High-severity wildfires in temperate Australian forests have increased in extent and aggregation in recent decades.

Wildfires have increased in size and frequency in recent decades in many biomes, but have they also become more severe? This question remains under-examined despite fire severity being a critical aspect of fire regimes that indicates fire impacts on ecosystem attributes and associated post-fire recovery. We conducted a retrospective analysis of wildfires larger than 1000 ha in south-eastern Australia to examine the extent and spatial pattern of high-severity burned areas between 1987 and 2017. High-severity maps were generated from Landsat remote sensing imagery. Total and proportional high-severity burned area increased through time. The number of high-severity patches per year remained unchanged but variability in patch size increased, and patches became more aggregated and more irregular in shape. Our results confirm that wildfires in southern Australia have become more severe. This shift in fire regime may have critical consequences for ecosystem dynamics, as fire-adapted temperate forests are more likely to be burned at high severities relative to historical ranges, a trend that seems set to continue under projections of a hotter, drier climate in south-eastern Australia.

Introduction

Wildfire shapes landscape patterns and ecosystem processes as it determines both vegetation distribution and structure [1, 2]. Changes in wildfire activity may alter mortality and regeneration patterns, initiating new successional pathways that ultimately lead to shifts in vegetation composition and landscape attributes [3]. Many studies over the past decades have reported a change in wildfire activity including increases in the frequency, size, and duration of wildfires, as well as the length of the fire season [4-8]. Such increases have been linked to climate change, which influences key fire drivers like fuel accumulation and availability [9-11]. Models based on climate change projections suggest that this trend in increasing fire activity will continue into the future [3, 12–15] posing threats to forest resilience, including shifts to lower density forests or non-forest states [16-18].

Fire severity is a wildfire attribute that quantifies the degree of environmental change caused by fire including immediate fuel consumption and carbon emissions and longer-term impacts on vegetation mortality, successional pathways, and soil substrate [19]. Wildfire severity is spatially heterogeneous and can range from partial litter consumption and light scorching of understorey vegetation to near complete mortality of canopy trees [19-21]. Fire severity and the spatial configuration of severity classes have critical implications for fire-related resilience and potential degradation of ecosystems [21-25]. Wildfire severity is related to fire intensity, which is driven by fuel, climate, and weather [26-29]. As such, fire severity, as for other components of fire regimes, has likely been affected by changing climates in recent decades [30, 31]. In contrast to the large number of studies that have documented recent increases in wildfire area and frequency [9, 32–34], comparatively fewer studies, mostly focused on North America forests, have investigated trends in fire severity, some indicating increases while others indicating no change or decreases [35-37]. Changes in wildfire severity can influence ecological processes by affecting the trajectory of postfire vegetation succession, leading to reductions in forest cover and even conversions to non-forested vegetation [38, 39]. A better understanding of changes in fire severity is crucial to foresee the future pathways of forest systems [40-44].

Australia is one of the most fire-prone countries worldwide [45, 46] with 30.4 million hectares burned across Australia in 2019–2020 alone [47]. Studies have highlighted how climate change has and will continue to impact Australian fire weather and fire activity [31, 48, 49] with fires predicted to become larger and more frequent [50-52]. Whether fires have also become more severe remains largely undocumented. This study’s principal objective was to examine patterns in high-severity fires in temperate forests of the state of Victoria, south-eastern Australia over the last three decades. Specifically, we addressed three questions: 1) Has the area burnt by high- severity fire in temperate forests of Victoria increased in the last 30 years?; 2) Has the spatial configuration of high-severity patches in the landscape changed in the last 30 years; and 3) Are the observed trends consistent across bioclimatic regions?

Materials and methods

Study area and forest types

This study was conducted across the state of Victoria, south-eastern Australia, an area that encompasses 237,659 km2, ranges from 0 to 1986 m a.s.l in elevation and comprises several geographical bioregions with differing geology, soils, climate, and predominant vegetation (Table 1 and Fig 1) [53]. Climate across Victoria is temperate with warm to hot summers (average maximum temperature between 16 °C and 30 °C; [54]). The annual mean temperature ranges from 12.6 °C in the south-east region to 14.7 °C in the north and north-west regions of the state [55]. The mean annual precipitation varies from 500 to 2,200 mm, with precipitation over 1000 mm in the mountainous areas of the Great Dividing Range [56]. Over the past few decades, Victoria has become warmer and drier, consistent with global trends, and these trends are likely to continue [57-59].

Table 1

Characteristics of the bioregions in the study area affected by the selected 162 fires.

Bioregion		Major forest types a	Height (m)	Projective Foliage Cover (%)	Regeneration strategy b	Elevation (m)	MAT (°C)	MAP (mm)	No of fires	Total burnt area (ha)	Total high-severity burnt area (ha)
AA	Australian Alps	High Altitude Shrubland/ Woodland	15	10–30	R	844–1996	4.5–12.6	712–1996	9	1,426,791	290,073
		Riverine Woodland/Forest	15	10–30	R
MDD	Murray Darling Depression	Lowan Mallee	7	10–30	R	265–690	12.8–17.2	265–702	52	514,689	358,238
		Riverine Woodland/Forest	15	10–30	R
SCP	South East Coastal Plain	Riverine Woodland/Forest	15	10–30	R	492–1260	11.4–14.9	494–1306	10	40,375	8,745
SEC	South East Corner	Moist Forest	30	70–100	S	664–1184	7.3–15.2	656–1292	17	170,045	18,700
		Riverine Woodland/Forest	15	10–30	R
SEH	South Eastern Highlands	Grassy/Heathy Dry Forest	10–30	10–30	R	681–1922	6.6–14.8	645–1942	17	995,133	170,452
		Moist Forest	30	70–100	S
VM	Victorian Midlands	Forby Forest	15–30	30–70	R	418–1411	8.5–15.3	418–1490	46	404,363	156,083
VVP	Victorian Volcanic Plain	Moist Forest	30	70–100	S	477–1026	11–14.9	476–1026	11	165,003	79,022

Bioregion name and acronym [53], major forest types in each bioregion affected by the selected wildfires, height, projective foliage cover and regeneration strategy of the dominant species in each forest type, elevation range, mean annual temperature (MAT) and annual precipitation (MAP) range [65]; Number of wildfires included in this study (i.e. 162 wildfires greater than 1000 ha, occurred between 1987 and 2017 and with available Landsat imagery) and their cumulative total [64] and high-severity burnt area (as estimated in this study).

a Major forest types were adopted from EVD names and associated structural data [66]. Dominant tree species were derived from the Ecological Vegetation Classes (EVC) benchmarks database [67];

b R: resprouter; S: obligate seeder, classifications based on predominant fire-response traits of dominant tree species [62, 68, 69].

Fig 1

Map of study area.

(i) Victoria highlighted (grey) in the map of Australia; (ii) Locations of study areas within the state of Victoria in south-eastern Australia. Red points rrepresent the centroids of the 162 wildfires investigated in this study. Colours relate to bioregions (Acronyms are defined in Table 1).

Vegetation affected by the studied wildfires was predominantly comprised of a range of Eucalyptus forests of varying composition, structure and post-fire regeneration strategies [60] (Table 1). These included Mallee, with low canopy height (7 m) and sparse canopy cover (25%), Woodlands with medium canopy height (15 m) and sparse canopy cover, Open forests, with medium to tall canopy height (10–30 m) and mid-dense canopy cover (30–70%) and Closed forests, with tall canopy height (30 m) and dense canopy cover (70–100%) [61]. Obligate seeder tree species are dominant in Closed forest whereas resprouter eucalypts (basal or epicormic) are dominant in all other forest types [60, 62, 63].

Fire history dataset

We used the wildfire history data available from the Victorian Department of Environment, Land, Water & Planning (‘DELWP’; [64]). Data contained the spatial extent of wildfires since 1926 and, for the most recent fires (from 1998 onward), the start date of the fire. For this study we selected the subset of wildfires that occurred between 1987 and 2017 and that had a minimum burned area of 1000 ha to ensure the fire size was sufficient to include multiple fire-severity levels. That amounted to 211 wildfires that were used to assess changes in the number of fires per year and mean fire size between 1987 and 2017. Each fire was classified according to its dominant bioregion [53]. For the purpose of assessing changes in fire severity, 32 of the 211 wildfires were discarded because pre- or post-fire remote sensing images were unavailable, and 11 were discarded because clouds covered more than 25% of the fire affected area, which may affect the spatial metrics assessed in our study. In total, a subset of 162 wildfires, with at least two fires per year over the past three decades, was used to generate fire-severity maps and analyse changes in severity patterns.

Remote sensing dataset and spectral indices

Wildfire severity of the selected 162 fires was mapped using Landsat TM, ETM+ and Landsat 8 imagery (30 m spatial resolution, all from Landsat Collection 1, Tier 1). Pre- and post-fire images were selected for each wildfire based on the recorded fire start dates, which were predominantly in the summer months (December to February). Images were selected within two months before and after the fire to minimise differences in forest phenology and general atmospheric conditions at the time of acquisition. When only the fire year but not start date was recorded (~13% of the fires), we conducted a visual inspection of all images available for the fire season, identified the image where the fire scar was first visible and selected that image and the previous one as post- and pre-fire images respectively for that event. A total of 347 Landsat images including 228 scenes of Landsat 5 (TM), 36 scenes of Landsat 7 (ETM+), and 83 scenes of Landsat 8 (OLI/TIRS) were selected and obtained through the US Geological Survey (USGS) EarthExplorer at http://earthexplorer.usgs.gov as higher level surface reflectance products for each fire. The images were masked for clouds and shadows using the Fmask algorithm [70], which has an accuracy of about 96% [71].

Four spectral indices, namely NBR, NDVI, NDWI, and MSAVI, and their temporal differences (i.e. delta versions, which calculate the change between pre-fire and post-fire spectral index values) were computed for each of the 162 wildfires. These indices are commonly used to assess fire severity [72-76] and were identified by the authors, in a previous study, as the optimal spectral indices for mapping fire severity in the forest types of the study area [77].

Fire severity mapping

Severity of the wildfires in Victoria has not been consistently recorded, with historic fire severity mapping only available for nine years in the period between 1998 and 2014 [78]. To generate fire severity maps for the 162 selected wildfires ensuring the consistency of the classification we used a Random Forest model based on spectral indices that had been previously trained and validated by the authors for the same study area [61]. The reference fire-severity dataset used for training and validation was comprised of 3730 plots from eight large wildfires (>5,000 ha) that occurred between 1998 and 2009 and covered 13 forest types differing in species composition, canopy cover, canopy height and regeneration strategy. These forest types match those affected by the 162 wildfires of this study. Fire severity of the 3730 reference plots had been assessed in situ or visually interpreted on very high resolution orthophotos by the Department of Environment, Water & Planning (DELWP) [78]. Severity was classified as Unburnt: less than 1% of eucalypt and non-eucalypt crowns scorched; Low severity: light scorch of 1–35% of eucalypt and non-eucalypt crowns; Moderate severity: 30–65% of eucalypt and non-eucalypt crowns scorched; or High severity: 70–100% of eucalypt and non-eucalypt crowns burnt [79]. Overall, the reference data included a minimum of 20 plots for each forest type and fire-severity class combination. The Random Forest model was trained with 60% of the data and used 12 predictor variables, which included the four optimal SI indices (dNBR, dNDVI, dNDWI, and dMSAVI) and their pre- and post- fire values. Model accuracy was tested on the remaining 40% of the data that had been left for model validation. Accuracy for high-severity mapping was very high, with a commission error (plots wrongly attributed to high severity) of 0.06 and an omission error (high severity plots incorrectly classified) of 0.18.

Metrics of high-severity fire

Based on the high-severity maps of each of the 162 wildfires, we calculated eight landscape metrics to characterize the extent and spatial configuration of the high-severity burned area. Extent metrics included total and proportional high-severity burned area. Spatial configuration metrics were calculated at the patch level, i.e. areas of high-severity fire surrounded by different severities within the wildfire boundary. Spatial configuration metrics included two patch size metrics (mean patch size, coefficient of variation of patch size), two fragmentation metrics (number of patches, and edge density—a measure of shape complexity) and two aggregation metrics (clumpiness and normalized landscape shape index–NLSI, S1 Table of S1 File). Edge density is the ratio between the total length (m) of the edges of the high-severity patches and the fire size (i.e. total wildfire area burnt at any severity; ha). Low edge density values represent simple shape (e.g. circular) and/or large patches, while large values indicate irregular and/or less continuous patches [80]. Clumpiness and NLSI, both unitless, quantify patch aggregation. The former is based on the likelihood of adjacent pixels belonging to the same class, whereas the later measures the deviation from the hypothetical minimum edge length of the class. Increasing levels of aggregation (i.e. increasing clumsiness and decreasing NLSI) represent more compact and simpler-shaped patches [80, 81]. These metrics describe different aspects of landscape configuration but were not completely independent and therefore should be interpreted jointly (S1 Table of S1 File). Spatial pattern metrics were obtained using the ‘landscapemetric’ package [82] in the R statistical software [83].

Data analysis

Linear regression models were used to evaluate the trends in high-severity fire metrics from 1987 to 2017, with individual fires as the sampling unit. We built two groups of models, a state-wide model (n = 162 fires) and separate bioregion models. The response variables for both groups of models were the extent or landscape configuration metrics of the high-severity burned area. Predictor variables included year and fire size (i.e. total wildfire area, ha) as fixed effects and bioregion as a random effect, which was only included in the state-wide mixed effects models. Fire size was included as covariate in all models as it can be related to burn patterns [27] and was not correlated with fire year (Pearson’s r = -0.01). Data were transformed when needed to meet assumptions of normality (S1 Table of S1 File). All statistical tests were conducted in the statistical programming language R [83].

Results

Changes in area and proportion of high-severity fire over time

Based on the fire history dataset (n = 211), the number of wildfires per year larger than 1000 ha between 1987 and 2017 increased significantly (P = 0.012), a trend that was mostly due to an increase since 2000 (Fig 2). In contrast, we detected no significant change in total fire size (i.e. all fire severities combined) over that period.

Fig 2

Changes in the number of fires per year and fire size between 1987 and 2017.

Data includes all wildfires ≥ 1000 ha from DEWLP fire history dataset (n = 211) [64]. Solid black line indicates significant relationship (P<0.05), dashed grey line indicates no significant relationship.

Between 1987 and 2017 the area burnt by high-severity fire increased significantly (PYear <0.001) even when accounting for total fire size (PFire size <0.001; Fig 3 and S1 Fig of S1 File). The same trend was observed for the proportion of the area burnt by high-severity fire (PYear <0.001; Fig 3). Estimated changes in the area and the proportion of area burnt by high-severity fire over time by bioregions were positive and significant (or marginally significant 0.05 < P <0.1) in all cases (Fig 3 and S2-S3 Figs of S1 File). The studied bioregions supported quite distinct forest types, from wet, tall, and highly productive to dry, open, and less productive. This suggests that the observed increases in the area burnt by high-severity fire was ubiquitous across regions and did not depend on local environmental conditions or forest types.

Fig 3

Changes in the area and proportional area of high-severity fire from 1987 to 2017.

Left panels: Area and proportional area burnt by high-severity fire in each of 162 wildfires (line represents significant relationship between variables). Right panels: Standardized coefficients for high-severity area (top, log transformed) and the proportion high-severity area (bottom, arcsine transformed) indicating the relationship between area burnt and time. Each panel displays results for a single model for all regions (“Victoria”) and for individual bioregions (Acronyms of bioregions are defined in Table 1); Dot points represent mean estimated coefficient along with the 90th (solid line) and 95th (dashed line) percentile intervals. Coefficients denote significant changes when interval does not include zero.

Changes in spatial patterns of high-severity fire

We detected no changes in fragmentation of wildfires between 1987 and 2017 as evidenced by no significant increases in the number of high-severity patches, a result that was consistent across all bioregions (Figs 4 and 5 and S4 Fig of S1 File). In contrast, edge density, which is related to patch shape complexity, increased over time across Victoria (PVictoria = 0.006), although this trend was only (marginally) significant for the SEC, VM, VVP bioregions (0.05 < PYear < 0.1; Fig 5 and S5 Fig of S1 File). While mean high-severity patch size did not change significantly, the coefficient of variation of patch size, which was related to fire size, increased in all models (PYear<0.05 and PFire size<0.001; Figs 4 and 5 and S6-S7 Figs of S1 File). Accordingly, we detected an increase in the size of the largest patch (PYear = 0.005; S8 and S9 Figs of S1 File). The level of patch aggregation measured through increased clumpiness and/or decreased Normalized Landscape Shape Index (NLSI), also increased from 1987 to 2017 (Figs 4 and 5 and S10 and S11 Figs of S1 File). This trend, which was significant both at the state and bioregion level, suggests the patterns in high-severity fire changed from a more random, highly-dispersed distribution of patches towards fewer, larger patches of irregular shape that were more aggregated within the fire boundaries.

Fig 4

Changes in high-severity spatial metrics over time.

Each subplot displays a scatterplot between the Year of the fire and the defined high-severity spatial metric. Dots represent each of the 162 wildfires. Values are the results for single mixed effects models where Year and Fire size are fixed effects and Bioregion is a random effect. Lines represent significant (solid black) or not significant (dashed grey) linear relationships.

Fig 5

Estimated coefficients for high-severity spatial metrics by bioregions.

Each panel displays results for a single model for all regions (“Victoria”) and for individual bioregions (Acronyms of bioregions are defined in Table 1); Dot points represent mean estimated coefficient along with the 90th (solid line) and 95th (dashed line) percentile intervals. Coefficients denote significant changes when interval does not include zero. Spatial metrics were log transformed (Number of Patches, Mean Patch Area, Variation Patch Area, NLSI) or arcsine transformed (Edge Density).

Discussion

Our study assessed for the first-time changes in high-fire severity patterns since 1987 in Victoria, south-eastern Australia. We detected an increase in the area burnt at high-severity during that period and a shift in the landscape configuration of high-severity patches, which was consistent across most bioregions, encompassing a broad range of forest types.

The area of high-severity fire has increased

Our results showed an increasing trend in both total and proportion of high-severity burned area between 1987 and 2017 across various temperate forests types in south-eastern Australia. Our findings are in contrast to similar studies conducted in the US where either an increase in fire severity was not detected [37, 84] or the detected increase was due to increasing fire size [36]. Our results also show a covariation between fire size and the extent of the area burned by high-severity fire, a pattern that has been documented before in several north American forests [4, 27, 85–87].

The increasing trends in total and proportion of high-severity burned area at the state level were consistent across all bioregions, indicating that these changes occurred irrespective of forest type and climatic region. This is in contrast to the mixed fire-severity trends assessed across regions in North America [37, 88], which have been argued to be related to fire suppression policies masking climate-change effects [84, 88].

Changes in the area of high-severity fire like those described here have been predicted to occur as a result of climate change since decades ago [89-91]. Our results confirm for the first time that wildfires in south-east Australia are indeed becoming more severe and, given projections of a hotter, drier climate [59], this pattern seems set to continue in coming decades.

Trends in landscape configuration: Aggregation of high-severity patches

Our results showed changes in the landscape configuration of high-severity patches that were consistent at the state level and across bioregions. While we did not detect a significant shift in patch number or mean patch size, we noted an increase in patch size variability, patch shape complexity (measured as edge density) and patch aggregation (as evidenced by trends in clumpiness and NLSI). These changes suggest that the areas burned by high-severity fire have become more aggregated, more irregular in shape, and have a larger area occupied by the largest patch. Similar changes in spatial patterns of high-severity fire have also been reported in fire-severity research in North America [27, 88, 92], where increasing patch aggregation was related to the increased proportion of high-severity area [42].

Implications of increasing high-severity fire for temperate forests in south-east Australia

Our quantified increases in high-severity burned area can lead to concerns about the resilience of Victoria’s temperate forests [20, 93, 94], similar to those expressed for other forest types elsewhere [4, 92, 95]. High-severity fire influences ecosystem dynamics with effects on vegetation succession [25, 96, 97], biogeochemical processes [21, 26, 98], geomorphic processes [99, 100], and habitat availability and biodiversity [23, 101, 102]. Recent high-severity fires within our study area have led to increased mortality of fire-tolerant eucalypt trees and to an increase in the density of young trees vulnerable to subsequent fires [20, 63, 103]. If increasing trends in the extent of high-severity fire detected in our study continue, this indicates potential for large-scale changes in key structural attributes of even the most fire-tolerant forests.

High-severity fire impacts can be modulated by the size, shape, and configuration of high-severity patches. For instance, patch size and aggregation can influence runoff connectivity and post-fire sediment yields and affect the distribution of low- and moderate-severity patches that serve as refuges for fire-sensitive species [104-106]. Patch size and spatial configuration can also affect dispersal and subsequently influence vegetation succession potentially leading to forest-type conversions [107-109]. Delays in tree re-establishment following high-severity fires has been detected in non-serotinous forests of the United States and Canada due to a rapid and extensive shrub establishment via persistent soil seedbanks [109, 110]. Eucalypt forests in south-eastern Australia, including those affected by the studied wildfires, are dominated by either resprouter species that survive most fires, or obligate seeder species that rely on a canopy seedbank to regenerate after fire [63, 111]. Seed dispersal in both resprouters and obligate seeders eucalypt forests is limited to one or two tree heights, with seeds lacking attributes to facilitate animal or wind dispersal [112]. Resprouters’ seed viability decreases with fire intensity [113] and therefore regeneration in high-severity patches may depend on dispersal from adjacent moderate-severity or unburned patches (although see [20] indicating prolific regeneration from seed of resprouter eucalypts after a single high-severity wildfire). Increases in high-severity patch size though aggregation as observed in this study could hinder post-fire tree establishment by increasing distances from seed source and also altering the regeneration abiotic environment [114] contributing to feedbacks that result in an increased risk of forest-type conversion [115, 116]. Spatial configuration of high-severity patches can also influence regeneration of obligate seeder forests burnt by recurrent fires in quick succession (~20 years; [103]). In such circumstances, trees regenerating after the first fire would not have yet produced meaningful quantities of viable seed before a second fire [117], and eucalypt regeneration would rely on seed dispersal from adjacent patches. Lack of tree regeneration after short-interval fires in obligate seeder forests has been observed in the last decades with aerial sowing being required to address post-fire recovery in obligate seeder forests [118]. This highlights the impact that the observed changes in fire regimes have had on the resilience of eucalypt forests in south-eastern Australia [63, 103].

Conclusions

Changes in high-severity fire, its extent and spatial configuration, can alter a range of ecosystem processes that interactively determine post-fire recovery, including the conversion to non-forest alternative states. Our analysis showed an increase in both the total and proportion of high-severity burned area in Victoria between 1987 and 2017. Over that period, high-severity patches have become more aggregated and more irregular in shape. These trends were consistent across bioregions encompassing a diversity of forest types. Shifts in the spatial patterns of high-severity fire over time may have cascading effects on forest ecology, highlighting the increased threat posed by changing fire regimes to forests ecosystems.

(DOCX)

Click here for additional data file.

24 Jun 2020

PONE-D-20-11188

High-severity wildfires in temperate Australian forests have increased in extent and aggregation in recent decades

PLOS ONE

Dear Dr. TRAN,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Aug 08 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Krishna Prasad Vadrevu, Ph.D

Academic Editor

PLOS ONE

Additional Editor Comments (if provided):

Dear authors,

Please kindly address the important concerns raised by the reviewers as below:

-Address the novelty and terminology issues

-Clarifications relating to patch shape complexity and inherent image quality issues.

-Explanation regarding the use of fixed thresholds with spectral indices is not ideal for sensitive detection because it does not take into account local conditions (soil type, drought etc).

-More details and justification on the fire severity mapping technique.

Please see the detailed comments below.

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1) Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2) Please remove your figures from within your manuscript file, leaving only the individual TIFF/EPS image files, uploaded separately.  These will be automatically included in the reviewers’ PDF.

3) Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: No

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This study examines high severity fires in the state of Victoria Australia through a retrospective analysis of wildfires over the last 30 years. The authors conclude that the number of high severity fires has increased over time and speculate that this trend can be of great consequence to the fire-adapted temperate forests of Australia.

While I believe that studies of high severity fires are very important and highly relevant, I cannot see the novelty of this study, specifically, the difference between this study, which the authors claim is ‘for the first time’ (L 287 and L 311), and fire severity maps of [64]? From Table 1 and the Methods it seems that the temporal changes of fire severity have already been mapped on the State level prior to this work and those are of high quality based on in situ and high-resolution image confirmation.

The section on fire severity mapping (L 164) is not clear. Because [64] contains the total and high severity burnt areas (Table 1), and years of fires (the Methods), why authors repeated the analysis using Landsat images? Yet analysed changes in the number of fires and fire size using the data from [64], Fig 2? In the Methods section the authors say they excluded 43 fires and only 162 fires were analysed (L 129), while Fig 2 is based on 211 fires from [64], L 227. What analysis is based on [64] and on images processed in this study?

There is also unclear terminology, e.g. what is the ‘patch’? I understand it reflects high severity burnt areas, but only early in the text it says so (L 86?) and no definition is given when the patch analysis is described.

Please clarify what is the difference between ‘fire area’ and ‘total fire size’ L 209-212? From Fig 2 it looks like unit of ‘fire size’ is ha, and by reading further it seems that the total fire size includes all severity classes, L 222, but it can be only guessed if ‘fire area’ relates to high severity areas? Then how does fire area relate to patch? Is it an aggregation of patches?

Fig 4, because ‘patch’ is not clearly defined, it’s not clear if the analysis relates to the burnt area, high severity area or..?

In the discussion, can it be that the changes in patch shape complexity relate to the image quality as the data from earlier years would come from less sophisticated satellite images such as Landsat 5 rather than indicate increasing severity of fires?

L 329, the reference [86] from California’s conifer forests is not highly relevant to the forest resilience statement regarding fire adapted Eucalyptus of Australian forests. Are there conifer forests in the state of Victoria, subjected to high severity burns?

In conclusion, studies of fire severity changes over time are highly important and relevant yet a clear separation of novelty of this work vs already conducted state level analysis is required.

Reviewer #2: This is a very well written, rigorous and timely paper that leverages the increasing ease of analysing the Landsat satellite archive to examine trends and spatial patterns of severe fire in Victoria, Australia. Fire managers and ecologists are increasingly recognising that fire severity is a vital metric to understand, beyond traditional measures of burnt area. I recommend acceptance of this paper subject to additional comment by the authors on the following issues:

Line 221; official fire history records and databases tend to decline markedly in quality the further back in time one looks; the trend in number of fires per year may therefore to some extent be impacted by how well records of fires were kept in the 1980s - can the authors comment on the quality of the dataset in this regard? Has satellite burnt area mapping confirmed this trend in Victoria independently of the fire history database?

General comments; this study focuses specifically on "wildfires"; no specific mention of prescribed/hazard reduction burns is made so I am assuming they are explicitly excluded. While prescribed fires are intended to be of low severity, and usually are, this is not always the case. Can the authors comment on whether they explicitly excluded prescribed fires, and if so, how can they be sure these excluded fires did not have high severity patches that escaped analysis?

Reviewer #3: Tran and coauthors investigate changes in fire severity (fires >1,000 ha) in south-eastern Australia over the last 30 years. They find that fire severity has increased through time, both in absolute terms as well as in terms of the proportion of area burnt. They also investigate several other properties of fire severity such as patch size, number and clustering, as well as regional variation.

Fire severity provides a clear link between fire and its effects on vegetation and ecosystems more broadly, yet outside of the U.S. there are few studies that have examined long term trends in fire severity. As the authors recognise, given widespread interest, evidence of the existence of trends in fire severity also fills an important gap in our understanding of wildfire and climate change (commentary on the existence of such trends in the absence of evidence notwithstanding).

While I commend the authors for tackling this subject I have serious concerns about the methods they use to measure fire severity. Their use of fixed thresholds with spectral indices is not ideal for sensitive detection because it does not take into account local conditions (soil type, drought etc). As far as I can tell they have not calibrated their severity measurement in this way. This is doubly important because the whole point of looking at changes over time is separating the signal (severity changes) from the noise (eg changes due to climate or based on local differences). Thus there is concern that their method omits key elements of both spatial and temporal variation.

Despite these concerns, the authors’ validation metrics appear reasonable, suggesting their work nevertheless captures some properties of fire severity. However, the performance appears systematically worse than other methods now available (Gibson et al. 2020, Collins et al. 2018, 2020). Thus there are both theoretical and practical reasons for preferring an alternative approach.

A lesser but also important issue is that their fire severity mapping technique is based on work outlined in conference proceedings. Although the proceedings are listed in journal citation databases, I don’t think it is appropriate that a foundational piece of this study comes from there. The method deserves proper scrutiny and I don’t have confidence that the conference proceedings provide that, nor is it reasonable for reviewers to consider this conference proceeding in addition to the manuscript itself.

I have some other more minor comments on the manuscript but do not feel it appropriate to raise them in light of these more substantial issues.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Grant James Williamson

Reviewer #3: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

15 Oct 2020

[PONE-D-20-11188] - (Revision 1) - Reply to Reviewers

Title: High-severity wildfires in temperate Australian forests have increased in extent and aggregation in recent decades

Thank you very much for the invitation to revise our manuscript. We are encouraged that the Reviewers appreciated the work and thank them for their constructive comments that have led to significant improvements in the manuscript.

We have made several changes in response to the Reviewer’s comments as well as several minor changes to further clarify our approach, noting that the paper’s key findings remain unchanged. We trust that the changes as detailed below fully address the Reviewers’ comments, and that the paper will now be accepted for publication.

Please note references to line numbers in the reviewers’ comments refer to our original submission, whereas line numbers in our responses refer to the revised version (with track changes). Underlined text in responses indicates new text.

Reviewer # 1

Our thanks to Reviewer 1 for their insightful comments. We have addressed Reviewer 1’s comments as follows:

1) Comment 1: While I believe that studies of high severity fires are very important and highly relevant, I cannot see the novelty of this study, specifically, the difference between this study, which the authors claim is ‘for the first time’ (L 287 and L 311), and fire severity maps of [64]? From Table 1 and the Methods it seems that the temporal changes of fire severity have already been mapped on the State level prior to this work and those are of high quality based on in situ and high-resolution image confirmation.

Response:

The text detailing the methods was not sufficiently clear, which has led to this confusion: As indicated in the text, the wildfire history dataset available from the Victorian Department of Environment, Land, Water & Planning (‘DELWP’; [64]) contains the spatial extent of the wildfires since 1926. Fire severity mapping in Victoria has not been conducted consistently, with only some fires being assessed for severity in the period between 1998 and 2014. This fire severity information is contained in a spatial layer that we did not cite in the text but that has now been included ([78]) To conduct this study, we had to generate the severity mapping of the selected 162 fires, and we did it by implementing a random forest classification model. The model was trained with the severity data available in the spatial layer provided by the government. Therefore, one of the novelties of this study was the generation of severity maps for all the wildfires larger than 1000 ha that occurred between 1987 and 2017 and for which there were satellite images available, which has indeed been done here for the first time.

Change: (L170-173) “Severity of the wildfires in Victoria has not been consistently recorded, with historic fire severity mapping only available for nine years in the period between 1998 and 2014[78]. To generate fire severity maps for the 162 selected wildfires ensuring the consistency of the classification we used Fire severity was mapped using a Random “

([78]: Department of Environment, Land, Water & Planning - DELWP. Aggregated Fire Severity Classes from 1998 onward. Melbourne, Victoria, Australia: Department of Environment Land Water and Planning; 2017. Available from https://discover.data.vic.gov.au/dataset/aggregated-fire-severity-classes-from-1998-onward).

Table 1 might have contributed to this confusion as the caption did not clearly indicate that the values were relative to the 162 wildfires, and that the data in the final column were from this study (not the State data).

Change: (L136-140) To avoid such confusion, we have modified the caption. “Table 1. Characteristics of the bioregions in the study area affected by the selected 162 fires […] Number of wildfires included in this study (i.e. 162 wildfires greater than 1000 ha, occurred between 1987 and 2017 and with available Landsat imagery) and their cumulative total [64] and high-severity burnt area (as estimated in this study).”

2) Comment 2: The section on fire severity mapping (L 164) is not clear. Because [64] contains the total and high severity burnt areas (Table 1), and years of fires (the Methods), why authors repeated the analysis using Landsat images? Yet analysed changes in the number of fires and fire size using the data from [64], Fig 2? In the Methods section the authors say they excluded 43 fires and only 162 fires were analysed (L 129), while Fig 2 is based on 211 fires from [64], L 227. What analysis is based on [64] and on images processed in this study?.

Response:

As above, we agree that the text was not sufficiently clear. From the fire history dataset, we identified a total of 211 wildfires between 1987 and 2017 that met the criteria of being >1000ha. Those were used to analyze changes in the number and total extent of the fires in the studied period. Of those 211, we generated severity maps for the 162 for which there were pre and post fire Landsat images available and cloud free. Thus, fire severity analysis was conducted only for those 162.

Change: We have modified the text to clarify when each dataset was used. (L125-134) (Methods): “That amounted to 211 wildfires that were used to assess changes in the number of fires per year and mean fire size between 1987 and 2017. Each fire was classified according to its dominant bioregion [53]. For the purpose of assessing changes in fire severity, 32 of the 211 wildfires were discarded because pre- or post-fire remote sensing images were unavailable, and 11 were discarded because clouds covered more than 25% of the fire affected area, which may affect the spatial metrics assessed in our study. In total, a subset of 162 wildfires, with at least two fires per year over the past three decades, was used to generate fire-severity maps and analyse changes in severity patterns.”

(L239-240) (Results): “Based on the fire history dataset (n=211), the number of wildfires per year larger than 1000 ha between 1987 and 2017 increased significantly (P= 0.012)”

3) Comment 3: There is also unclear terminology, e.g. what is the ‘patch’? I understand it reflects high severity burnt areas, but only early in the text it says so (L 86?) and no definition is given when the patch analysis is described.

Response:

We agree that a clear definition of the term ‘patch’ is missing. The term is used to refer to areas burnt by high-severity fire surrounded by a different severity within the wildfire perimeter. We have added this definition in the description of the spatial configuration metrics.

Change: (L200-202) ”Spatial configuration metrics were calculated at the patch level, i.e. areas of high-severity fire surrounded by different severities within the wildfire boundary.”

4) Comment 4: Please clarify what is the difference between ‘fire area’ and ‘total fire size’ L 209-212? From Fig 2 it looks like unit of ‘fire size’ is ha, and by reading further it seems that the total fire size includes all severity classes, L 222, but it can be only guessed if ‘fire area’ relates to high severity areas? Then how does fire area relate to patch? Is it an aggregation of patches?

Fig 4, because ‘patch’ is not clearly defined, it’s not clear if the analysis relates to the burnt area, high severity area or..?.

Response:

We agree that the use of both terms ‘fire area’ and ‘total fire size’ is confusing. We have revised the text to consistently use the term ‘fire size’, which has been defined as the total wildfire area (ha). In contrast, we use the qualifier ‘high-severity” to clearly indicate when we refer exclusively to the area burnt at high-severity.

Changes: (L206-209) (Methods): “Edge density is the ratio between the total length (m) of the edges of the high-severity patches and the fire size (i.e. total wildfire area burnt at any severity; ha)”

(L225-231) (Methods):“Predictor variables included year and fire size (i.e. total wildfire area, ha) as fixed effects and bioregion as a random effect, which was only included in the state-wide mixed effects models. Fire size was included as covariate in all models as it can be”

Figure 4 shows the spatial metrics, which always refer to the patches of high-severity fire. For simplicity and readability, we do not include the qualificative ‘high-severity” in the axes labels. However, we will include a clarification in the caption

Change: (L277-283) “Fig 4. Changes in high-severity spatial metrics over time. Each subplot displays a scatterplot between the Year of the fire and the defined high-severity spatial metric. Dots represent each of the 162 wildfires. Values are the results for single mixed effects models where Year and Fire size are fixed effects and Bioregion is a random effect. Lines represent significant (solid black) or not significant (dashed grey) linear relationships.”

5) Comment 5: In the discussion, can it be that the changes in patch shape complexity relate to the image quality as the data from earlier years would come from less sophisticated satellite images such as Landsat 5 rather than indicate increasing severity of fires?

Response:

The Landsat Program represents the world's longest continuously-acquired collection of space-based moderate-resolution land remote sensing data and thus it provides essential land change data and trending information not otherwise available. The program has been designed to ensure its capability to track changes overtime is preserved. To that end the technical prescriptions (e.g. spectral bands, bandwidths, spatial resolution) of its sensors have remained consistent through the different Landsat missions. That consistency has made the Landsat time-series one of the most widely used to monitor land surface changes overtime.

In accordance, we have no reason to suspect that the changes observed in the high-severity spatial patterns could be related to the characteristics of the sensors.

In addition to that, all images were obtained from the Landsat Collection 1 Tier 1, which according to the USGS have the highest available data quality and are considered suitable for time-series analysis. Tier 1 includes Level-1 Precision and Terrain corrected data that have well-characterized radiometry and are inter-calibrated across the different Landsat instruments. The georegistration of Tier 1 scenes is consistent and within prescribed image-to-image tolerances of ≦ 12-meter radial root mean square error (RMSE) (https://www.usgs.gov/core-science-systems/nli/landsat/landsat-collection-1) .

Change: (L145-146) “Wildfire severity of the selected 162 fires was mapped using Landsat TM, ETM+ and Landsat 8 imagery (30 m spatial resolution, all from Landsat Collection 1, Tier 1)”.

6) Comment 6: L 329, the reference [86] from California’s conifer forests is not highly relevant to the forest resilience statement regarding fire adapted Eucalyptus of Australian forests. Are there conifer forests in the state of Victoria, subjected to high severity burns?

Response:

We intended to show that increases in high-severity fire are concerning because of their impact on resilience worldwide but also agree that a reference related to the impact of fire severity on Victorian temperate forest would be relevant. We have modified the text to clarify our message.

Change: (L356-359): “Our quantified increases in high-severity burned area can lead to concerns about the resilience of Victoria’s temperate forests [20, 93], similar to those expressed for other forest types elsewhere [4, 92, 95].”

7) Comment 7: In conclusion, studies of fire severity changes over time are highly important and relevant yet a clear separation of novelty of this work vs already conducted state level analysis is required.

Response:

As indicated before, state fire history dataset only contained the extent of the wildfires. The generation of fire severity maps and the analysis of the high-severity fire metrics is all an original work developed in this study.

Reviewer # 2

Our thanks to Reviewer 2 for their positive assessment that the paper ‘is a very well written’, ‘it is rigorous and timely paper that leverages the increasing ease of analysing the Landsat satellite archive to examine trends and spatial patterns of severe fire in Victoria, Australia’, and ‘Fire managers and ecologists are increasingly recognising that fire severity is a vital metric to understand, beyond traditional measures of burnt area’. We have addressed Reviewer 2’s comments as follows:

8) Comment 1: Line 221; official fire history records and databases tend to decline markedly in quality the further back in time one looks; the trend in number of fires per year may therefore to some extent be impacted by how well records of fires were kept in the 1980s - can the authors comment on the quality of the dataset in this regard? Has satellite burnt area mapping confirmed this trend in Victoria independently of the fire history database?

Response:

We thank the referee for the useful comment. We agree that the accuracy of the records has changed over time, something that may be particularly true for small fires that were unaccounted for. However, we believe this may not have impacted the quality of our dataset as we focused on wildfires larger than a 1000ha. Furthermore, the trend was still consistent when reducing the study period to remove the initial years where the records could have been more uncertain. Unfortunately, no satellite burnt area mapping has been conducted to confirm this trend independently.

9) Comment 2: General comments; this study focuses specifically on "wildfires"; no specific mention of prescribed/hazard reduction burns is made so I am assuming they are explicitly excluded. While prescribed fires are intended to be of low severity, and usually are, this is not always the case. Can the authors comment on whether they explicitly excluded prescribed fires, and if so, how can they be sure these excluded fires did not have high severity patches that escaped analysis?

Response:

Our study focused only on “wildfires” and thus we explicitly excluded prescribed fires from the fire history data available from the Victorian Department of Environment, Land, Water & Planning of Victoria state at the beginning of the wildfire selection. It was out of the scope of this study to investigate any changes in the severity of the prescribed fires, most of which did not meet the minimum 1000ha criteria.

Reviewer # 3

Our thanks to Reviewer 3 for their positive assessment that the paper’s ‘evidence of the existence of trends in fire severity also fills an important gap in our understanding of wildfire and climate change’ and ‘examined long term trends in fire severity’. We have addressed Reviewer 3’s comments as follows:

1) Comment 1: Fire severity provides a clear link between fire and its effects on vegetation and ecosystems more broadly, yet outside of the U.S. there are few studies that have examined long term trends in fire severity. As the authors recognise, given widespread interest, evidence of the existence of trends in fire severity also fills an important gap in our understanding of wildfire and climate change (commentary on the existence of such trends in the absence of evidence notwithstanding).

While I commend the authors for tackling this subject I have serious concerns about the methods they use to measure fire severity. Their use of fixed thresholds with spectral indices is not ideal for sensitive detection because it does not take into account local conditions (soil type, drought etc). As far as I can tell they have not calibrated their severity measurement in this way. This is doubly important because the whole point of looking at changes over time is separating the signal (severity changes) from the noise (eg changes due to climate or based on local differences). Thus there is concern that their method omits key elements of both spatial and temporal variation.

Response:

We apologize for the lack of clarity in the description of the methods implemented on the fire severity mapping. Fire severity classification was not conducted based on fixed threshold with spectral indices, as understood by Reviewer 3. Instead, and as indicated in the ‘Fire severity mapping’ section, we used a Random Forest (RF) classification model trained with 2238 reference plots (60% of the entire reference dataset) from eight large wildfires plots between 1998 and 2009 that covered all forest types encompassed in this study. A total of 12 predictor variables were included in the RF classification model: the identified four best performing spectral indices for the studied forest types (dNBR, dNDVI, dNDWI, dMSAVI) and their pre- and post- fire values; The RF algorithm was validated on an independent set of 1492 reference plots (40% of the reference dataset), yielding a high classification accuracy for the high-severity fire, with a commission error (plots wrongly attributed to high severity) of 0.06 and an omission error (high severity plots incorrectly classified) of 0.18. RF classification models are known to outperform single thresholding and are being increasingly and more widely implemented.

On the other hand, as explained in the methods section, we limited the influence of spatial variation by using a change detection approach, where severity is classified at the pixel level based on the differences between pre-and post- fire signal. The influence of temporal variation was also reduced by selecting pre- and post-fire images that were not more than 3 months apart, thus minimizing the impact of phenology and atmospheric conditions.

Change: (L170-182) For clarification, we have added details to the description of the fire severity classification. “Fire severity was mapped using a Random Forest model based on spectral indices that had been previously trained and validated by the authors for the same study area [61]. The reference fire-severity dataset used for training and validation was comprised of 3730 plots from eight large wildfires (>5,000 ha) that occurred between 1998 and 2009 and covered 13 forest types differing in species composition, canopy cover, canopy height and regeneration strategy. These forest types match those affected by the 162 wildfires of this study. Fire severity of the 3730 reference plots had been assessed in situ or visually interpreted on very high resolution orthophotos by the Department of Environment, Water & Planning (DELWP). Severity was classified as […]”

2) Comment 2: Despite these concerns, the authors’ validation metrics appear reasonable, suggesting their work nevertheless captures some properties of fire severity. However, the performance appears systematically worse than other methods now available (Gibson et al. 2020, Collins et al. 2018, 2020). Thus, there are both theoretical and practical reasons for preferring an alternative approach.

Response:

As clarified above, in this study we have used a Random Forest classification model to map fire severity, which is the same method implemented by Gibson et al. 2020, Collins et al. 2018, 2020. We trust that therefore there is no further concerns regarding the severity mapping conducted in this study.

3) Comment 3: A lesser but also important issue is that their fire severity mapping technique is based on work outlined in conference proceedings. Although the proceedings are listed in journal citation databases, I don’t think it is appropriate that a foundational piece of this study comes from there. The method deserves proper scrutiny and I don’t have confidence that the conference proceedings provide that, nor is it reasonable for reviewers to consider this conference proceeding in addition to the manuscript itself.

Response:

We agree that the methods used to conduct a study should be sufficiently explained in the text of the manuscript so that there is no need for the readers to access further publications. That is why the ‘Fire severity mapping’ section includes a detailed explanation of the development, validation and accuracy of the random forest model used in this study. Further details have been added to the text, thus in our opinion the information now provided is sufficient for the reader to understand the method and gauge the robustness of the approach.

The conference paper, which was published by SPIE remote sensing after a review process, compared the classification accuracy between Random Forest model and the SI threshold approach, which confirmed that Random Forest models outperform thresholding.

4) Comment 4: I have some other more minor comments on the manuscript but do not feel it appropriate to raise them in light of these more substantial issues.

Response:

We will be glad to address the minor comments in any subsequent revision

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

4 Nov 2020

High-severity wildfires in temperate Australian forests have increased in extent and aggregation in recent decades

PONE-D-20-11188R1

Dear Dr. TRAN,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Krishna Prasad Vadrevu, Ph.D

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: Yes: Grant James Williamson

9 Nov 2020

PONE-D-20-11188R1

High-severity wildfires in temperate Australian forests have increased in extent and aggregation in recent decades

Dear Dr. TRAN:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr Krishna Prasad Vadrevu

Academic Editor

PLOS ONE