International Journal of Wildland Fire

Fig. 1. Workflow for biome-scale fire severity mapping framework in northern Australia. The process integrates MODIS surface reflectance, MODIS-derived burnt area mapping, active fire data, field data and classification and validation steps to generate annual fire severity maps and a confusion matrix.

Fig. 2. North Australia depicted as four main zones, from north to south: the High Rainfall, Low Rainfall, Southern Savannas and Australian Mainland. Also delineated are the three north Australian pyro-geographic regions defining the Early and Late Dry Season thresholds: (R1) Kimberley/Top End West (1 January to 30 June); (R2) Top End East/West Queensland (to 30 July); and (R3) Northeast Queensland (to 31 August) delineated by the lines of longitude at 132°E and 142°E, respectively. The validation waypoints collected from 2011 to 2016 are included.

Fig. 4. Scatterplots illustrating the Fire Severity class values of post-fire NIR vs RdNIR (relativised difference of pre and post NIR) derived from the supervised Random Forest classification in the Early Dry Season (EDS) on the left, and the Late Dry Season (LDS) on the right, 2016. Although there appears to be considerable overlap, there are distinct individual values of Severe and Mild reflectance supported by the validation.

🔥New in IJWF:
Edwards et al. present a biome-scale fire severity mapping framework for Australia’s tropical savannas using MODIS data and 6478 field sites, achieving 93% accuracy. A major step for biodiversity, emissions, and Indigenous-led fire programs.
🔗 doi.org/10.1071/WF25044

#IJWildlandFire

🔥 Call for Papers: Wildland–Urban Interface Fires 🌳🏠

Submission deadline extended to 1 March 2026.
We welcome research on fire dynamics, risk modelling, exposure & impacts, mitigation, recovery, and community resilience.

🔗: connectsci.au/wf/pages/cal...
#FireScience #WUI #IJWildlandFire

Graphic showing the three layers of relationships investigated in this paper, from top to bottom: synoptic weather patterns – surface fire weather – vegetation fires. Mean sea level pressure (MSLP) anomalies and Canadian Fire Weather Index (FWI) values are for June 26, 2018, the day the Saddleworth Moor Fire in England was declared a major incident, which remains one of the largest fires experienced in the UK at 18 km2 ( Graham et al. 2020). The vegetation fire layer shows all spring (blue) and summer (orange) vegetation fires >1 ha recorded between April 2009 and April 2020.

Summary of fire data from the incident recording system (IRS) database. (a) Total number of fires (orange line) and total burned area (hectares; pale orange bars) of crop, grassland and heathland/moorland vegetation fires in England, 2009–2020. (b) Total number of fires across the entire data period (2009–2020) in each land cover category during spring (March–May) and summer (June–August). (c) Total number of fires across the entire data period (2009–2020), on each day of the year. Dashed line represents the division between spring and summer (31st May).

Ranked percentile curve score (intercept of Theil–Sen regression ± 95% confidence interval) for surface weather metrics on fire days. Weather variables obtained from E-OBS (‘raw’ weather metrics) are MnT = mean daily temperature; MxT = maximum daily temperature; GR = global radiation; RH = relative humidity. Weather indices obtained from the Canadian Fire Weather Index System (CFWIS) are FWI = fire weather index; FFMC = fine fuel moisture code; DMC = duff moisture code; DC = drought code; ISI = initial spread index; BUI = build-up index. Higher values indicate that the variable performs better at predicting fire occurrence.

🔥New in IJWF:

Little et al. reveal that persistent high-pressure systems drive spring vegetation fires in England, while summer fires are less weather-dependent. Forecasting fire risk? Look to both surface and synoptic signals.

🔗 doi.org/10.1071/WF25158

#IJWildlandFire

Fig. 1. Characteristic velocities associated to the forces governing wildfire behaviour on a sloping terrain: wind inertia (parallel to the ground), buoyancy (vertical) having two components that are parallel and normal to the direction of fire propagation. Ue: effective wind speed, Uw: prevailing wind speed, UB: buoyancy characteristicvelocity.

🔥New in IJWF:
Accary et al. argue that fire-induced wind—often ignored in models—plays a key role in extreme fire behaviour. They call for new experiments and simulations to quantify its feedback and improve predictions.
🔗 doi.org/10.1071/WF25258

#IJWildlandFire

Initial Attack Assessment index (IAA) for Californian fires retrieved by IRWIN from 2020 to 2023 (n = 26,907) considering the initial attack success (left: success (fire size <4 ha); right: fail (fire size >4 ha). Wildfires were independently simulated using WFA to obtain their corresponding IAA. The actual reported fire size is represented by graduated circles.

Schematic modeling process that includes the three main research objectives related with the wildfire data subsets employed and their corresponding univariate logistic regression models.

(a) Fire simulations (n = 360) were automatically conducted for the period of 6–9 January 2025. The colored dots represent the IAA category assigned to each simulation. (b) The number of incidents per IAA level, along with the IA success rate (%), and the names of escaped (>4 ha) and large wildfires are shown across IAA classes (1–5). Fire names are followed by their respective burned area in hectares.

🔥New in IJWF:

Cardil et al. present an Initial Attack Assessment (IAA, 1–5) to flag fires likely to escape initial suppression. Analysis of 26,907 California ignitions shows higher IAA(especially terrain and fire-behavior)lead lower initial-attack success.

🔗 doi.org/10.1071/WF24160
#IJWildlandFire

Predicted wildfire susceptibility maps generated by the Graph Convolutional Network (GCN) using climate data of the different Shared Socioeconomic Pathways (SSPs) for the year 2060. The vegetation is unchanged. The overall geographical distributions of susceptibility are consistent across all three scenarios. An area of focus is analyzed in depth to understand local effects of climate predictions (hashed area).

Time series of climate variables average over the northern region across different Shared Socioeconomic Pathways (SSPs) scenarios. SSP1–2.6 displays a higher average temperature, average minimum temperature, average humidity and precipitation, but a lower average maximum temperature.

Predicted 2060 species suitability distribution maps generated by the MaxEnt using bioclimate data of the different Shared Socioeconomic Pathways (SSPs) for the year 2060. The vegetation adapts to the climate scenarios.

🔥 New in IJWF:

Ren et al. model future wildfire susceptibility in Portugal with climate and vegetation change. Results show climate-only projections miss local shifts, while climate-driven eucalyptus expansion can increase risk even under low-emission.

🔗 doi.org/10.1071/WF25092
#IJWildlandFire

Example of an identified fire event (a), and corresponding burned (b, c), and unburned (d–f) areas used in the calculation of burn severity. (a) Landsat 5 post-fire mean image composite collected June–September 2001, displayed with a false color composite using shortwave infrared 2, near-infrared and red bands to highlight fire effects. Burned area polygons for the fire, delineated by (b) the Landsat Burned Area (LBA) product, and (c) Monitoring Trends in Burn Severity (MTBS) datasets. Examples of unburned areas, commonly used to calculate burn severity offset values, are shown in black and include (d) a manually delineated unburned polygon of similar vegetation composition to the fire event, as well as an automated ring buffer (180 m outer, 0 m inner) surrounding (e) LBA, and (f) MTBS fire perimeters.

Performance of models relating field-collected Composite Burn Index (CBI) and satellite-derived burn severity measures. We contrast model RMSE from (a) dNBR (differenced Normalized Burn Ratio), and (b) RdNBR (Relativized dNBR) values with and without offset correction across spectral indices, image selection method and offset type used. Offsets generated manually (white) and from ring buffers surrounding Monitoring Trends in Burn Severity (MTBS; black) and Landsat Burned Area (LBA; gray) fire perimeter datasets are evaluated.

The relationship between field-collected and satellite-derived burn severity data for the bestperforming automatically generated buffer offset correction methodology. (a) Observed Composite Burn Index (CBI) burn severity and associated differenced Normalized Burn Ratio (dNBR) values calculated from mean annual image composites, offset with the unburned dNBR from a 100 m buffered ring surrounding Landsat Burned Area (LBA) fire perimeters. The red dashed line indicates the fitted exponential relationship between CBI and dNBRoffset, with associated root mean squared error (RMSE) shown. (b) Exponential model-predicted dNBR compared with associated observed dNBRoffset values, with red dashed 1:1 identity line and associated R2 value overlaid.

Mean dNBR (differenced Normalized Burn Ratio) offset values, dNBRunburned, across tested image selection and offset delineation methodologies for 141 fire events. We compare the influence of inner and outer ring buffer sizes derived from Monitoring Trends in Burn Severity (MTBS) and the Landsat Burned Area product (LBA) fire perimeters. The outer buffer indicates the maximum distance from the fire perimeter and the inner buffer defines the minimum distance from the fire perimeter, in between which are the pixels used to calculate the offset value.

🔥New in IJWF:

Menick et al. test phenological offset corrections for Landsat dNBR/RdNBR across CONUS CBI plots, showing when offsets improve burn-severity–CBI relationships and when automated buffers bias severity low.

🔗 doi.org/10.1071/WF25066
#IJWildlandFire

Flowchart showing how each TreeMap2016 forest stand was assigned a hazard level based on simulated fire behavior and effects outputs from the Fire and Fuels Extension to the Forest Vegetation Simulator.

Simulated treatment pattern on a synthetic landscape comprised of stands from TreeMap2016. Each treatment unit is ~49 ha and the total treated area is about 22% of the total area.

🔥 New in IJWF:

Johnston et al. assess four fuel treatments using the Avoided Wildfire Emissions framework, showing that underburning and thinning + underburning meaningfully reduce future wildfire emissions, especially where annual fire probability is high.

🔗doi.org/10.1071/WF25026
#IJWildlandFire

Composite burn index map for fire perimeters analysed in this study (a); inset map of the fire location (b); classified composite burn severity classes across the study area and severity for large patches are presented in insets (i), (ii) and (iii).

Partial dependence plots of each variable with the respective relative contributions to the model in parentheses. The dashed lines represent the partial dependence values, while the solid line shows a fitted curve with loess smoothing applied. The density distribution of the sampled test data is shown at the bottom of each plot for continuous variables and alongside predicted response for a categorical variable. The red dots represent the predicted response for each vegetation type. CBI: Composite Burn Index.

Bar plot showing the mean forest cover (%) for different forest age classes in Quebec. Analysis is based on forest cover data from four studies ( Hansen et al. 2013; Sexton et al. 2013; Matasci et al. 2018; Feng et al. 2022) and age class data from Maltman et al. (2023). The error bars represent the standard deviation (s.d.) of the mean cover for each age class.

🔥New in IJWF:

Mackey et al.

The 2023 Québec fires burned 4.5 M ha of boreal forest. Using Sentinel-2 CBI mapping the authors show burn severity peaks on dry topographic positions, under extreme fire weather, and in 20–40-year forests.

🔗 doi.org/10.1071/WF24175
#IJWildlandFire

(a) Post-treatment forest structure and two people sampling fuels from the 0 to 1.7 m and 2.8 to 4.5 m plots within the clustered design. (b) A plot placed between two whiskers to mark the distance intervals and the strings stretched across the PVC sampling frame to denote the randomly generated 20 × 20 cm subplot for collecting the plot’s observation. Black markings on the PVC sampling frame are at 20 cm intervals to create the grid of 36 subplots. (c) Two people collecting a dead fine fuel sample into a pre-labeled and pre-weighed polyethylene resealable bag.

The observation period was organized into three phases: early summer (Julian day 138–189), mid-summer (Julian day 192–236) and late summer (Julian day 243–287) for the following plots. The dashed lines on the axes depicting FMC indicate the commonly used 30% moisture of extinction threshold for dead understory forest fuels ( Rothermel et al. 1986). (a) Empirical cumulative density functions of the three distributions (early summer, mid-summer, late summer) of observed fuel moisture content. (b) Kernel density plots of the three intraseasonal period distributions. (c) Boxplots colored by intraseasonal period depicting the distribution of FMC on each observation day. The box indicates the inter-quartile range (25th to 75th percentiles). The mid-band indicates the median, and the whiskers indicate points within 1.5 times the interquartile range. Points outside the whiskers are outliers. Note: to visualize the data more effectively, the axes depicting FMC (%) were visually constrained to 150%, preserving all values while focusing on a more relevant range.

The smoothed, marginal effects of understory cover, canopy cover, heat load index and precipitation on FMC transformed to show the fitted GAM function on the response scale. The x axis shows values of the covariate, the rug indicates the distribution of covariate observations, and the y axis shows expected FMC values. The gray bands correspond to the 95% confidence interval and represent model uncertainty in the transformed (response scale) estimate. Though negative values are not possible in FMC or under a Gamma distribution, the confidence interval below 0 in the precipitation plot is an artifact of transformation and reflects wide uncertainty.

🔥New in IJWF:

Ohlson et al. map fine dead fuel moisture across a Colorado mixed-conifer forest using 1-h fuel samples from 80 plots. They reveal strong fine-scale spatiotemporal variability shaped by canopy, understory, and aspect.

🔗 doi.org/10.1071/WF25...

#IJWildlandFire

Fire frequency between 1989 and 2021 and land use and land cover (LULC) in 2021 from the Carajás National Forest, Eastern Amazon, Pará, Brazil. Forest areas are not delineated; i.e. they correspond to areas outside the anthropised and canga areas. The embedded pie chart shows the accumulated fire scars per LULC class; note that profound changes in LULC during the observation period ( Fig. 1) were considered for construction of the chart.

Time series data of the annual proportion of burned areas (%) within Campo Ferruginosos National Park and Carajás National Forest and possible fire drivers (Amazonian and regional deforestation, annual precipitation and climatic severity) for the period 1989–2021 (a) and correlations between potential fire drivers and the annual proportion of burned areas, separated by area (b). The dashed lines in (a) represent the 5-year trends. The dashed lines in the scatterplots (b) represent the linear relationships between the rainfall, climatic severity, and deforestation trends and the annual proportion of burned areas.

🔥New in IJWF

Sanjuan et al. analyse 33 years of fires in Amazonian canga, showing stark contrasts between long-protected Forest and recently protected Forest. They find land-use history drives fire occurrence while dry-season rainfall controls intensity.

🔗 doi.org/10.1071/WF24...

#IJWildlandFire

Fig. 1. The geographic, environmental and management context of the four study locations in the Australian deserts. (a) The main vegetation types (data from National Vegetation Information System (NVIS) 6.0 downloaded July 2023) and rainfall isohyets (data from Bureau of Meteorology (BoM)). Non-shaded areas represent either minor vegetation types, or non-desert vegetation of the northern savannas. (b) Areas that are owned or managed/comanaged by Indigenous groups; and areas that are managed for conservation as part of the National Reserve System by the government, private organisations, and Indigenous groups. Non-shaded areas on the map are non-Indigenous tenures not managed for conservation. Data on Indigenous ownership and management were obtained from Australian Bureau of Agricultural and Resource Economics and Sciences; and data on protected areas from Collaborative Australian Protected Areas Database (downloaded 12 March 2023). (c) The annual rainfall for the four study locations, over the study period 1997–2019 (data resourced from Scientific Information for Land Owners (SILO), downloaded 18 Jan 2023). Note the high rainfall in 2000–2001 and 2010–2011, especially at Newhaven and Tanami.

Fig. 2. Annual fire extent and fire season. (a) The annual fire extent in the managed and baseline periods.(b) Annual fire extent in relation to variation in rainfall during the preceding 2 years; data for Pinpi, Katjarra and Newhaven are grouped, and shown separately to data from Tanami. (c) The proportion of each year's fire extent that occurred in the hot season, in conditions of low, moderate, and high rainfall in the preceding two years; data for the baseline period and the managed period are shown separately. Data for Pinpi, Katjarra and Newhaven are grouped, and shown separately to data from Tanami. (d) An example of the accumulated cool (blue shading) and hot season (orange shading) fires for the baseline and managed periods is shown for Pinpi.

🔥New in IJWF:
Cliff et al. evaluate a decade of Indigenous-led fire management across 4 Australian desert sites. The shift to cooler-season, smaller burns improved fire heterogeneity and cultural engagement.
🔗 doi.org/10.1071/WF25...

#IJWildlandFire

🔥New in IJWF:

Desservettaz et al. provide a comprehensive review addressing Australian firefighters’ concerns about bushfire smoke. The article outlines health risks, gaps in PPE protection, off-gassing, dermal exposure, and mitigation strategies.

🔗 doi.org/10.1071/WF25138

#IJWildlandFire

1/4 - Check our new paper in @ijwildlandfire.bsky.social

We use MOEA optimisation to support decisions for #wildfire risk reduction (e.g., where should we use prescribed burning to reduce risk?). Optimisation improves fuel treatment plan effectiveness by 80–280% and adapts to changing objectives.

Top-left plot shows the performance of Pareto-optimal fuel treatment plans (black points) for the BP-area objective that aims to reduce burn probability (BP) to the entire Adelaide Hills region. Light grey points represent non-Pareto fuel treatment plans that were considered as part of the optimisation process. Other plots show selected fuel treatment plans matching a range of different levels of area treated (AT), as indicated by crosses on the Pareto front. The selected solutions show that the specific treatment blocks chosen in a plan can change significantly for different levels of AT. Comparisons are made between plans with an AT of 1.01 or 1.07% and 4.91 or 5.17%. Areas that are unique between the plans at the two levels of comparison are shown in colour and areas chosen by both plans are shown in black. Reduction in BP is calculated using the baseline BP as modelled with the metamodel.

🔥New in IJWF:
Radford et al. introduce a simulation-optimisation framework using neural network metamodels and NSGA-II to create fuel treatment plans that reduce burn probability by up to 284%, this method balances risk reduction and resource use.

🔗 doi.org/10.1071/WF25080

#IJWildlandFire

🔥New in IJWF:

Desservettaz et al. review the complex composition and health risks of bushfire smoke for firefighters, offering evidence-based guidance on exposure reduction, PPE use, and decontamination.

🔗 doi.org/10.1071/WF25138

#IJWildlandFire

In short: connectsci.au brings a new look and features, but with the same high standards and publication ethics.
Browse International Journal of Wildland Fire:
connectsci.au/wf

Powered by @csiropublishing.bsky.social , connectsci.au prioritises accessibility, discoverability & functionality, incl:
- better search filtering across article types&subjects
- journal article split screen view
- nuanced email notification options, inc. saved search alerts: connectsci.au/sign-in

Today @csiropublishing.bsky.social launched ConnectSci, a new global destination for trusted science content, hosting our journal, eBooks and a science news service.
You can now find International Journal of Wildland Fire here:
connectsci.au/wf
So, what's new for readers and authors?

Fig. 1. Geometric structure of symmetric canyon (a); experimental set-up (b); and schematic diagram (c).

🔥New in IJWF:

Fan et al. investigate canyon fire dynamics under varied terrains, revealing critical slope thresholds (α ≥ 27.5°, δ ≥ 20°) for eruptive fire. Strong convective heating ahead of the fire front drives rapid spread, challenging strategies.

🔗 doi.org/10.1071/WF24...
#IJWildlandFire

The most common barrier identified in each dispatch zone (top); and the most common barrier constructed through human landscape modification (bottom).

🔥Most Read

Epstein & Seielstad analyse WFDSS text from 6,630 large US wildfires (2011–2023). Barriers appear in 75%—mainly roads, burn scars and fuel variation. Prior fires more often stopped spread than treatments.

🔗 doi.org/10.1071/WF25051
#IJWildlandFire

Fuel treatment sequence and categorization. In a two-stem process, large number of specific treatments were categorized and combined based on their sequence, and then further combined based on the dominant characteristic of the modification of fuels.

🔥Most Read

Fallon et al. present a novel methodology to assess fuel treatment effectiveness in California forests. Using FTEM and FACTS data, they show 61% of treatments modified fire behavior, with fire or removal-based treatments most effective.

🔗 doi.org/10.1071/WF24220
#IJWildlandFire

Wildfire risk assessment in Sichuan Province, China: hazard modeling approach considering different combinations of classification criteria and connection values of factor attributes Background Current wildfire risk research has primarily focused on hazard assessment, lacking a comprehensive framework that integrates vulnerability and adaptive capacity. Moreover, the influence of ...

🔥Most Read

Wagner et al. highlight elevated PTSD, depression, anxiety, and sleep problems in wildland firefighters, but baseline risks unclear. They urge clearer separation of general vs fire-specific stress, focus on everyday exposures, long-term impacts.

🔗 doi.org/10.1071/WF24159
#IJWildlandFire

Wildfire hazard maps generated using different connection methods coupled with the logistic regression (LR) algorithm. Subfigures (a–f) are based on wildfire samples as classification criteria of factor attributes, and subfigures (g–l) are based on the whole area (PS: probability statistics; FR: frequncy ratio; IV: information value; CF: certainty factor; WOE: weights of evidence; EBF: evidential belief function).

🔥 New in IJWF:

Yue et al. present a comprehensive wildfire risk framework for Sichuan, China, integrating hazard and vulnerability. Using six statistical connection methods with logistic regression, they identify the Point-IV-LR model as most effective.

🔗 doi.org/10.1071/WF25089
#IJWildlandFire

🏠🔥New in IJWF:

Gjedrem et al. (2025) examine fire risk perception and garden adaptation in Tasmania’s WUI. Residents often underestimate hazards, but personalised garden hazard reports motivated change despite knowledge, resource, and emotional barriers.

🔗 doi.org/10.1071/WF24213
#IJWildlandFire

Brief:

Wildland-urban interface fires pose a challenge, driven by climate change, expanding settlements, and evolving fire regimes. Recent fires in LA, highlight the need to advance scientific understanding, policy, and adaptive management strategies to mitigate risks and enhance resilience.

🔥Call for Papers: Wildland-Urban Interface Fires Special Collection in IJWF!🌳🏠🌳

Share research on fire dynamics, risk modelling, resilience & more. If you have any questions, reach out to the team.

📆 Submit by 31 Dec 2025

More Information: www.publish.csiro.au/wf/content/C...

#FireScience

🌿 New in IJWF:

Krix et al. (2025) developed impact models for the Australian Fire Danger Rating System to predict structure loss from wildfires. Using structure density, cleared land, canopy height and terrain ruggedness.

🔗 doi.org/10.1071/WF24148
#IJWildlandFire

🌳New in IJWF
Wong et al. show that heat yields in wet sclerophyll fuels vary widely by species and season—especially in live understorey fuels. Fixed values in fire models may overestimate fire intensity.

📄 doi.org/10.1071/WF24227
#IJWildlandFire

🌿 New in IJWF:

O’Grady et al. shows how Machine Learning with Landsat can reconstruct fire histories across US military lands. Models achieved >93% accuracy, offering local-scale insights into ignition patterns & fire management for defense landscapes.

🔗 doi.org/10.1071/WF24214
#IJWildlandFire