woods on the basis he did not refer to fire pattern indicators suggesting that north of the active heap there was a hybrid between a backing fire and a flank fire; he did not identify flank indicators north of the active heap, cool heap and the dead tree; and, if Mr Woods could differentiate between the different indicators, he should have located flank indicators behind the postulated point of origin. [1084] Dr Marsden-Smedley’s half-back, half-flank opinion concerned fire movement at about the half-way point between the straight back of the fire and the straight flank taking into account the relative angles of the prevailing wind, the active heap and the geolocated tree. The diagrammatic depiction of an elliptical fire shape in the Cheney and Sullivan text delineates the backing fire as about 45 degrees either side of the point of origin and Mr Fenwick treated the backing fire as encompassing that area. As set out above, the description in the NWCG Guide of a backing fire vector includes microscale fire pattern indicators and less damage relative to advancing and lateral areas. It describes lateral fire vectors as having the characteristics of backing or advancing fires with lower intensity flanks leaving backing type indicators exhibiting a more subtle and wider transition zone than higher intensity flanks which may leave indicators consistent with advancing fire spread. The delineation between backing and flanking indicators is thus not absolute or precise. [1085] Mr Woods considered that a backing fire under circumstances of variation in wind speed and direction would become more of a lateral fire in some circumstances. In the photographs of micro fire pattern indicators taken by the CFS investigators, Mr Woods identified transition zones where the fire changed direction because of the influence of wind, including a transition zone that pushed the vector as more of a lateral fire up towards the ridge on the north-eastern side. Those micro fire pattern indicators had disappeared by the time he conducted his site visit. As set out above, Mr Woods also compared grass stem indicators in CFS photographs which identified transition zones where the fire changed 185 Reasons [1083]-[1091] (footnotes omitted). -- 67 of 69 -- [2026] SASCA 50 The Court 66 direction. Indications of grass stem fire pattern indicators confirmed to Mr Woods that even in the backing area, the fire changed from a backing fire to somewhat of a lateral fire and back again.\ Accordingly, while tracking the fire movement as a backing fire vector from the active heap, Mr Woods’ evidence accounted for movement within that vector in the way described. Mr Woods did not inspect the particular area shown in the photographs taken by Mr England. Mr Woods said Mr England’s evidence mirrored what he would have expected to see on the fire ground. [1086] As set out above, Mr Woods’ process of selecting fire pattern indicators entailed selecting indicators representative of a particular vector. He said that in accordance with the NWCG Guide, it was not necessary to document fire pattern indicators inconsistent with the overall fire direction. [1087] Thus, it does not necessarily follow that Mr Woods’ evidence about locating and identifying backing fire pattern indicators and his conclusion there was a backing fire vector from the active heap moving towards the area of the cool heap and the dead tree and Dr Marsden-Smedley’s view that the fire movement from the active heap towards the geolocated tree was more half-back half-flank cannot be reconciled. [1088] However, if I am wrong in that view, to the extent of any such inconsistency I rely on Mr Woods’ opinions based on the fire pattern indicators. He engaged in the detailed analysis to which I have referred, identified and interpreted evidence on the ground at his site visit and in the investigation photographs and supported his analysis with cogent explanations for his interpretation of the fire pattern indicators he identified, his explanation of the transition zones and the conclusions he reached. Findings on fire pattern indicators and fire movement [1089] For the reasons set out above, it is not possible for me to make findings about where and how the fire moved from the time it started to the location photographed by Mr England at 1:18 pm. [1090] I find that fire pattern indicators indicated an advancing fire vector in an approximately southerly vicinity of the active heap, lateral fire vectors to the sides of the active heap, approximately to the east and west, and a backing fire vector to approximately to the north of the active heap and, accordingly, between the active heap and tree B. The fire pattern indicators identified by the investigators were consistent with a backing fire moving to the north from the active heap and an advancing fire moving to the south of the active heap. Scars on trees identified by Dr Marsden-Smedley pointed to the same conclusion as did the macro and micro fire pattern indicators identified and interpreted by Mr Woods. The fire pattern indicators were inconsistent with fire having originated in or in the vicinity of tree B. They were consistent with the fire having originated in the vicinity of the active heap. [1091] Based on all the matters to which I have referred above, I find the fire originated from the active heap. 201 The appellants submitted that the judge ‘did not address a central plank of the appellants’ attack on Woods, namely that if (as Woods claimed and the Judge accepted) fire pattern indicators are so reliable in identifying fire behaviour, then this must mean that the backing fire indicators identified north of the active heap -- 68 of 69 -- [2026] SASCA 50 The Court 67 establish that the fire was a backing fire’.186 In our respectful view, the judge’s reasons, set out above, did address that proposition. 202 The appellants’ submissions on appeal did not provide a basis upon which to conclude that the judge’s analysis of Mr Woods’ evidence about transition zones was wrong. Further, even if the better view of Mr Woods’ evidence was that the fire progressed as a backing, not hybrid flanking and backing, fire, this was not destructive of the judge’s reasoning concerning the CG&C rate. The reasons why the judge did not consider that the backing spread rate was inconsistent with the active heap theory were, as has been shown, multi-layered. 203 Standing back, and considering the appellants’ contentions in combination, they do not demonstrate error in relation to the causation finding. The judge’s findings that the heap was lit in mid-2020, was constructed in a way that facilitated smouldering combustion, was not surrounded by fire breaks, and was capable of leading to flaming combustion in the weather conditions on 11 January 2021, have not been challenged. There was strong support from experts with relevant experience, and who were considered to have presented as clear and compelling witnesses, for the proposition that the active heap was the likely origin. 204 Of the other potential causes, the lightning tree B theory was found to be inherently very unlikely, and no other plausible causes were hypothesised by any of the experts. That of course did not demand acceptance of the active heap theory, but had the lightning tree B theory remained a plausible theory, this would likely have stood in the way of an actual sense of persuasion that the active heap was the cause, given that the support for the active heap theory was circumstantial. 205 Having regard to the basal and intermediate findings made by the judge that were not the subject of challenge and the expert evidence given by experts who the judge considered gave cogent and reliable evidence based upon their relevant expertise, the ultimate findings about the origin of the fire have not been shown to be contrary to compelling inferences or inconsistent with incontrovertible facts. It has not been demonstrated that the findings were wrong. Disposition 206 The appeal must be dismissed. 186 Appellants’ written submissions [67]. -- 69 of 69 --