The availability of high-resolution, geo-referenced digital data underscores the need for advanced tools to model wildfire dynamics. A critical task is transforming these vast datasets into actionable insights for stakeholders. Recent advancements in computational science, particularly in the development of innovative algorithms, are essential for understanding and addressing wildfire behaviour and vulnerability.
This session aims to bring together experts from geosciences, climatology, forestry and territorial planning to enhance our understanding of these critical fire-related dynamics and to explore innovative strategies for mitigation and resilience. By fostering interdisciplinary collaboration, we seek to advance the science of wildfire prediction, prevention, and post-fire recovery, ultimately contributing to more effective responses to the growing threat posed by wildfires in a changing climate.
We welcome contributions on topics such as:
• Methodologies for recognizing, modelling, and predicting wildfire spatio-temporal patterns.
• Pre- and post-fire assessments, including fire mapping, severity evaluations, and risk management.
• Long-term analysis of wildfire trends in relation to climate change and land use changes.
• Fire spread modelling and studies on fire-weather relationships.
• Post-fire vegetation recovery and phenology.
Join us in advancing the study of wildfires and developing strategies to mitigate their impact.
Western North America (WNA) has emerged as a global wildfire hotspot. While quasi-stationary atmospheric blocking drives persistent fire-favorable conditions, synoptic recurrent Rossby wave packets (RRWPs) represent a critical but underexplored driver of wildfire extremes. This gap is deepened by an apparent paradox that synoptic-scale circulation is projected to weaken under climate change while extreme wildfires intensify. Here we jointly analyze transient RRWPs and quasi-stationary blocking to classify extreme wildfire events in WNA. We then assess how these changing circulation patterns translate into fire risk using a novel wildfire-triggering efficiency framework powered by machine learning. Our results show that RRWPs contribute to wildfire extremes at magnitudes comparable to blocking, together explaining nearly two-thirds of events. Blocking shows only weak changes and RRWPs clearly weaken in WNA, but their wildfire-triggering efficiency is strongly enhanced by thermodynamic amplification. Under SSP5–8.5, blocking-related extreme wildfires increase by 45.9% and RRWP-related events by 37.1% by 2100. These findings establish a more complete picture of circulation controls on wildfires and identify thermodynamics as the primary driver of increasing wildfire risk in a warming future.
How to cite: He, C., Chen, H., and Liu, Y.: Weakened circulation yet stronger wildfires in Western North America, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3663, https://doi.org/10.5194/egusphere-egu26-3663, 2026.
The wildfire regime in South-Central Chile (SCC, 30º to 44ºS) has changed in recent decades due to changes in land use, climate conditions, and characteristics of weather extreme events. While during 1976-2016, the mean annual burned area was ~54,000 ha, during the last decade a sequence of seasons multiplied that value, in particular 2016-2017 with 570,000 ha and 2022-2023 with 450,000 ha. To the north of this region, the fire regime is fuel-limited (e.g. amount and connectivity of biomass), while to the south, it is primarily climate-limited (i.e. plenty of wet fuels). In contrast to all other Mediterranean regions worldwide, SCC has a very low rate of natural ignitions (<1% of wildfires), whereas 99% of fires are caused by humans. In SCC, large-scale plantations of flammable exotic species and invasive trees and shrubs have modified the fuel structure particularly since the mid-1970s, leading to an increase in fire risk. Within this context and beyond climate variability, in this work we unveil crucial aspects on the relationship between wildfires and weather variability in SCC.
As a first task, we identify weather patterns associated with relatively large wildfires (>520 ha, N~800) within 7 SCC sub-regions, previously delimited according to climate, topography, and land use. Using historical wildfire records (including start date, duration, and burned area) from the National Forestry Corporation (CONAF) spanning 1984-2025, we describe the mean local 15-days evolution of weather conditions centered on the start dates of wildfires. For this, we use daily ERA5 data, including maximum temperature, minimum specific humidity, mean sea-level pressure, and maximum surface wind intensity. Furthermore, within each subregion, a cluster analysis reveals distinct mean weather sequences and typical thresholds for these variables related to wildfires. While subtle weather variability is detected in the northern part of SCC, for the southern part of SCC our analysis reveals the relevance of mid-latitud synoptic variability–in particular blocking patterns induced by migratory anticyclones–, as well as associated mesoscale phenomena, especially coastal lows and foehn wind systems. Moreover, prominent differences in wildfire characteristics are found between distinct extreme weather events, such as heat waves and single hot days.
As a second task, we explore the intra-seasonal evolution leading to selected weather patterns associated with wildfires in SCC. We find groups of events that reveal different sequences of significant mid-latitude and tropical circulation anomalies up to 14 days before the wildfire start dates. For each group, we show that the corresponding weather-fire relationship is in fact mediated by a distinct trajectory of the Fire Weather Index (FWI). Finally, we suggest a scheme based on the Madden-Julian Oscillation (MJO) index and the Standardized Extra-Tropical Index (sETI) to monitor intra-seasonal atmospheric teleconnections favoring weather fire in SCC.
How to cite: Jacques-Coper, M., Ruiz, N., Suazo-Alvarez, M., Segura, C., Mendiburo, C., Pérez, M., González-Reyes, A., de la Barrera, F., and Holz, A.: Wildfires and Weather Variability in South-Central Chile, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15094, https://doi.org/10.5194/egusphere-egu26-15094, 2026.
The Canadian Fire Weather Index (CFWI) system is a wildfire risk evaluation tool used in several countries. This index estimates fire intensity based on meteorological variables. We use the CFWI framework to investigate how global warming will impact the risk of forest fires in Canada. We calculate the CFWI indices in equilibrium 5000-year integrations of the Canadian Earth system model (CanESM5) with different prescribed atmospheric CO2 levels (pre-industrial to 4x pre-industrial). We find that higher atmospheric CO2 levels lead to higher fire weather index (FWI) values and longer fire seasons across Canada. The yearly maximum FWI values also tend to increase with CO2, suggesting that global warming will raise the risk of extreme wildfire. The FWI increase is mainly driven by temperature: higher CO2 levels and temperatures lead to more efficient and sustained drying periods, resulting in more flammable, drier fuel for forest fires. However, more CO2 in the atmosphere also leads to more precipitation, higher relative humidity, and slower wind speeds, resulting in regional differences in the response of CFWI to changes in CO2. We further conduct a regional analysis of fire indices to examine how global warming will impact Canada at the provincial level. This model-based information will be useful to evaluate the risk of wildfire across Canada in the future, and a similar analysis could be applied in other world regions.
How to cite: Garduno, D., Weaver, A., Whaley, C., Abraham, C., and Netherton, S.: The effect of global warming on forest fires in Canada, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15326, https://doi.org/10.5194/egusphere-egu26-15326, 2026.
In 2022, southwestern France experienced an exceptional fire season, with a burned area 14 times higher than the 2006–2023 average. Here, we assess how unusual were the fire weather conditions observed during wildfires of different sizes and how anthropogenic climate change (ACC) has altered —and will further alter— the probability of fire-weather conditions associated with the top-3 largest fires in 2022 (Landiras-1: 12,552 ha; Landiras-2: 7,124 ha; La Teste-de-Buch: 5,709 ha).
To do so, we used the daily Fire Weather Index (FWI) computed from the SAFRAN reanalysis (Système d’Analyse Fournissant des Renseignements Atmosphériques à la Neige) —cross-validated with ERA5— and a nationwide fire record dataset (BDIFF, Base de Données sur les Incendies de Forêts en France: 2006–2023). Using the generalized extreme value (GEV) theory, we then quantified the rarity of FWI conditions associated with the top-3 largest fires across different spatiotemporal scales. Our results show that the rarity of those conditions generally increases with the resolution, with return periods increasing from ~6 to ~34 years, from ~22 to ~89 years and from ~6 to ~101 years when moving from the coarser to the finest spatiotemporal scale for Landiras-1, Landiras-2 and La Teste-de-Buch fires, respectively. Finally, we used four GCMs (IPSL-CM6A-LR, CanESM5, MIROC6 and NorESM2-LM) from the CMIP6 DAMIP and ScenarioMIP experiments to examine how ACC has made those FWI conditions more or less probable from 1950–2100. By 2022, ACC had at least doubled the likelihood of those FWI conditions, and will make them, by the end of the century (under SSP2-4.5), at least 10–100 times more probable, depending on the models. Our study underlines the growing influence of ACC in the risk of extreme fires in France across a range of scales.
How to cite: Zhu, S., Barbero, R., Pimont, F., and Renard, B.: Quantifying the current and future likelihood of the 2022 extreme wildfires weather conditions in France with anthropogenic climatechange, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5436, https://doi.org/10.5194/egusphere-egu26-5436, 2026.
Terrestrial near-surface wind speed research (NSWS, at 10 m above ground) has largely focused on the global-scale stilling phenomenon observed over recent decades. However, much less attention has been paid to whether this phenomenon is also present in the wind regimes associated with wildfire activity. In this context, the recently reported reversal of near-surface wind trends towards increasing wind speeds introduces additional uncertainty regarding the potential impacts of wind on global wildfire regimes.
In this study, we assess the ability of commonly used reanalysis products, such as ERA5 and CERRA, to represent observed wind variability at weather stations across the Iberian Peninsula for 1984-2021. According to our results, most reanalyses fail to reproduce the trends and multidecadal variability of NSWS observed at more than 700 weather stations. In contrast, the use of a high-resolution (3-km) NSWS dataset produced using a U-Net based on partial convolutions, trained to reconstruct the wind field from station-based wind observations, better captures these temporal trends and variability. We then analyse the wind changes observed during wildfire events in Spain over recent decades, examining their relationship with large-scale climate oscillation modes. Finally, we explore whether observed trends in wildfire-related winds are consistent with the stilling–reversal framework.
How to cite: Plaza-Martin, N. P., Alba-Manrique, À., Plésiat, É., Azorín-Molina, C., and g. Pausas, J.: Wind variability influencing wildfires in Spain, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12564, https://doi.org/10.5194/egusphere-egu26-12564, 2026.
Wildfires are natural phenomena affecting ecosystems and causing negative impacts on human health and biodiversity. In the Mediterranean region, wildfire regimes are strongly influenced by local climatic conditions, leading to pronounced inter and intra-annual variability in wildfire occurrence.
Owing this link, the study explores for the first time the climatic drivers influencing the monthly burned area (BA) during the summer fire season (May - September) in northern Italy at the three scales of spatial resolution: 0.11, 0.25 and 0.50 degrees. We then build multi-regression data-driven models to define the main BA predictors for the investigated area. The summer monthly percentages of burned area at the three resolution for the 2008-2022 period were derived from the GPS-based BA perimeters. A total of 150 daily precipitation and maximum and minimum ground station series were collected, converted at monthly scale, reconstructed, homogenised and spatialised at 0.11°, 0.25° and 0.50° resolution using the Universal Kriging with auxiliary variables. Several climatic indices were subsequently computed for precipitation, temperature and drought. To identify the best BA predictors, we first performed the Pearson’s correlation test, for each pixel, between the monthly BA series and the climatic indices calculated for three different aggregation periods: concurrent summer (2008-2022), 6 months before the fires (winter 2007-2021) and 12 months before the fires (summer 2007-2021). Multilinear regressions models were computed using every possible combination of the best predictors. The best regression models were selected through an out-of-sample procedure, and the model performance was tested by comparing the predicted BA with the observed data, estimating explained variance and correlation. Finally based on the CORINE Land Cover map, the vegetation classes that were most susceptible to wildfires, and their typical elevation ranges, were identified.
This study shows that summer fires in northern Italy are concentrated in July and August and are predominantly located in the southern part of the study area, at elevations between 100 and 600 m a.s.l. In particular, the lower rates of the Ligurian and Tuscan Apennines exhibit a fire return period of 1 to 2 years, in contrast to the Alps, where it exceeds 6 years. Sclerophyllous, Sparse, and Open Forests appear to be the vegetation classes most susceptible to fire in these fire-prone regions. Modelling results for the 2008–2022 period indicate that the most accurate predictions were performed at 0.11° of resolution and fires are driven by drought conditions caused by water stress than by high temperatures. Indeed, the most significant predictors of burned area were the two drought indices and water balance, recorded both for the current period (June to July) and for the preceding 6 months period (December to March).
How to cite: Baronetti, A., Fiorucci, P., and Provenzale, A.: Defining climatic drivers for the prediction of summer wildfires in northern Italy, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-529, https://doi.org/10.5194/egusphere-egu26-529, 2026.
Over the past years, boreal forests of Canada have been increasingly affected by large and high-severity wildfires, with recent fire seasons recording unprecedented burned areas across the country. Alongside these extreme wildfires, the ecosystems have been forced to recover under frequent climate extreme events, including prolonged droughts and intense heatwaves, which have often occurred compounded. In this study, we propose a preliminary framework to analyse the association between meteorological conditions and their impact on post-fire recovery over three major eco-regions of Canada - Western Canada, the Great Plains, and Eastern Canada. Considering the period 2001-2025, we first estimate the post-fire vegetation recovery rates using a mono-parametric model based on the remotely-sensed Enhanced Vegetation Index (EVI). Then, we apply a Random Forest (RF) modelling approach that integrates SHAPely Additive exPlanations (SHAP), aiming to explain how seasonal meteorological variables, which include air temperature, precipitation, snow depth, and solar radiation, influence the forest recovery process.
Among the three eco-regions, the recovery model exhibits a consistently strong performance. Forests in Western Canada generally show faster post-fire recovery, contrasting with slower recovery rates observed in the Great Plains, although considerable intra-regional contrasts are found. The RF models and the associated SHAP-based results effectively identify key meteorological drivers of burned forest recovery, showing an overall good performance across the three regions. The model tends to give higher importance to variables that strongly control the growing season in boreal ecosystems, namely solar radiation and air temperature during transitional seasons, particularly in spring. In Western Canada, solar radiation and air temperature roughly constitute the most influential features on recovery, whereas in the Great Plains and Eastern Canada, autumn precipitation emerges as the primary controlling feature. Additionally, both precipitation and air temperature extremes in winter and summer frequently appear as secondary drivers of recovery rate, highlighting that climate extreme events may display an important modulating effect on post-fire recovery.
Our preliminary framework provides a novel approach to estimate the recovery rate of burned vegetation across Canada based on a time-series analysis, rather than space-for-time substitution methods. Furthermore, the application of machine-learning techniques combined with SHAP provides new insights related to seasonal and regional roles of meteorological variables in modulating post-fire vegetation recovery processes.
This work was performed under the framework of DHEFEUS project, funded by Portuguese Fundação para a Ciência e a Tecnologia (FCT) (https://doi.org/10.54499/2022.09185.PTDC).
How to cite: Ermitão, T., Russo, A., Bastos, A., and Gouveia, C.: Impact of Meteorological Conditions on Post-fire Recovery of Boreal Forests across Canada, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14044, https://doi.org/10.5194/egusphere-egu26-14044, 2026.
Mediterranean ecosystems are increasingly exposed to frequent and high-severity wildfires, driven by rising temperatures, prolonged droughts, and land-use change, making wildfire one of the dominant disturbance agents shaping forest structure and function. There is a concern that frequent and high-severity wildfires may threaten the resilience of forests, even in fire-prone forest ecosystems, and their ability to recover to pre-fire levels. This has implications for carbon storage, biodiversity conservation, water regulation, and the long-term provision of ecosystem services on which both local communities and broader society depend. The availability of long-term multispectral satellite time series has demonstrated the ability to estimate the instantaneous impact of fires on forests and the recovery trajectories. Yet, spectral recovery is two-dimensional and does not necessarily mean functional, structural or compositional recovery which may be slower than simply tracking the greenness index trajectories. GEDI lidar metric display a larger variety of fire responses that spectral metrics but are only available since 2019. This study combines structural GEDI metrics with a Landsat-based historical forest disturbance to estimate the structural recovery of forests post fire in Greece from the 1985. Overall, we find post-fire vegetation recovery in Greece, using GEDI biomass, height, canopy cover, and foliage height density, likely takes 50 or more years. Low-intensity and small spatial scale fires recover within the first 20-30 years, while high-intensity and large fires show forest recovery likely >50 years. There is also some evidence of a lack of recovery trajectory or a new ecosystem state within the first 40 years for some regions. This work demonstrates how integrating lidar with long-term spectral archives can provide regional scale post-fire structural recovery assessments, can provide critical information to constrain terrestrial biosphere models predicting fire impacts and forest recovery, and can begin providing more targeted data locally to regionally for fire management, restoration practices and climate mitigation.
How to cite: Antonarakis, A.: Long-term Structural Recovery of Wildfire-affected Forests in Greece using GEDI and Landsat, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21174, https://doi.org/10.5194/egusphere-egu26-21174, 2026.
The prevailing paradigm is that recovery of post-fire soil-hydrologic properties is dominated by pedogenic processes that drive soil structure formation, a critical process for regaining soil hydrologic functioning. The primary drivers for soil structure formation are climate and vegetation required for soil biological activity. Evidence shows that following perturbations of agricultural soils (e.g., compaction) or abrupt land use change soil structure recovery may take years to decades. In contrast, measurements from post-fire soils show that recovery of critical soil hydrologic properties (notably soil saturated hydraulic conductivity) is rapid (generally within 1 to 3 yrs) and occur at rates faster than expected from soil structure regeneration. To reconcile the rapid post-fire recovery rates, we propose a new conceptual framework for recovery of post-fire soil-hydrologic properties driven primarily by accelerated erosion of the unstable and structureless pyrolyzed surface soil layer. In this framework, initial recovery occurs not by redeveloping new structure in the pyrolyzed surface soil layer, but rather by removing it, thus exposing minimally-affected sublayers as new soil surfaces. Based on wildfire characteristics, a typical depth of pyrolyzed soil layer is estimated to be a few centimeters (<5 cm) depending on fuel load, burning times and heat transport. A tentative peak temperature of 300 C (torrefaction limit) defines the extent of loss of binding organic carbon thus creating a fragile and easily transported layer by wind or water erosion. Examples of the proposed mechanism in several Western US post-fire landscapes will be presented with discussion of various landscape geomorphic controls (topography, post-fire rainfall, ash transport and more).
How to cite: Or, D. and W. McCoy, S.: Enhanced erosion of pyrolyzed soil surfaces drives rapid recovery of post-fire landscape hydrologic functions, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4293, https://doi.org/10.5194/egusphere-egu26-4293, 2026.
The projected increase in lightning-ignited wildfires (LIWs) during the 21st century highlights the need to improve our understanding of the mechanisms and processes governing these natural fires. However, results from large-scale LIW studies are often limited by uncertainty in identifying the specific lightning discharge responsible for each ignition. Here, we present a simple and flexible classification system that ranks LIWs according to the level of confidence in the lightning event causing the ignition.
We first used a probabilistic index to identify the most likely lightning event igniting each wildfire. This index was combined with a set of filters based on eight criteria, including holdover time (the time between lightning-induced ignition and fire detection) and the distance between the reported lightning location and the fire ignition point, to define four confidence classes. The lowest-confidence class applied no filters and retained all lightning events selected by the probability index (one per fire). The remaining three classes applied increasingly strict filters, yielding progressively higher confidence levels. This classification framework was applied to LIWs from four study regions: Switzerland, Catalonia (Spain), California and Nevada (United States), and the whole continental United States. In addition, two LIWs with ignitions documented by video footage were used for validation.
Relative to the unfiltered class, intermediate confidence classes retain approximately one-quarter to two-thirds of lightning discharges, whereas the highest-confidence class retains only 5-20%. This reflects a trade-off between sample size and confidence. The proposed confidence classification provides an initial framework that can be further refined, and offers a way to increase the robustness of LIW analyses, thereby supporting improved investigations of the factors controlling lightning-induced wildfire ignitions.
How to cite: Moris, J. V., Pérez-Invernón, F. J., Camino-Faillace, P. A., Gordillo-Vázquez, F. J., Pineda, N., Pezzatti, G. B., Conedera, M., Zhu, Y., Lapierre, J., Hunt, H. G. P., and Veraverbeke, S.: Defining confidence classes for lightning discharges igniting wildfires, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6513, https://doi.org/10.5194/egusphere-egu26-6513, 2026.
Wildland fire smoke transport is governed by a complex interplay between fire heat release, atmospheric boundary layer (ABL) turbulence, synoptic forcing, and terrain. Despite substantial advances in coupled fire–atmosphere modeling, the role of the ambient, evolving ABL state in controlling plume rise and transport under realistic fire conditions remains insufficiently resolved, largely due to the extreme computational demands of event-scale large-eddy simulation (LES). This study addresses this gap by conducting a high-resolution LES of the atmospheric boundary layer over complex terrain during the Mosquito Wildland Fire (California, September 2022), followed by a plume simulation whose forcing is constrained by satellite observations.
We perform a multi-domain Weather Research and Forecasting (WRF-LES) simulation spanning 24 hours (08–09 September 2022) over the Sierra Nevada, capturing the diurnal evolution of boundary-layer depth, turbulence intensity, wind shear, and regime transitions under realistic synoptic and topographic forcing. The ABL simulation is validated against four ASOS surface stations and NOAA Twin Otter airborne observations, demonstrating accurate reproduction of near-surface thermodynamics and vertical wind shear. The results reveal pronounced transitions from convective to shear–buoyancy-driven regimes, strong inversion-layer shear, terrain-modulated low-level jets, and vertically coherent turbulent structures extending several kilometers above the surface.
Using the resolved ABL state at noon local time, we then simulate the release of a buoyant plume for one hour using an active-scalar LES formulation. The plume is represented as an idealized, steady circular heat source at the ground, with surface heat flux prescribed to match satellite-derived fire radiative power (FRP) from MODIS. This approach isolates the influence of the ambient ABL on plume evolution while maintaining physically realistic forcing. Independent evaluation against MISR stereo plume-height retrievals shows strong consistency between simulated and observed plume-top heights (~3–4 km), vertical gradients, wind shear, and downstream transport pathways. Importantly, MISR plume heights reflect time-integrated plume evolution over several hours of advection, allowing meaningful comparison with the short-duration LES plume simulation.
The results demonstrate that plume rise, vertical penetration, and horizontal transport are primarily controlled by the evolving ABL structure—specifically boundary-layer depth, inversion-layer shear, turbulent kinetic energy distribution, and terrain-induced flow modulation—once the fire heat release is constrained to realistic values. Sensitivity analysis shows that while plume source size and buoyancy magnitude influence near-source behavior, ABL regime and shear dominate plume fate at kilometer scales.
This study provides one of the first event-scale demonstrations that resolving the real atmospheric boundary layer under complex terrain is a prerequisite for physically meaningful wildfire plume simulation. By combining validated ABL LES with satellite-constrained plume forcing, the work establishes a robust foundation for future fully coupled fire–atmosphere modeling and advances understanding of two-way ABL–buoyancy interactions in wildfire environments
How to cite: Bhaganaagar, K.: Using Large-eddy-simulation at event-scale to evaluate the ABL-widllandfire-plume interactions of Mosquito Wildland Fire, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4562, https://doi.org/10.5194/egusphere-egu26-4562, 2026.
Mediterranean ecosystems face escalating wildfire challenges as climate change intensifies extreme temperature conditions across Southern Europe, making fire danger zone identification increasingly critical for ecosystem management. This research develops a satellite-based modeling framework integrating spatial analysis techniques to comprehensively map fire danger zones across Sicily's Messina province. The study focuses on this fire-prone region where the convergence of fuel availability, multiple ignition sources, and extreme environmental conditions create favorable scenarios for wildfire events. This methodology employed European Forest Fire Information System data spanning the period 2012-2024 (excluding 2015 due to data unavailability) to analyze wildfire patterns across Messina's 326,689 hectares. The research implemented a six-step analytical framework: temporal binary coding for fire occurrence pattern identification, multi-layer spatial union of administrative and burned boundaries, raster conversion with cumulative summation, integrated forest type mapping, coordinate reference system standardization, and comprehensive vegetation-based area calculations. This methodological approach achieved high spatial accuracy while ensuring analytical consistency across heterogeneous landscape types. Results reveal substantial wildfire impact across the study region, with 30,654 hectares affected representing 9.38% of Messina's total area. Fire frequency analysis demonstrated a significant increasing trend, growing from 64 events in 2012 to 382 events in 2023. Spatial analysis identified 1,470 distinct fire events distributed throughout the provincial area. Vegetation impact analysis revealed differential vulnerability patterns, with agricultural lands most affected (34.84% of burned area), followed by Mediterranean maquis (25.88%) and oak forests (19.98%). Mountain pine forests exhibited the highest reburn vulnerability (35.32%), while beech forests demonstrated complete resistance to repeated burning. The modeling approach has so far successfully identified fire danger zones and vulnerability patterns across Messina's diverse ecosystem types, providing valuable data for targeted fire prevention strategies and ecosystem restoration priorities. This research contributes important insights to fire danger zone mapping and establishes a methodology applicable to similar wildfire-prone region across Southern Europe.
Key words: Fire danger zones, Spatial modeling, Mediterranean ecosystems, Burn frequency, Vegetation vulnerability
How to cite: Adewuyi, A. B. and Barbati, A.: Identification and mapping fire danger zones using modeling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-196, https://doi.org/10.5194/egusphere-egu26-196, 2026.
Recent wildfire extremes in northern Canada indicate a shift in lightning-driven ignition processes beyond episodic variability. This study examines the atmospheric conditions responsible for the increasing occurrence of dry lightning—cloud-to-ground lightning accompanied by negligible precipitation—across Yukon, the Northwest Territories, and Nunavut. By integrating cloud-to-ground lightning observations with ERA5 reanalysis, we identify a dominant thermodynamic configuration controlling dry-lightning frequency. Dry lightning increases most strongly when anomalously warm near-surface temperatures coincide with enhanced mid-tropospheric moisture (700–500 hPa), forming a pronounced vertical contrast. This structure supports deep convective electrification while limiting surface wetting through efficient sub-cloud evaporation. In contrast, conventional instability and wind-based indices exhibit limited explanatory power for long-term dry-lightning variability. The extreme 2023 wildfire season exemplifies this ignition-efficient configuration rather than representing a rare anomaly. Projections from the CMIP6 multi-model ensemble indicate that continued surface warming and increasing mid-tropospheric moisture will shift this thermodynamic state toward the climatological mean under future warming, particularly under high-emissions scenarios. A physically constrained regression framework suggests that dry-lightning occurrence may increase by more than 50% by the late 21st century. These findings demonstrate that northern Canada is transitioning toward a climate state in which lightning-induced wildfire ignitions are structurally favored. Accounting for evolving vertical thermodynamic conditions is therefore essential for anticipating future high-latitude wildfire risk.
Acknowledgement
This work was funded by the Korea Meteorological Administration Research and Development Program under Grant RS-2024-00404042 and the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2024-00343921).
How to cite: Bae, J., Wang, S., Yoon, J., and Son, R.: Dry Lightning and Escalating Wildfire Risk in Northern Canada: The 2023 Extreme Fire Season and Future Projections, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4655, https://doi.org/10.5194/egusphere-egu26-4655, 2026.
Wildfires are an upcoming threat across Central Europe, driven by shifting climate regimes, extended drought periods, and rising temperatures. Effective fire management depends on a solid understanding of fire behavior, which creates a demand for reliable fire growth models. Fire modelling in this region poses several challenges, especially if the models were developed for different environmental regions (e.g. North America). The availability of high-resolution fuel data, fuel models and information on local fuel moisture and wind patterns - all important drivers for fire spread prediction – can cause additional difficulties in predicting fire behavior. Well-documented fire events can provide reliable information for model calibration and validation, but such case studies are scarce in Central Europe.
Therefore, this study investigates the applicability of several fire growth models (Farsite, Prometheus, SimtableTM, PhyFire) for the specific environmental conditions in Central Europe based on a set of pre-defined evaluation criteria. The selected models are applied to two well-documented fire cases to assess their ability in predicting spatial and temporal fire growth under varying environmental conditions in Central Europe. The analysis reveals differences in suitability among the models and underscores the need for region-specific calibration. Furthermore, improved data availability regarding documented fire cases and wind velocity and direction are demanded. These results help to identify the needs for an advanced wildfire growth modelling in Central Europe and supports more informed fire management decisions and training in future.
How to cite: Kuhnen, K., Andrade, M. S., Müller, M., and Vacik, H.: Lessons learnt from the application of various wildfire growth models for the environmental conditions in Central Europe, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9904, https://doi.org/10.5194/egusphere-egu26-9904, 2026.
Wildfire activity in the Iberian Peninsula (581,353 km²) is highly heterogeneous due to strong gradients in climate, topography, vegetation, disturbances, land use, and management. This spatial variability challenges fire modeling, risk assessment, and fuel reduction strategies across contrasting regions. Previous efforts to map fire regimes succesfully used clustering of historical or remote-sensed fire data. However, the resulting zones were often large and spatially fragmented, rendering them challenging to integrate into landscape scale stochastic wildfire modeling.
To address this, we delineated pyroregions, defined as spatial units with generally homogeneous fire regime conditions, to support subsequent fire weather characterization and the definition of modeling domains for stochastic wildfire simulations. Our objective was to generate contiguous spatial units that exhibit both similar historical fire incidence and consistent fire-weather and topographic characteristics. To achieve this, we populated subwatersheds (obtained from HydroBASINS; n = 4,409; mean area 13,391 ha) with contemporary fire regime descriptors derived from burned area and ignition records –sourced from national (AGIF for Portugal and EGIF for Spain) and European (EFFIS) databases– complemented with fuel moisture (Camprubí et al., 2022; 10.5281/zenodo.6784663) and weather data (ERA5-Land reanalysis data). Descriptors included annual ignition density, annual and summer burned area, wind direction distributions, and fuel moisture content for live woody and fine fuels in the period 2001-2024. Pyroregions were obtained via spatially constrained agglomerative clustering with Ward linkage, enforcing contiguity using a Queen connectivity matrix, which ensured that merges occurred only between adjacent subwatersheds. Following a two-step aggregation scheme, we first delineated 16 broad pyroregions representing major wildfire-regime zones and then partitioned them into 78 similarly sized subareas (pyromes; mean area 7,570 km²) for modeling applications. Finally, boundaries were refined to reduce sharp transitions associated with subwatershed geometry and to produce smoother contours. The resulting map captured transboundary similarities and contrasts in fire regimes and revealed clear structure, including altitudinal gradients and a marked Atlantic to Mediterranean contrast, with large contiguous regions over the inner mesetas and major depressions, and a near continuous coastal belt.
How to cite: Rodrigues, M., Mulavizada, F., Alcasena, F., Molina, J. R., Lamelas, T., and de la Riva, J.: Delineating Iberian pyroregions using agglomerative clustering of fire regime descriptors, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7489, https://doi.org/10.5194/egusphere-egu26-7489, 2026.
Accurately predicting wildfire behavior at geographical-to-regional scales using coupled atmosphere-fire models has the potential to enhance the operational activities of Météo-France, which provides fire danger assessment in support of the French civil protection services. Current fire danger indices primarily rely on meteorological variables and do not include an explicit representation of biomass fuels, despite the fact that extreme wildfire events often result from the combined effect of atmospheric conditions and fuel state.
In this study, we investigate how to integrate a detailed representation of surface fuels into the coupled Meso-NH/BLAZE modeling system (Lac et al., 2018; Costes et al., 2021), by taking advantage of high-resolution vegetation modeling from the SURFEX land surface system (Masson et al., 2013) and by defining fuel models for the vegetation types of the ECOCLIMAP database (Faroux et al. 2013). This study focuses on the French Mediterranean area for two main reasons: i) this is a wildfire-prone area that has experienced intense fire activity in recent years and that is projected to face increased fire danger due to climate change in the next decades (Fargeon et al. 2020); and ii) it has been monitored for several decades by the ONF (French forest services) through a dense observational network, providing extensive measurements of Live Fuel Moisture Content (LFMC).
We implement the Rothermel heterogeneous rate-of-spread (ROS) formulation (Andrews, 2018) in the coupled atmosphere-fire model associated with dynamic fuel models (Scott and Burgan, 2005), in order to represent both dead and live components of the biomass fuels, and to dynamically transfer the herbaceous fuel load from live to dead components as a function of the LFMC to reproduce seasonal curing. We thus analyze the added value of including a live component of biomass fuels and the role of the LFMC in the ROS predictions. Preliminary results indicate that accounting for the live fuel component part of fuels generally reduces the simulated ROS, as higher live fuel content tends to inhibit combustion. Moreover, simulations using dynamic fuel models propagate less extensively than non-dynamic fuel models.
Beyond the explicit modeling of fire-fuel interactions, we also examine the Fire Weather Indices (FWI) based on the Canadian approach (Van Wagner et al., 1985) and adopted by Météo-France to assess meteorological fire danger. By analyzing their relationship with simulated ROS, we aim to establish a first quantitative link between fire danger indicators and physically-based fire behavior predictions.
How to cite: Peyrot, M., Le Moigne, P., and Rochoux, M.: Advancing surface fuel representation for operational wildfire spread modeling at Météo-France, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7424, https://doi.org/10.5194/egusphere-egu26-7424, 2026.
How to cite: Ambroz, M. and Mikula, K.: Forest fire propagation modelling by evolving curves on topography incorporating data assimilation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20807, https://doi.org/10.5194/egusphere-egu26-20807, 2026.
Accurate fuel model mapping is essential for supporting the prediction of forest fire ignition and fire propagation. Although standardized fuel model classifications are widely applied in fire sciences, their performance is often limited when evaluated against field observations, largely due to the high variability within the different fuel categories. Especially in Central Europe there are less experiences with the application of different fuel model classification due to the lack of experiences in the predicting fire behavior under the specific environmental conditions and the lower number of larger fire events. This study addresses these needs by proposing a validation framework to ensure that fuel models assigned to a certain forest patch or landscape allow to represent real-world fire behavior.
To develop the fuel model map for this study, experts combined field measurements on fuel loads with the results of the interpretation of aerial imagery to classify fuels, assigning classes for each 10x10m pixel according to the Scott and Burgan (2005) fuel models based on their interpretation. The proposed validation framework of the fuel model map for this study integrates observed field data from forest fires and prescribed burns in the past to estimate selected fire behavior parameters, such as flame length and rate of spread (ROS). These field observations serve as a ground truth to evaluate the accuracy of a developed customized fuel map using expert-based knowledge. Additionally, we simulate fire behavior with the BehavePlus package for the expert-assigned fuel models, to determine if the simulated parameters match the observed field data, thereby validating whether the fuel model assigned to a given area is both appropriate and provides physically realistic fire behavior. Furthermore, we utilize the Rothermel R package, which implements the mathematical equations of the Rothermel (1972) fire spread model, to reverse-analyze field data and identify the most probable fuel model for a given condition. In a next step, we compare the fuel models suggested by the algorithmic with the fuel models assigned by the expert judgments and the fire behavior parameters derived from BehavePlus.
The results of this study show that customized fuel models based on expert knowledge outperform standardized fuel classifications in representing real-world fire behavior. Reverse fitting of field data using the Rothermel’s model is likely to show differences between algorithmically derived parameters and expert-assigned fuel models, particularly in complex and heterogeneous landscapes. Overall, the integration of field observations with expert-based fuel modeling is expected to reduce uncertainty in fire behavior simulations by: i) comparing simulated fire behavior parameters to field observations; and ii) using the Rothermel R package to validate expert-assigned fuel models, diagnose mismatches and refine fuel assignments.
How to cite: Silva Andrade, M., Kuhnen, K., M. Müller, M., and Vacik, H.: Validating expert-based fuel model by field observations and simulations , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20218, https://doi.org/10.5194/egusphere-egu26-20218, 2026.
Wildfires initiate a hazard chain that significantly alters landscapes and geohydrological processes. In addition to the extensively documented effects on vegetation, soil erosion, and debris flows, steep and rocky terrains may experience delayed yet persistent slope instabilities. However, post-wildfire hazard assessment frameworks still predominantly use the soil burn severity indicators for any type of mass wasting processes, while the response of rock masses and their contribution to post-fire hazards remain underrepresented.
This study addresses this gap by proposing an integrated, rapid assessment approach to evaluate post-wildfire rock slope instability. The motivation of this study is the necessity of cost-effective and timely tools that support emergency response and short- to medium-term risk management in mountainous Mediterranean environments where infrastructure, settlements, and transportation corridors are exposed to post-fire hazards.
The proposed methodology combines Sentinel-2 Level-2A multispectral imagery with field-based observations. Burn severity was mapped using the differenced Normalized Burn Ratio (dNBR), and field surveys were conducted to validate spectral classifications and to identify fire-induced rock degradation indicators. In contrast to conventional soil burn severity observations, special attention was given to rock-specific responses. The rock burn severity indicators were semi-quantitatively evaluated and integrated within a GIS-based framework to identify potential slope sectors with increased rockfall susceptibility.
Results show that wildfire-induced thermal alteration can significantly weaken carbonate rock surfaces and discontinuities without necessarily leading to rapid slope failures. Wildfire functions as a conditioning mechanism that elevates the susceptibility of rock slopes to subsequent triggers, including rainfall infiltration, runoff concentration, and solar radiation cycles.
The study emphasizes the importance of incorporating rock-specific burn severity indicators into post-wildfire rock slope stability assessments. Such an approach supports more comprehensive risk inventories and improves prioritization of mitigation and monitoring strategies. The findings contribute to ongoing efforts to integrate field observations and remote sensing.
How to cite: Kadakci Koca, T.: Assessing Wildfire-Induced Changes in Rock Slopes Using Field Observations and Satellite Data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21604, https://doi.org/10.5194/egusphere-egu26-21604, 2026.
Wildfire dynamics are a highly coupled system depending on not only wildland fuel characteristics but also on topography and weather. Steep terrain features like canyons have been widely reported to produce significant effects on wildfire dynamics, such as sudden fire accelerations. However, these effects are poorly studied and not correctly captured in current models. Furthermore, observational data of wildfire dynamics in steep terrain is extremely scarce. In this work, we will present the study design and preliminary results from a canyon fire field experiment conducted in California (USA) in October 2022. The experiment was set up so that a high-intensity head fire was started and allowed to spread freely up a canyon of approximately 1 km in length and 300 m in elevation difference. The vegetation primarily consisted of chaparral shrubs. Fire dynamics were monitored using airborne multispectral infrared sensors. Vegetation was characterized before and after the burn through airborne lidar scans. Additionally, fire-weather interactions were investigated leveraging Doppler lidar and radar sensors as well as in-situ micrometeorological towers. A fire eruption was observed when the fire entered the canyon, providing evidence of terrain-induced modifications to fire behavior. Datasets like this one are key to study the complex interactions between fire dynamics, vegetation properties, terrain characteristics, and weather dynamics, and constitute an important resource for model development and validation.
Acknowledgements: This work was supported by the U.S. National Science Foundation (NSF) under award number 2053619, the NSF-IUCRC Wildfire Interdisciplinary Research Center, and the EU COST Action NERO (CA22164). The authors also thank the California Department of Forestry and Fire Protection (CAL FIRE) for coordinating the field experiment.
How to cite: Valero Pérez, M. M., Clements, C. B., Klofas, A., Giesige, C. C., Goldbeck-Dimon, E., Manoj Santhi, S., Stockmans, T., Yip, J., Arreola Amaya, M., and Olivera Prieto, P.: Characterizing wildfire dynamics in steep terrain: a canyon fire field experiment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20063, https://doi.org/10.5194/egusphere-egu26-20063, 2026.
Wildfires represent one of the most critical threats to Mediterranean forests, making timely detection and continuous monitoring a priority for risk mitigation and environmental management. Despite significant advances in satellite-based fire monitoring, current approaches remain constrained by a fundamental trade-off between spatial and temporal resolution in available remote sensing data. Geostationary satellite systems offer high-frequency observations that are well suited for near-real-time monitoring, yet their coarse spatial resolution limits their effectiveness for applications requiring fine-scale spatial detail. Addressing this limitation is particularly relevant for wildfire monitoring, where early-stage events often occur at small spatial scales. In this presentation, we introduce a learning-based framework for spatial resolution enhancement of high-temporal infrared satellite observations. The approach explores multiple model families, including autoencoder-based architectures, residual channel attention networks, and generative models such as neural operator diffusion, to reconstruct fine-scale thermal structure from coarse measurements while preserving temporal consistency. The best model configurations are tested in the context of wildfire monitoring, using higher-resolution thermal products from NASA VIIRS as reference data. Results indicate improved representation of fire-related signals, with implications for better early detection and monitoring applications. Detailed methodological developments and quantitative evaluations will be presented in a forthcoming publication.
How to cite: Zenonos, A., Sciare, J., Dovrolis, C., and Ciais, P.: Spatial Resolution Enhancement of Geostationary Thermal Observations for Wildfire Monitoring, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2007, https://doi.org/10.5194/egusphere-egu26-2007, 2026.
As one of the key links in maintaining the balance of ecosystems, natural fires in nature are often extensive and unpredictable. When they get out of control and turn into wildfires, the threats they pose to ecosystems, the atmospheric environment, and human health are incalculable. Fires lead to a continuous reduction in forest coverage, while a large amount of harmful gases produced by forest combustion are emitted into the atmosphere. This causes enormous harm to the ecological environment, economic development, and the safety of human lives and property. Therefore, timely and accurate detection of forest fires, as well as grasping specific characteristics such as the exact occurrence time, location, and spatiotemporal evolution of fires, helps to explore the causes and patterns of fires, and is of great significance for the sustainable management of forests supported by fire prevention management.
This study proposes a novel fire detection algorithm integrating spatiotemporal information, utilizing data from Himawari-8, a next-generation geostationary satellite. By combining contextual information and a dynamic threshold detection method, the algorithm achieves real-time detection and scientific prediction of fire points through improving the slope deviation of infrared channels. A forest fire that occurred in Yuxi City, Yunnan Province, from April 11 to April 15, 2023, was selected as a research case for fire detection analysis. The results demonstrate that the proposed fire point detection method reduces edge false detections compared to WLF, the official fire point product of Himawari-8. Meanwhile, it shows significantly higher recognition accuracy and a notably lower false detection rate than the pre-improved algorithm.
The experimental results show that this improved forest fire detection algorithm can quickly and effectively detect fire point information. Compared with the pre-improved algorithm, it has higher detection accuracy. Meanwhile, the improvement of infrared gradient provides new ideas and methods for realizing effective disaster situation monitoring.
How to cite: Xue, Y.: A Novel Spatiotemporal Fire Detection Algorithm Based on Himawari-8 Satellite Data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2348, https://doi.org/10.5194/egusphere-egu26-2348, 2026.
The characterization of fire behavior from observations and its coupling with plume dynamics and atmospheric composition remains a major challenge for coupled fire–atmosphere modeling systems. In this context, that is the frame work of the EUBURN initiative, this work presents recent developments in the processing and exploitation of MTG-FCI (Meteosat Third Generation - Flexible Combined Imager) observations for the derivation of fire behavior descriptors, and exercise of validation against airborne infrared measurements acquired during the SILEX experimental airborne campaign conducted in southern France in summer 2025.
A dedicated processing framework based on the Fire Event Tracker (FET) algorithm is introduced. FET performs a spatio-temporal clustering of FCI hotspot detections provided by LSA-SAF to delineate individual fire events and derive event-scale fire behavior descriptors, including fire duration, Fire Radiative Energy (FRE), and time series of Fire Radiative Power (FRP), Forward Rate of Spread (FROS), and Fire Line Intensity (FLI). During the SILEX campaign, FET was operated in Near-Real-Time (NRT) and coupled with the ForeFire–MesoNH modeling system through automated now-casting system (FireCast) to simulate plume rise and dispersion, supporting the design of flight plans for the SAFIRE ATR42 research aircraft.
This summer, FET was also made operational over Portugal in collaboration with the Portuguese civil protection authority (ANEPC), with support from the VOST association. In this operational context, FET products mainly consisted of event-scale FRP time series that were used to monitor fire activity and detect reactivation during prolonged fire episodes.
More recently, FET has been extended to a retrospective processing mode, allowing the integration of the complete 2025 LSA-SAF hotspot archive over the Mediterranean basin. This provides a unique dataset of fire behavior descriptors at the scale of fire regime zones, from which initial sub-regional analyses are presented.
To support satellite product validation and provide high-resolution fire behavior characterization, Middle Wave Infrared (MWIR) thermal cameras were operated onboard the ATR42 during SILEX. These airborne observations provide meter-scale snapshots of active fire fronts and their radiative structure, enabling the assessment of sub-pixel fire heterogeneity and radiative variability and serving as a reference for evaluating FCI-derived FRP and their linkage to FET-derived fire perimeters.
In addition, FCI-derived FRE estimates are compared with fuel consumption measurements obtained by INRAE through post-fire field sampling at the Sigean site. This comparison provides an experimental evaluation of the consistency between satellite-based radiative estimates of biomass consumption and ground-based measurements, contributing to efforts to constrain relationships between FRE, fuel properties, and consumed biomass.
Overall, this work supports the development of an integrated fire characterization framework combining satellite and airborne observations, with direct relevance for the validation of coupled fire–atmosphere modeling systems such as ForeFire–MesoNH. By jointly addressing fire behavior, plume development, aerosol emissions, and atmospheric chemistry, the EUBURN project contributes to advancing event-based wildfire representations in next-generation fire–atmosphere and air quality models.
How to cite: Paugam, R., Parent, G., Filippi, J.-B., Benali, A., Gomes, J., Xu, W., Dutra, E., Hofmann, M. P., Ruffault, J., Pimont, F., André, F., Boulanger, D., Retornard, V., Meraner, A., Denjean, C., and Penot, V.: Event-Scale Fire Behaviour Characterization from MTG/FCI Observations and Airborne Observation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14984, https://doi.org/10.5194/egusphere-egu26-14984, 2026.
Forest fires are emerging as an increasingly severe threat to terrestrial ecosystems worldwide, with a reported 246% increase in fire occurrences across the western United States over the past decade. This rapid escalation highlights the urgent need for robust, objective, and scalable forest monitoring approaches capable of detecting fire disturbances in a timely manner. Synthetic aperture radar (SAR), with its all-weather, day-and-night imaging capability, offers significant advantages for operational forest monitoring, particularly in fire-prone regions. In this study, we employ Sentinel-1 C-band SAR data to monitor forest dynamics and map fire-affected areas, with a specific application to the 2025 California forest fires. Sentinel-1 Single Look Complex (SLC) data acquired between 19 June 2024 and 16 June 2025 were processed using the InSAR Scientific Computing Environment (ISCE). The SLC data was used to derive gamma-nought backscatter, alpha angle, and entropy. A statistical change detection framework based on the cumulative sum (CuSUM) method was implemented to identify the timing of fire-induced disturbances. For each pixel, residuals were computed as deviations from the temporal mean, and their cumulative sums were tracked over time. Abrupt shifts exceeding a predefined threshold were interpreted as change events, with the corresponding acquisition dates assigned as pixel-wise change dates. The threshold was adapted to scene-specific characteristics to mitigate false alarms arising from seasonal variability. The algorithm was applied to multitemporal stacks of SAR backscatter, α (alpha) scattering angle, and entropy, producing raster products in which pixel values represent estimated disturbance dates. Validation was conducted using independent vector-based building damage data derived from CALFIRE and compiled by Environmental Systems Research Institute, Inc. (ESRI) for the January 2025 California fires. A comprehensive accuracy assessment was performed by comparing SAR-derived fire-affected areas with the reference data. The results demonstrate that SAR-derived polarimetric parameters provide complementary information for detecting fire disturbances, with VH backscatter yielding the highest agreement (precision: 0.7, F1 score: 0.4) with reference data. Overall, this study presents an efficient and scalable SAR-based framework for near-real-time mapping of forest fire-affected areas, supporting timely disaster response and contributing to sustainable forest management and risk mitigation strategies.
How to cite: Sabir, A. and Khati, U.: Forest fire damage assessment using Sentinel-1 dual-polarimetric SAR data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16366, https://doi.org/10.5194/egusphere-egu26-16366, 2026.
Fire regimes are changing in many parts of the world, with particularly notable shifts in Europe: Regions such as Scandinavia where wildfires historically played a limited role are increasingly experiencing wildfire activity, while parts of Southern Europe face worsening conditions. These developments strengthen the need for integrated information that supports decisions across the full disaster management cycle. OroraTech has developed an end-to-end wildfire product suite that combines satellite observations, numerical modelling, machine learning and AI to support wildfire preparedness, response, and recovery.
Before a fire occurs, the platform focuses on disaster preparedness through medium-range wildfire hazard forecasting up to one week in advance. These forecasts integrate meteorological drivers, fuel characteristics, and historical fire occurrence patterns using data-driven and physics-informed approaches to identify areas of elevated hazard. In addition, scenario-based fire spread simulations allow users to explore potential fire behaviour under varying ignition locations, environmental conditions, and mitigation measures such as fire breaks, enabling proactive planning and evaluation of response strategies.
During an active fire, the system provides operational support. Near real-time active fire detection is delivered via OroraTech’s proprietary thermal infrared satellite constellation, combined with detections from more than 30 additional satellite missions to maximise temporal coverage and robustness. These observations are used to update dynamic fire spread simulations, supporting tactical decisions such as fire break placement and resource allocation. Active fire intelligence is enriched with contextual layers including land cover, topography, and short-term weather forecasts, among others.
After containment, the product suite delivers burned area mapping to support impact assessment, reporting, and recovery planning. Providing consistent pre-, during-, and post-fire products within a single platform enables a continuous and coherent view of wildfire events, supporting stakeholders across the entire wildfire lifecycle.
How to cite: Liesenhoff, L., Wahbe, J., Pörtge, V., Rashkovetsky, D., Bereczky, M., Feuerbacher, K., Würl, K., Langer, M., and Gottfriedsen, J.: The OroraTech Wildfire Solution: Fire Management based on the Forest Satellite Constellation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3365, https://doi.org/10.5194/egusphere-egu26-3365, 2026.
Wildfire danger reflects the interaction of processes acting across a wide range of spatial and temporal scales, from rapid weather-driven variability to slower fuel, hydrological, and climate-mediated controls. This contribution examines how recent advances in artificial intelligence, when combined with structured Earth System Data Cubes, can be used to improve wildfire danger forecasts and also to better understand the mechanisms that drive their variability across scales.
We build on two complementary datacube paradigms: (i) regional, high-resolution daily cubes (e.g., Mesogeos at 1 km × 1 day over the Mediterranean) to resolve local meteorology–fuel–human interactions, and (ii) global sub-seasonal to seasonal cubes (e.g., SeasFire at 0.25° × 8-day, integrating climate, vegetation, oceanic indices, and human factors) to represent large-scale context and teleconnections.
For short lead times, we show that deep learning models that jointly exploit meteorological forcing and surface state information (e.g., vegetation condition and wetness proxies) consistently outperform operational meteorology-only approaches such as the Fire Weather Index. Importantly, explainable AI methods help diagnose which drivers dominate different fire episodes, revealing physically plausible and event-dependent controls rather than fixed empirical relationships. At subseasonal-to-seasonal horizons, predictability increasingly depends on slow-varying land-surface conditions and remote climate signals. Here, we discuss multi-scale learning approaches that fuse local predictors with coarser global fields and climate indices, enabling skillful forecasts of burned-area patterns at multi-month lead times without assuming homogeneous predictability across regions or biomes.
Finally, we argue that improved accuracy alone is insufficient for operational use. We therefore emphasize uncertainty-aware modelling, drawing on Bayesian deep learning to quantify epistemic and aleatoric uncertainties, improve forecast calibration, and support decision-making under risk through interpretable predictions accompanied by explicit confidence information.
How to cite: Papoutsis, I.: AI for wildfire danger forecasting at different spatiotemporal scales, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13976, https://doi.org/10.5194/egusphere-egu26-13976, 2026.
Reliable short-term wildfire forecasting is essential for early warning, timely air-quality management, and mitigating wildfire-related health impacts and economic losses. However, global prediction remains difficult because wildfire occurrence is rare and highly heterogeneous across fire regimes. The Fire Weather Index (FWI) is widely used as a benchmark, but it mainly reflects weather-driven fire danger and does not explicitly represent fuel or fire dynamics, limiting predictive accuracy. Physics-based coupled models can resolve fire–atmosphere interactions, yet they typically require prescribed ignition information and are too computationally expensive for global deployment. Data-driven methods enabled by satellite and reanalysis data offer an efficient alternative. However, many conventional ML approaches treat grid cells as independent samples, which limits learning of neighborhood interactions and multi-day preconditioning. Recent DL studies improve representation learning, but many remain regional and lack unified spatiotemporal dependency modeling. Thus, global spatiotemporal frameworks tailored to the rare and sparse nature of wildfire occurrence remain scarce.
Here we present the STA-Net, a novel global daily wildfire forecasting framework built on a harmonized multi-source dataset spanning 2013–2024. The dataset integrates meteorology, vegetation, lightning, and topography information on a unified 0.5° global grid. Through modeling of spatiotemporal dependencies and imbalance-aware training, The STA-Net learns coherent features that capture multi-day environmental preconditioning and neighborhood-driven fire evolution, which enables accurate next-day wildfire forecasts at the global scale. It also supports short-range forecasts at 1–7 day lead times, although predictive skill decreases progressively as lead time increases.
The STA-Net outperforms the FWI and representative data-driven baseline models, including XGBoost (non-spatiotemporal), LSTM (temporal-only), and 2D-CNN (spatial-only). On an independent global test set, the STA-Net achieves an AUC of 0.97 and maintains stronger discrimination than FWI across all 14 GFED fire regions. Two 2024 case studies in Bolivia and Canada further show that the STA-Net captures the spatial footprint and concentrated high-risk cores of catastrophic outbreaks, supporting event-level generalization beyond aggregate metrics. Using F1 as the primary rare-event metric, the STA-Net achieves the highest score among the data-driven baselines (F1 = 0.65). An ignition–spread–persistent (I–P–S) stratification attributes the largest improvement to spread fire, where neighborhood propagation is central, providing direct evidence for the effectiveness of the STA-Net’s spatiotemporal modeling.
Beyond forecasting, we perform predictability attribution across fire types and regions. SHAP analyses under an IPS stratification show that persistent fire prediction is dominated by prior fire states, spread fires depend on coupled fuel–environment conditions, and ignition is driven mainly by vegetation and land-surface properties with a stronger role of soil moisture. Region-aggregated attribution further indicates that FRP and NDVI are consistently influential predictors, while secondary drivers vary by region and fire regime, with meteorological controls shifting in importance and lightning density contributing more strongly in regions with frequent lightning-driven ignitions.
Overall, the STA-Net provides a high-skill and scalable approach for global short-term daily wildfire forecasting together with transparent attribution of predictive drivers, supporting wildfire risk management and emission forecasting.
How to cite: Wu, T., Zheng, J., Ye, J., Huang, Z., Zhu, M., Chen, W., and Xue, Z.: Global Short-term Daily Wildfire Forecasting and Predictability Attribution using a new Spatio-temporal Deep Learning Framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12016, https://doi.org/10.5194/egusphere-egu26-12016, 2026.
Wildfire trends vary by region and are influenced by climate, vegetation, and human activity. Regional trends over the past 20 years have varied, though overall have driven a 60% increase in global wildfire carbon emissions, primarily from carbon-dense boreal forests. In addition to releasing carbon, wildfires alter surface albedo, aerosols, and vegetation dynamics, producing complex climate feedbacks. Representing patterns and quantities of burned areas across the globe is thus crucial to accurately predict future climate, but is difficult due to the nonlinear and spatially heterogeneous nature of wildfire drivers. In this work, we develop an artificial intelligence (AI)-based model to predict patterns and quantities of burned areas across the globe, with the goal of integrating it within the University of Victoria Earth System Climate Model (UVic-ESCM v2.10). Our model consists of a deep neural network trained with a new custom, spectral-based loss function (DNN-FFTLoss). We compare it with deep neural networks trained with a mean-square error loss function (DNN-MSE) and random forests (RF), using a consistent set of climate and vegetation predictors from the UVic-ESCM v2.10.Training is performed using climate and vegetation predictors from CMIP6 simulations (1850–2100, including multiple Shared Socioeconomic Pathway (SSP) scenarios) alongside satellite-based Global Fire Emissions Database (GFED) 4 burned area observations (2001–2015). Transfer learning is then performed using the GFED4 dataset to impose observational constraints, reduce biases, and improve burned area predictions and the representation of fire-climate interactions. A comparison with the independent test year (2014) reveals that the DNN-FFTloss more accurately reproduces the spatial and seasonal variability of global burned area than the DNN-MSE and RF. However, the DNN-FFTloss still exhibits regional biases, overestimating burned area in Northern and Southern Africa and Australia and underestimating it in Europe. Nevertheless, these discrepancies are reduced relative to the other architectures. Additionally, the global cumulative density function of burned area is best captured by the DNN-FFTloss, indicating improved representation of both high- and low-burn regions. All model configurations show reduced skill temporally during the spring transition (e.g., March-April), when global Pearson correlations drop to 0.3 for the DNN-MSE model and 0.6 for the DNN-FFTloss model. Overall, the DNN-FFTloss better represents the global behaviour of wildfire burned area and will provide new insights into how climate change alters wildfire regimes and their impact on terrestrial carbon storage.
How to cite: Chalifour, O., Boussard, J., and Matthews, D.: An AI-driven approach to enhancing wildfire representation and climate feedbacks in the UVic-ESCM v2.10, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4913, https://doi.org/10.5194/egusphere-egu26-4913, 2026.
The increasing number of wildfires in Italy presents a growing challenge for environmental protection and infrastructure resilience. Among the most vulnerable assets is the high-voltage transmission network: during wildfire events, in fact, overhead lines must often be preemptively deactivated to facilitate aerial and ground-based firefighting and to preserve infrastructure integrity and grid stability. This necessity creates a critical conflict between emergency response requirements and the continuity of electricity supply.
While anthropogenic activities and human negligence remain the primary drivers of ignition, the meteorological conditions leading to fire spread have worsened in recent years due to persistent summer heatwaves and prolonged droughts. To monitor and predict wildfire danger, various meteorological indices have been developed, most notably the Canadian Fire Weather Index (FWI). However, while these indices are essential for daily operational monitoring, they are inherently limited by not considering fuel availability and terrain characteristics. Consequently, high FWI values may be recorded in areas with no combustible biomass, such as urban areas, highlighting the limits of purely weather-based fire danger assessments.
To improve fire danger characterization, a susceptibility map was developed on a 100-meter resolution grid covering the entire Italian territory. To achieve this, a random forest model was trained on non-meteorological, high-resolution data using a balanced dataset constructed from areas burned between 2010 and 2023, and an equal number of randomly sampled non-fire locations. These features included land use, topography (elevation, slope, and aspect), latitude, and proximity to critical infrastructure (roads and power lines). The model demonstrated high predictive performance, achieving an accuracy of 0.95 on a 30% hold-out test sample; feature importance analysis revealed that latitude, elevation, and land-use class are the primary drivers of fire susceptibility. Finally, the model has been applied across the entire Italian Peninsula, yielding a high-resolution map of burning probability for each grid cell.
To evaluate its operational effectiveness, the susceptibility map was validated against two case studies where wildfires directly caused the deactivation of critical power lines. The results demonstrate that the map significantly refines the spatial accuracy of coarser meteorological alerts based solely on the FWI. By integrating fuel and topographic data with weather-based indices, the model successfully narrows the focus to specific high-risk segments of the grid, thereby reducing 'false alarm' areas and providing a more targeted decision-support tool for transmission system operators.
This susceptibility map provides an important foundation for a comprehensive wildfire alert system, bridging the gap between broad meteorological forecasts and local-scale infrastructure needs. By refining established weather indices with high-resolution environmental and topographic data, the model allows for a level of situational awareness compatible with the needs of power grid operators within the growing challenges of Mediterranean climate.
How to cite: D'Amico, F., Bonanno, R., Collino, E., Viterbo, F., and Lacavalla, M.: Wildfire Susceptibility in Italy: High-Resolution Mapping for Power Grid Resilience, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6395, https://doi.org/10.5194/egusphere-egu26-6395, 2026.
Accurate wildfire risk assessment is essential for disaster mitigation and landscape management, particularly in Mediterranean ecosystems. A number of wildfire risk maps for Cyprus use expert-driven indices, single-model statistical methods, and data from remote sensing. However, there is currently no standardized, high-precision Fire Risk Index (FRI) that comprehensively considers multiple risk factors and provides accurate, consistent predictions across different areas. This study introduces an innovative multi-stage machine learning framework designed to develop a comprehensive Static Fire Risk Index (FRI) for Cyprus. The methodology consists of two primary phases: the creation of four thematic sub-indices and their subsequent integration through an ensemble meta-modeling approach. More specifically, a topographic risk index was derived from derivatives of an EU Digital Elevation Model (DEM) (25 m spatial resolution), namely slope, elevation, aspect, plane curvature, and classification of landforms. A vegetation-moisture risk index was generated using multi-temporal satellite imagery from Landsat 8 and 9 to calculate the Leaf Area Index (LAI), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and Normalized Difference Moisture Index (NDMI). Fuel flammability index was assessed using a comprehensive vegetation type map, while an anthropogenic risk index included factors such as population density, proximity to roads, transmitter stations, picnic sites, power lines, and built-up regions to address human-induced fire risks. The historical fire location data from 2015 to 2024 were extracted from VIIRS sensors to facilitate the development of machine learning models. Initially, four thematic fire risk indices were generated: Fuel Flammability, Vegetation Moisture, Topography, and Anthropogenic Risk. These indices were subsequently standardized into five ordinal fire danger classes, ranging from 1 (Very Low) to 5 (Very High).
To determine the most effective integration strategy, eight distinct machine learning architectures were benchmarked: Random Forest (RF), XGBoost, LightGBM (LGBM), Decision Trees (DT), Support Vector Machines (SVM), Artificial Neural Networks (ANN), K-Nearest Neighbors (KNN), and Logistic Regression (LR). Model bias and uncertainty were assessed using cross-validation with historical fire occurrences, along with an examination of prediction residuals and spatial error patterns. The performance evaluation, which focused on accuracy (83%) and Area Under the Curve (AUC) (0.87), revealed that tree-based ensemble models (RF, XGBoost, LGBM, and DT) significantly outperformed both baseline and kernel-based algorithms. Consequently, these four top-performing models were chosen for the final fusion stage.
A "Soft Voting" ensemble method was used to combine the predictions of the chosen models. This approach involved pixel-wise averaging of fire occurrence probabilities, which effectively minimized individual model bias and improved spatial stability. The resulting continuous probability map was then reclassified into five distinct threat classes using the Jenks Natural Breaks optimization method. Validation against historical fire data demonstrated that this consensus-based methodology provides superior predictive reliability in comparison to single-algorithm models. The final Fire Risk Index (FRI) map acts as a high-resolution decision-support tool, allowing fire management authorities to prioritize resources in high-vulnerability zones through a mathematically robust and standardized classification system.
How to cite: Suresh Babu, V., Sarris, A., and Stagonas, D.: Development of an Integrated Static Fire Risk Index for Cyprus Utilizing Tree-Based Ensemble Classifiers: A Soft-Voting Approach, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3053, https://doi.org/10.5194/egusphere-egu26-3053, 2026.
Wildfires critically alter hydrological regimes in Himalayan watersheds, yet their quantitative impacts remain poorly understood. This study integrates remote sensing–derived burn severity data with the SWAT model to assess postfire hydrological responses in the Central Himalayan Kosi River Basin (2013–2019). Burn severity information derived from Landsat-8 Operational Land Imager (OLI) imagery was used to update leaf area index (LAI) and curve number (CN) parameters within SWAT model to represent fire-induced surface modifications. The model showed satisfactory performance (R² = 0.67 calibration; 0.66 validation). Results indicated that extensive burns, particularly in 2013 and 2016, increased surface runoff by 20–34% and water yield by 13–20%, while reducing evapotranspiration by 17–24% and recharge by up to 7%. The findings highlight that Subbasin 16 experienced repeated moderate-to high-severity burns throughout 2013–2019 and exhibited the most intense and consistent fire effects. This subbasin is hydrologically more sensitive and likely contribute disproportionately to surface runoff and erosion during postfire periods. Therefore, targeted reforestation and soil stabilization efforts should be prioritized to reduce postfire runoff and erosion. These findings collectively emphasize ongoing postfire hydrological changes caused by vegetation loss and soil degradation, highlighting the importance of remote sensing–SWAT integration for postfire watershed management amid rising wildfire frequency.
How to cite: Das, B., Mamgain, S., Roy, A., Sharma, A., Sen, S., and Mukherjee, S.: Wildfire Severity and Post-Fire Hydrological Responses in a Central Himalayan Watershed: Integrating Remote Sensing and SWAT, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1102, https://doi.org/10.5194/egusphere-egu26-1102, 2026.
Fires play a critical role in shaping ecosystems, driving biogeochemical cycles, and influencing atmospheric composition. In many regions historically affected by fire, the frequency, intensity, and size of fires have undergone rapid change in recent decades, especially in high-latitude forests. Meanwhile, wildfire extremes are now emerging across many of the world’s forests and fire-sensitive ecosystems including regions such as the Amazon, Congo, Indonesia, and the Pantanal. Many of these ecosystems have evolved with little or no fire, increasing the impacts of these fires’ potential risk of climate-driven tipping points. It is therefore essential to accurately represent wildfire dynamics within Earth system models to quantify their influence on carbon–climate feedbacks and predict ecosystem responses, including potentially rapid and irreversible ones, to environmental change.
Modelling and understanding wildfires processes remain challenging due to complex interactions among climate, vegetation, human activity, and land-use change. The Joint UK Land Environment Simulator (JULES) provides a robust framework for simulating the dynamics of terrestrial hydrology, vegetation, carbon storage, and the surface exchange of water, energy, and carbon. Complementary Machine Learning (ML) techniques allow development of model emulators, enabling large-scale data processing and quantification of model uncertainty for a comprehensive analysis of potential wildfire driving factors.
Here, we will present an ML-based emulator for the JULES-INFERNO model to: (1) Analyse and understand the key climatic drivers for wildfire, characterising recent trends (such as the size, frequency and intensity of wildfires) across JULES model simulations; and (2) Evaluate and identify the potential for monitoring early warning signals for tipping points by combining model simulations, remote sensing data and Artificial Intelligence. The analysis and evaluation will contribute to a better understanding for wildfire processes and provide comprehensive information for policy makers.
How to cite: Zheng, L., Parker, R., Liu, Z., Ghent, D., Kelley, D., and Burton, C.: Understanding the drivers of wildfires using JULES model simulations and machine learning emulators, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3099, https://doi.org/10.5194/egusphere-egu26-3099, 2026.
The number of wildfires along the Mediterranean and Aegean coasts increases each year, impacting regional industries and ecosystems. In particular, the wildfire that occurred in Izmir, located in western Türkiye, on June 29-30, 2024, with peak temperatures exceeding 40°C and wind gusts reaching 22 m/s, spread to residential areas, resulting in the temporary closure of the city's airport and disrupting aviation operations. Therefore, predicting regional fire hazard risk based on meteorological data has become crucial, and many studies have been conducted in this area. The Canadian Fire Weather Index System (FWI) estimates forest fires based on the effect of fuel moisture and weather conditions. In this work, the risk of forest fires in Türkiye's Aegean and Mediterranean coastal regions has been estimated for future years using FWI data produced using high-resolution regional climate models supplied by the Copernicus Climate Change Service. The future years between 2026 and 2096 were compared under optimistic (RCP 2.6), moderate (RCP 4.5), and pessimistic (RCP 8.5) emission scenarios, with the 1971–2005 reference period. The results of this study showed that the number of extreme risk days (FWI > 45) increases from 50.48 days to 55.22 days (9.4% increase) under the RCP 2.6 scenario, to 57.26 days (13.4% increase) under the RCP 4.5 scenario, and to 61.71 days (22.2% increase) under the RCP 8.5 scenario when compared to the reference period. More significantly, according to the RCP 8.5 scenario, the risk level in coastal regions is estimated to reach 234.92 days annually, meaning that the risk of fires along the Aegean and Mediterranean coasts may last almost 65% of the year. In order to manage fire hazards in the Aegean and Mediterranean regions, where the risk of fire is extremely high, strategies that prioritize low-emission policies and carefully regulated tourism activitiesare crucial, as evidenced by the difference between RCP 2.6 and RCP 8.5 scenarios. The RCP 8.5 scenario also confirms that heat waves and altered precipitation patterns have increased the frequency and severity of these risks. These results indicate that the fire hazards will increase in the future, highlighting the importance of detailed information on fire risk assessment over the coastal areas of Türkiye’s Aegean and Mediterranean regions. In this context, the next phase of this study will focus on utilization of a Random Forest-based Inference Engine model to increase 12.5 km resolution of the EURO-CORDEX data to a 1 km spatial resolution in order to improve fire risk assessment. The model aims to identify non-linear wildfire risk patterns by correlating FWI components with local geographic features using an ensemble of decision trees. The proposed system is intended to operate as a Decision Support System (DSS) by automatically classifying extreme weather clusters, providing real-time resource allocation strategies.
How to cite: Aktaş, C. and Özdemir, E. T.: Spatio-Temporal Projection of Forest Fire Risk in the Aegean and Mediterranean Basins of Türkiye (2026–2096), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4032, https://doi.org/10.5194/egusphere-egu26-4032, 2026.
Flash droughts are characterised by rapid soil-moisture depletion driven by elevated atmospheric evaporative demand from higher air temperatures, low humidity, stronger solar radiation and wind, especially when precipitation is limited. This heightened atmospheric evaporative demand enhances evapotranspiration, accelerates moisture depletion in the root zone and intensifies vegetation water stress. As plants dry and weaken, their flammability rises, creating a feedback loop that elevates wildfire risk during prolonged heat and drought conditions.
This study investigates the relationship between flash drought and wildfire dynamics using two commonly used methods of flash drought detection across diverse land-cover types in the continental United States. We show that the frequency and spatial patterns of flash drought and its relationship with wildfire is significantly influenced by the method used for flash drought detection. Flash drought events identified by the Standardized Evapotranspiration Stress Ratio (SESR) capture atmospheric evaporative stress, while Root Zone Soil Moisture (RZSM) reflects sustained soil drying that directly increases fuel flammability. Approximately 53% of fires occurred after flash droughts identified using SESR definition, whereas RZSM classified about 10%, with each producing different spatial footprints.
To quantify how flash drought alters fire evolution, we applied Kaplan–Meier survival analysis to time-to-burn, estimating the probability that pixels remain unburned as a function of time since ignition under flash drought and non-flash drought conditions, and used Cox proportional-hazards models to derive hazard ratios (HR), which measure the relative instantaneous burning rate under FD (HR > 1 indicating faster spread). Grasslands and croplands show the highest vulnerability to flash drought–related fires due to their fine, continuous fuels that rapidly dry and ignite, with stronger acceleration and earlier spread under RZSM identified flash droughts (HR ≈ 1.45 in grasslands, 1.33 in croplands, 1.84 in open shrublands; woody savannas ≈ 1.17), while SESR effects are small or near zero in several covers (HR ≈ 1.05 in croplands and grasslands; ≈ 0.99 in woody savannas).
We therefore recommend incorporating rapid soil-moisture drying dynamics into wildfire risk models and enhancing real-time monitoring to strengthen early warnings and fire management, especially in ecosystems prone to swift drying and ignition.
How to cite: Zeraati, M., Fowler, H., Manning, C., and White, C.: Flash Droughts and Wildfire Interactions: Influence of Detection Methods on Fire Risk and Speed Across U.S. Landscapes, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7555, https://doi.org/10.5194/egusphere-egu26-7555, 2026.
Vegetation growing along railway corridors creates conditions in which fires can ignite and spread rapidly, even though steam locomotives—the historical source of many railway fires—are no longer in regular use. This study examines vegetation fires occurring near railway lines in the Czech Republic over the last 20 years, with the aim of understanding their temporal patterns, links to weather conditions, and spatial concentration. The analysis draws on detailed incident records from the national railway infrastructure manager and combines them with meteorological, geographic, and operational data to identify the factors that influence fire occurrence.
The results show that fires tend to cluster in the warmer part of the year, particularly from spring through late summer, and most often in the afternoon. Their occurrence is strongly associated with prolonged periods of elevated temperatures, limited precipitation, and low relative humidity. Logistic regression further revealed that infrastructure characteristics play a significant role: electrified lines, areas near railway stations, and sections with heavy freight traffic exhibit a markedly higher likelihood of fire. Conversely, higher elevations and greater distance from built-up areas reduce the probability of ignition.
Using the KDE+ method (https://www.kdeplus.cz), we identified more than 300 hotspots where fires repeatedly occurred, despite these locations representing only a very small fraction of the national rail network. These hotspots are typically situated in regions with warmer climates and on lines with substantial train movements. The findings indicate that even modern railway operations can generate ignition sources, such as sparks from braking systems.
Given projected increases in temperature and drought frequency due to climate change, vegetation fires along railways are likely to become more common. The identification of high‑risk segments therefore provides a valuable basis for targeted vegetation management and other preventive measures aimed at reducing the impacts of fires on railway operations and surrounding ecosystems. As part of our current research, we are developing an early‑warning system that integrates weather forecasts, fuel models, and operational data to alert railway managers to elevated fire risk in advance.
How to cite: Bíl, M., Nezval, V., and Andrášik, R.: Environmental and Operational Drivers of Vegetation Fires Along the Czech Rail Network, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10852, https://doi.org/10.5194/egusphere-egu26-10852, 2026.
Forecasting fire danger is essential for early warning, fire management, and planning in several climate-sensitive industries. In Australia, fire regimes are highly seasonal and regionally diverse, creating a complex land-atmosphere interaction driven by extreme climate variability. This study is a preliminary investigation into the relationship between MODIS burned-area data and datasets that can act as predictors, such as seasonal Canadian Fire Weather Index (FWI) forecasts on multiple lead times from the ECMWF; coupled ocean-atmosphere climate modes – Indian Ocean Dipole (IOD) and El Niño-Southern Oscillation (ENSO); satellite-derived fuel-related variables NDVI, NBR, FAPAR at national and subnational (climate biome) scales, to inform the development of a region-adaptive forecasting framework.
Spatio-temporal correlation and spatial autocorrelation are assessed between gridded datasets, with time-series analysis focusing on lagged teleconnections and cross-correlation. In the case of the forecast-driven FWI diagnostic comparisons with reanalysis FWI is undertaken to provide context for forecast skill. These diagnostics are employed to investigate whether Australian fire regimes are governed by a dual-constraint system with a fuel-accumulation and climate-driven phase, in which antecedent fuel accumulation as well as weather triggers are the primary drivers.
The purpose of this study is to reveal the extent to which FWI’s ability to predict danger varies across biomes, highlighting the need for fuel-related inputs. Lagged analysis is used to inform the optimal temporal scale for predicting fire danger in Australia. Diagnostic comparison with reanalysis data may identify potential biases in the ECMWF forecast dataset that play a role in its relationship with burned area, further highlighting the need for a region-adaptive framework to correct for local land-mediated influences. These preliminary findings will shape ongoing research into the use of different combinations of variables by regions.
How to cite: Charlton, C., Galizia, L., and Voulgarakis, A.: Evaluating region-dependent skill of seasonal Fire Weather Index forecasts in Australia, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14026, https://doi.org/10.5194/egusphere-egu26-14026, 2026.
Wildfires pose a major threat to forest ecosystems worldwide, leading to substantial losses in ecosystem services. Since forests play a critical role in climate change mitigation and climate regulation, quantifying the rate and completeness of post-fire recovery is essential for assessing long-term ecosystem functionality. However, robust approaches to characterize the timing and trajectories of functional recovery after fire remain limited, particularly at large spatial and temporal scales.
Satellite remote sensing provides a unique opportunity to address this challenge by enabling globally consistent, long-term monitoring of post-fire vegetation dynamics across different land cover types, complementing the limited spatial and temporal coverage of ground-based observations. Based on the Fire Climate Change Initiative (Fire CCI) dataset, fire events are identified globally and used to define the spatial and temporal framework for the analysis. For each fire event, post-fire recovery trajectories are constructed from satellite-derived vegetation indicators capturing complementary aspects of forest condition and ecosystem functioning, including vegetation greenness (NDVI, EVI), canopy structure (LAI), and photosynthetic activity (FPAR).
These recovery trajectories allow post-fire recovery rates and relative recovery levels to be quantified and compared across land cover types at the global scale, revealing spatial differences and variability in recovery dynamics. The framework thus provides a scalable approach to assess long-term changes in forest ecosystem functionality following wildfires and to evaluate how post-fire recovery dynamics vary across land cover types and over time.
How to cite: Everke, J., Zotta, R.-M., Bader, N., and Dorigo, W.: Global Monitoring of Post-Fire Forest Recovery Using Satellite-Derived Vegetation Indicators, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16930, https://doi.org/10.5194/egusphere-egu26-16930, 2026.
Wildfire activity is influenced by a wide range of factors, meteorological, topographical, vegetation-related, and anthropogenic, making its modeling a highly complex task. In this work, we present a methodology that integrates two distinct modeling approaches within a single tool: a Machine Learning-based ignition model and a physical fire spread model.
Outcome of our approach are event-based burn probability maps, derived by aggregating the outcomes of many fire-spread simulations initialized from stochastic ignition events generated by a Machine Learning ignition model. This model is trained on historical ignition records and integrates meteorological, vegetation, and anthropogenic variables to yield daily ignition probability maps. From each daily map, we sample stochastic ignition events and run the fire spread model for each, generating an ensemble of plausible outcomes whose aggregated footprint yields the final event‑based burn probability map.
This combined approach enables us to address separately two critical wildfire processes: ignition and spread. Utilizing a data-driven model allows us to account for anthropogenic influences on ignition through variables such as proximity to roads, power lines, and land use. Meanwhile, the complexity of fire spread is handled by a physical propagation model that considers key factors such as fuel continuity, terrain, and processes like spotting.
The tool is currently under development within the UNICORN project, funded by the EU Horizon Europe Programme (grant agreement No 101180172), and is being tested in the cross-border region of Northwest Spain and Northern Portugal, one of Europe’s most wildfire-prone areas.
How to cite: Senande-Rivera, M., Baladima, F., Brosnan, V., Guerrini, F., and Pinilla, M.: A hybrid modeling approach for wildfire danger assessment: combining data-driven ignition and fire spread models , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14322, https://doi.org/10.5194/egusphere-egu26-14322, 2026.
The South Pacific High (SPH) plays a crucial role in shaping the climate of South America by influencing atmospheric and oceanic processes in Chile, such as upwelling, precipitation regime, and affecting the frequency of extreme climate events like heatwaves and extreme wildfires in the Mediterranean (30º-36ºS; MCh) and South-Central Chile (37º-42ºS; SCCh). Despite the relevance of SPH on the Chilean and South American climate at different time scales, its temporal Intensity changes have been partially understood to date. Here, we used monthly mean sea level pressure data from ERA5, spanning 1940 to 2024, to estimate the monthly SPH intensity (SPHI) following Barrett and Hameed (2017). We consider the annual year from January to December months, while summer is taken from the previous December to the current February, March to May as autumn, June to July as winter, and September to November as Spring. We examined annual and seasonal trends in SPHI and explored the relationships between gridded products of precipitation (Pr), minimum (Tn), and maximum temperatures (Tx) derived from the Centre for Climate and Resilience Research CR2 at 5 km. In addition, monthly surface soil moisture (SSM) from ERA5 has also analyzed with the SPHI. We computed Pearson correlations between the SPHI and the environmental variables during 1961-2024. Our findings indicate a significant increasing trend (p-value < 0.01) in the SPHI at annual and seasonal scales since 1940. In addition, Pearson correlations indicate a significant and negative relationship between SPHI and Pr and Tn at annual and all-year seasons in both sub-regions. The linkages between SPHI and Tx and SSM recorder significant and negative correlations during winter and spring in both sub-regions. Our results indicate severe changes in the SPHI on annual to seasonal scales, and also remark the strong modulation of the SPHI on Pr regime in both sub-regions. Furthermore, also reveals the relevance of the SPHI on the Tn modulation at annual and seasonal scales. Finally, relationships between SPHI and SSM in the spring are crucial to understanding, given the previous development of favourable fire conditions associated with wildfire dynamics and drought conditions in both Chilean sub-regions.
How to cite: Gonzalez-Reyes, A., Suazo Alvarez, M., Jacques-Coper, M., Christie Browne, D., and Bravo-Lechuga, C.: Changes in the South Pacific High intensity since the mid-20th century: implications and environmental impacts in the Mediterranean and South-Central Chile, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14770, https://doi.org/10.5194/egusphere-egu26-14770, 2026.
Climate change is expected to significantly alter regional fire regimes and forest vulnerability in temperate regions, including western Europe. In Belgium, wildfires have historically been relatively rare, but recent regional studies show a potential threat to population due to an increase of fire-prone climate conditions (Cerac, 2025). In this study, we assess future fire occurrence and tree mortality in Belgium over the 21st century using high-resolution, downscaled climate projections from the CMIP6 ensemble. Daily temperature, radiation, precipitation, humidity, and wind fields are dynamically downscaled by the regional climate model MAR (Grailet et al, 2025) to drive the dynamical vegetation model CARAIB (Verma et al, 2025) to derive occurrence of vegetation ignition and tree mortality for selected widespread species in Belgium. The simulations are performed for multiple baseline emission scenarios from IPCC (SSP2-4.5, SSP3-7.0 and SSP 5-8.5).
We show the main behavior of fire ignition occurrence and tree mortality obtained from the modeling exercise, first with a verification of the capabilities on the past period (1980-2025) when possible, and then with the future modelled trends (till 2100), especially in relation with the increase in the frequency and duration of summer drought periods and of the compound heat-dry events. Limitations of this exercise will be discussed to frame our future work.
This work provides one of the first climate-driven assessment of future fire risk and forest mortality for Belgium in the wake of the national climate downscaling experiment Cordex.Be2 (https://cordex.meteo.be/). It highlights emerging threats to temperate belgian forest ecosystems and offers a frame for quantitative information to support long-term forest management and adaptation strategies.
Cerac (2025): https://www.cerac.be/sites/default/files/media/files/2025-02/rpt_wildfire_risks_in_belgium_20250228_cerac_ngi_en_v2.0.pdf
Verma et al (2025): https://www.sciencedirect.com/science/article/pii/S0301479725003056
Grailet et al (2025): https://gmd.copernicus.org/articles/18/1965/2025/
How to cite: Ghilain, N., Francois, L., Lecart, B., Dethinne, T., Jonard, F., and Fettweis, X.: Modeling fire occurrence and tree mortality in Belgium for the 21st century using downscaled CMIP6 climate simulations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19149, https://doi.org/10.5194/egusphere-egu26-19149, 2026.
The frequency and severity of wildfires are increasing, with damaging effects on infrastructure, human populations and ecosystems. To inform risk mitigation planning, climate change projections are essential for assessing future trends in fire weather - meteorological conditions conducive to wildfire ignition and spread - and subsequently for identifying areas likely to face heightened wildfire risk in the future. This is particularly important in regions where wildfires are emerging as a notable threat in areas not historically considered fire-prone. One such example is South Asia, a region home to two billion people and already facing significant challenges associated with climate and environmental change.
Here, we examine how fire weather is likely to respond to a changing climate in South Asia. We first evaluate the ability of 14 state-of-the-art Earth System Model (ESM) ensembles from the sixth phase of the Coupled Model Intercomparison Project (CMIP6) to realistically represent observed mean, variance, and spatial variability statistics in the Fire Weather Index (FWI), using the ERA-driven global fire danger reanalysis as a reference. Those ESMs demonstrating an acceptable performance are used to quantify changes in the characteristics of a series of FWI-derived annual indicators throughout the 21st century under four emissions scenarios defined by the Shared Socioeconomic Pathways (SSPs). These projections are also analysed in relation to Land Use and Land Cover (LULC) classifications for each scenario. We find that seasonal means and annual maxima of FWI are projected to increase by up to 10% by the end of the century under the highest emissions scenario, while the incidence of extreme fire weather may rise by as much as 20 days per year under SSP5-8.5. Regarding projected changes across different LULC types, our results reveal significant positive trends in FWI metrics over forest and grassland areas under all SSP scenarios.
Overall, our findings contribute to a better understanding of future fire weather in a region historically unprepared for wildfire threats. We conclude by discussing the implications of these results for a range of stakeholders and their potential to enhance planning and preparedness at national and regional scales across South Asia, supporting the development of long-term mitigation and adaptation strategies.
How to cite: Eden, J., Zahoor, Z., Dieppois, B., and Blackett, M.: Projected changes in fire weather across South Asia using CMIP6 models under multiple emission scenarios , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19978, https://doi.org/10.5194/egusphere-egu26-19978, 2026.
As wildfire frequency and intensity escalate globally due to climate change, the development of robust, scalable predictive models becomes critical for effective disaster risk reduction. This research evaluates the adaptability of the Spatio-Temporal Google Earth Engine (STGEE) framework, originally designed for soil erosion modelling, to generate Wildfire Susceptibility Indices (WSI) across morphologically contrasting environments. The study focuses on two distinct sample areas: the rugged, mountainous terrain of Los Guájares, Spain, and the flat, homogeneous landscape of Hortobágy National Park, Hungary.
The methodology employs a Machine Learning (ML) approach within the cloud-computing environment of Google Earth Engine (GEE). A key innovation of this study is the adaptive selection of mapping units based on geomorphological characteristics. For the mountainous Spanish region, Slope Units (SUs) bounded by drainage and divide lines are utilized to capture topographic effects such as wind patterns and fire acceleration. Conversely, a pixel-based approach (30m * 30m) is applied to the Hungarian plain to address the relative topographic homogeneity.
The modelling process integrates a dual-component database. The inventory dataset comprises historical fire extents derived from Landsat and Sentinel-2 (MSI) products, paired with randomly sampled pseudo-absences. These are correlated with a suite of multi-source environmental conditioning factors, including topographic metrics (elevation, slope, aspect, TWI), vegetation and fuel proxies (NDVI, EVI), hydrological status (MNDWI), climatic variables (LST, precipitation, wind speed), and anthropogenic drivers (distance to roads and settlements).
Predictive modelling is performed using the Random Forest (RF) ensemble algorithm, selected for its capacity to handle non-linear interactions and multi-collinearity. To ensure model robustness and mitigate spatial autocorrelation, performance is validated using Spatial K-fold Cross-Validation. Model accuracy is assessed via the Area Under the Receiver Operating Characteristic Curve (AUROC), while Variable Importance Measurement (VIM) based on Gini Impurity is used to identify dominant fire drivers.
Preliminary hypotheses suggest that susceptibility in Los Guájares is primarily driven by topographic factors, specifically slope and aspect, whereas the Hortobágy model is expected to show higher sensitivity to vegetation moisture content and anthropogenic proximity. By successfully applying a unified methodology to heterogeneous terrains, this research aims to demonstrate the versatility of the STGEE framework in supporting targeted fire prevention strategies across diverse landscape types.
---
This research was funded by the Vicerrectorado de Investigación (University of Granada) with the Plan Propio PP2022.PP-12 on the “Caracterización de propiedades clave en la relación agua-suelo para el estudio de la influencia del fuego en el balance hídrico y el carbono para el planteamiento de estrategias de restauración”. Also, it is based on work funded by COST Action (grant no. FIRElinks CA18135), supported by COST (European Cooperation in Science and Technology).
How to cite: Bertalan-Balázs, B., Bertalan, L., Rodrigo Comino, J., Balogh, S., and Abriha, D.: Comparative wildfire susceptibility modelling in heterogeneous terrains, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20076, https://doi.org/10.5194/egusphere-egu26-20076, 2026.
Wildfire policies commonly define agency responsibility for wildfire management, but policy effectiveness is difficult to evaluate because of multiple interacting factors. Our research aims to determine (1) if synthetic control estimations can serve as a data-driven approach to assess effects of wildfire policy interventions, and (2) if the wildfire regime in Italy has been altered in response to a policy intervention (Madia’s law) that in 2017 imposed changes in the wildfire management system in most regions. Using a control pool of European countries, and with and without consideration of fire weather, we demonstrate that synthetic control estimations can be a suitable approach to model counterfactual trends in fire activity following a policy intervention. In Italy, models suggest the attribution of higher burned area and average fire size in the first year after Madia’s Law policy intervention was effective, though the effect appears to a varying degree across regions. We conclude that synthetic control estimations can form a valuable complement to expert-based assessments of wildfire policies in a range of flammable landscapes, although challenges remain due to complex interacting factors.
How to cite: Kirschner, J. A., Kirschner, J., Ascoli, D., Moris, J. V., Boustras, G., and Spadoni, G. L.: Using Synthetic Controls to Evaluate Wildfire Policy Impacts: Evidence from Madia’s Law in Italy, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21833, https://doi.org/10.5194/egusphere-egu26-21833, 2026.
Wildfire impacts are increasingly driven by the joint occurrence and persistence of multiple meteorological drivers, such as atmospheric dryness and strong winds, rather than by isolated univariate extremes. Growing evidence shows that such compound conditions strongly influence wildfire characteristics, including event duration, spatial extent, and intensity, motivating the use of multivariate probabilistic frameworks for wildfire risk analysis [1,2]. Traditional approaches based on marginal extremes or linear dependence are often inadequate for representing tail dependence and joint exceedance behavior, potentially leading to biased estimates of rare but high-impact wildfire events [3,4].
This study develops a spatially explicit, event-based probabilistic framework for modeling wildfire-relevant meteorological drivers and derived event characteristics using copula-based dependence structures. The methodology follows a two-stage workflow. In the first stage, hourly gridded fields of a humidity-related variable and wind are transformed into per-cell daily time series, extracting daily extrema and duration metrics based on physically motivated thresholds. A combined condition identifies hours when both drivers are simultaneously active, enabling the construction of compound duration indicators. This spatially explicit, per-cell representation is consistent with established wildfire risk and susceptibility frameworks that rely on pixel-level meteorological and environmental descriptors and supports the consistent aggregation of local information into larger spatial units relevant for regional risk assessment and comparison [5].
In the second stage, extreme events are detected and modeled to build an event-based probabilistic dataset and generate long synthetic event catalogs. Event identification relies on return-period exceedance of annual maxima, combined with moving-window logic and minimum inter-event time constraints. Event-level descriptors, including maximum driver intensity and persistence, are used to quantify spatially aggregated impacts, consistent with prior work on joint modeling of wildfire duration and size [6,7]. Marginal distributions are fitted to event-level variables and transformed into the probability domain prior to dependence modeling, following established copula theory [3]. Multivariate dependence is then modeled using copulas, allowing synthetic events to be generated while preserving observed dependence structures among drivers and event characteristics [4,8].
The framework builds on recent advances in compound and multihazard analysis [1,2], copula-based frequency analysis [3], and comparative evaluations of multivariate extreme modeling strategies [9]. By exporting spatially aggregated event-impact matrices and event frequencies, the approach is designed for integration into downstream wildfire hazard and risk assessment engines. Preliminary results of a pilot implementations at regional level in Italy (Liguria, Tuscany, Marche), adopting a 40-years weather dataset (1981–2023), are shown.
References
[1] Zscheischler & Fischer (2020), Weather and Climate Extremes.
[2] Sadegh et al. (2018), Geophysical Research Letters.
[3] Salvadori & De Michele (2004), Water Resources Research.
[4] Bhatti & Do (2019), International Journal of Hydrogen Energy.
[5] Trucchia et al. (2022), Fire.
[6] Ghizzoni et al. (2010), Advances in Water Resources.
[7] Xi et al. (2020), Stochastic Environmental Research and Risk Assessment.
[8] Najib et al. (2022), Natural Hazards.
[9] Tilloy et al. (2020), Natural Hazards and Earth System Sciences.
How to cite: aflakian, P., Colavitto, B., Trucchia, A., Ghizzoni, T., and Fiorucci, P.: Event-Based Copula Modeling of Compound Fire-Weather Extremes for Wildfire Risk Assessment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11241, https://doi.org/10.5194/egusphere-egu26-11241, 2026.
Extreme Wildfire Events (EWEs) represent a growing threat in Mediterranean regions, yet their short-term hydrometeorological drivers remain less well constrained than those of more frequent, lower-intensity fires. Improving the discrimination between extreme and non-extreme wildfire behavior is therefore essential for advancing fire prediction, early warning, and risk management. This study investigates whether EWEs differ significantly from non-extreme fires in terms of their associated dynamic meteorological, vegetation, and hydroclimatic conditions, using Italy as a national-scale case study representative of Mediterranean fire regimes.
We analyzed a high-resolution wildfire geospatial dataset from the Italian Civil Protection Department, comprising 106,620 fire events recorded between 2007 and 2022 and a total burned area of approximately 1.37 million hectares (Moris et al., 2024). Fires smaller than 1 ha were excluded. To explicitly account for the contrasting statistical behavior of extreme and non-extreme wildfires, we adopted a two-regime modeling framework: i) the bulk of the burned-area distribution was modeled using Generalized Additive Models (GAMs); ii) EWEs were characterized using an Extreme Value Theory (EVT) framework in which burned-area exceedances above high percentile-based thresholds (90th, 95th, and 99th percentiles) were modeled with a Generalized Pareto Distribution.
Our analysis is supported by the integration of data-cube technology, which enables efficient extraction of high-resolution spatiotemporal data. Meteorological, vegetation, and drought-related variables were extracted at daily and 1 km resolution from the Mesogeos dataset (Kondylatos et al., 2023). Only dynamic variables were considered, including meteorological fields from ERA5-Land; land surface temperature, Normalized Difference Vegetation Index, and Leaf Area Index from MODIS; soil moisture from the European Drought Observatory. The Standardized Precipitation Evapotranspiration Index (SPEI) was additionally included as a complementary indicator of drought conditions.
Results indicate that EWEs are governed by processes that differ fundamentally from those controlling more frequent, lower-intensity fires. By isolating the tail behavior of burned area, the EVT framework reveals the dominant influence of drought intensity, near-surface air temperature, and wind speed under rare but high-impact conditions, relationships that are largely obscured when relying solely on bulk-based models such as GAMs. These findings highlight the importance of explicitly modeling wildfire extremes and provide a robust statistical basis for improving extreme-focused fire danger assessment, early warning, and risk management in Mediterranean regions.
Moris, J. V., Gamba, R., Arca, B., Bacciu, V., Casula, M., Elia, M., Malanchini, L., Spadoni, 481 G. L., Vacchiano, G. and Ascoli, D. (2024) A geospatial dataset of wildfires in Italy, 2007- 482 2022. Technical report, Zenodo.
Kondylatos, S., Prapas, I., Camps-Valls, G. and Papoutsis, I. (2023) Mesogeos: A multi467 purpose dataset for data-driven wildfire modeling in the Mediterranean. Advances in 468 Neural Information Processing Systems 36, 50661–50676.
How to cite: Tonini, M., Ghasemiazma, F., Turco, M., Trucchia, A., and Fiorucci, P.: Short-Term Hydrometeorological Drivers of Wildfires in Italy: Insights from Extreme Value Modeling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12302, https://doi.org/10.5194/egusphere-egu26-12302, 2026.
Wildfires are an increasing threat in Mediterranean regions, where extreme fire weather and long-term fuel accumulation are driving more frequent and severe events. In this context, fast and reliable fire spread simulations are essential to support both risk mitigation planning and real-time emergency management. PROPAGATOR is a stochastic Cellular Automata (CA) wildfire spread simulator designed to generate ensemble-based fire spread forecasts. The model, currently available as both an online application and open-source software, operates within a raster-based framework in which each cell is described by static attributes (e.g. fuel type, topography) and dynamic drivers (e.g. wind, fuel moisture). Fire propagation is modelled through a stochastic contamination process between burning and unburned cells, allowing the production of probabilistic maps of fire spread, as well as statistics on rate of spread and fireline intensity. PROPAGATOR also includes the capability to simulate the spotting phenomenon and suppression actions such as water drops or firebreak construction, making it suitable for both operational decision support during active fires and pre-event risk mitigation analyses.
A current limitation of operational applications of PROPAGATOR is its focus on surface fire propagation, with no explicit representation of vertical fuel structure or transitions to crown fire. Crown fires, however, are characterized by higher spread rates, greater energy release, and increased unpredictability, with major implications for suppression effectiveness and ecological impacts. To address this limitation, an enhanced version of PROPAGATOR has been developed by extending the model to a quasi-three-dimensional (2.5D) representation of fuels, enabling the simulation of crown fire processes within the stochastic CA framework. The proposed Crown Fire Module relies on established empirical and semi-empirical formulations for crown fire initiation and spread that are compatible with a cellular automata approach. Crown fire initiation is governed by surface fireline intensity and crown base height, while crown fire rate of spread depends primarily on canopy bulk density and fire behaviour. These mechanisms have been integrated into the propagation rules of PROPAGATOR, allowing dynamic transitions between surface and crown fire behaviour within a probabilistic modelling framework.
The implementation of these processes requires detailed information on both surface and canopy fuel structure and characteristics, which remains challenging at operational scales. To address this issue, we investigated the use of UAV-based LiDAR remote sensing to derive key fuel structure parameters using semi-automatic algorithms available in the literature. This approach offers a balance between spatial detail and areal coverage that is suitable for operational wildfire applications.
A pilot study conducted in the Venafro area (Molise, Italy), based on a past wildfire event with a comprehensive dataset describing fire evolution, provided high-resolution inputs to test the enhanced model. By explicitly simulating surface-to-crown fire transitions, the upgraded version of PROPAGATOR aims to improve decision support for wildfire risk management, supporting applications ranging from fuel treatment planning to operational response under extreme fire weather conditions.
How to cite: Trucchia, A., Colle, F., Perello, N., Fagugli, G., D’Andrea, M., Taccaliti, F., and Fiorucci, P.: Integrating UAV–LiDAR Fuel Data into Stochastic Cellular Automata PROPAGATOR for Crown Fire Modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14672, https://doi.org/10.5194/egusphere-egu26-14672, 2026.
Over the past decade, the operational unit BEYOND Centre in the National Observatory of Athens (NOA) has developed and presented an advanced wildfire monitoring and forecasting framework for Greece, namely FireHub. The system is ingesting real-time Meteosat Second Generation satellite data every five minutes through NOA/BEYOND’s in-house antenna, using SEVIRI Level 1.5 infrared bands (IR 3.9 and 10.8 μm) to detect ignition points with quantified confidence. A dedicated downscaling methodology refines detections to a much finer scale (300 m) than the native SEVIRI spatial resolution of 3 km. The system is further enhanced by integrating the Firehub Fire Information System (FFIS), which combines observations from VIIRS, MODIS, and Sentinel-2, providing a more comprehensive and reliable picture of the active fire state. To address early-stage satellite artifacts caused by clouds, smoke, or extreme temperatures, NOA/BEYOND has long coupled observations with fire propagation modeling, initially through the deployment of FLAMMAP alongside real-time meteorology, fuel types, and terrain information. While this hybrid approach proved accurate and well received, it faced constraints under the rapidly growing incident volume that required overwhelming computational resources. In addition, FLAMMAP relying on a static wind field defined only at ignition, limited the realism in complex and highly variable wind environments.
Under the framework of MedEWSa project, the entire system has been re-engineered from the ground up to overcome these limitations. The new architecture runs asynchronously and concurrently on high-performance computing nodes, leveraging optimized code and open data cubes to scale efficiently. FLAMMAP has been replaced by the FOREFIRE model, which incorporates wind variability in both space and time from ignition onward. Sensitivity tests demonstrate that fully dynamic wind simulations produce fire evolutions closer to observed burned scar maps than static approaches. Extensive testing across coastal zones, urban and suburban settings, and complex terrain, using multiple propagation schemes including Rothermel, Balbi, and the newly added FARSITE, has guided the selection of an operational configuration. In peak periods, dozens of fires were handled simultaneously and each ignition triggering parallel, automated propagation forecasts for the coming hours. During the 2025 fire season, the system ran in pseudo-operational mode, allowing a full evaluation to take place against the confirmed ignition points by the Hellenic Fire Service. Further developments are currently underway such as the switch to Meteosat Third Generation, in order to utilize the 1x1-km resolution scans every 10 minutes (2.5 minutes from 2027). Real-time monitoring and fire propagation outputs are presented as overlays with critical infrastructure layers, in order to support rapid action from first responders and informed decision-making by relevant authorities. The latest state will be presented just before the system’s inaugurate fire season as the operational platform of NOA/BEYOND.
How to cite: Bartsotas, N. S., Herekakis, T., Kanaki, V., Zachariadis, P., Tsoutsos, M.-C., and Kontoes, C.: Towards a Second-Generation Wildfire Detection and Forecasting Platform: Technical and Operational Advances in FireHUB., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6748, https://doi.org/10.5194/egusphere-egu26-6748, 2026.
Wildfires are a hazardous concern for human health and the environment, extensively studied for fire-prone regions for decades. However, in temperate Central Europe a significant gap remains in evaluation, assessment and understanding of the effects and risks on the geoenvironment, including post-fire pollutant cycling. The Harz Mountains in Central Germany face this environmental challenge, due to climate change, which is driving expansion of fire-prone regions and an increase in the frequency and number of wildfires. Especially since 2022, the region has experienced wildfires following natural disturbances such as bark beetle infestation, windthrow, as well as a high frequency of heat and drought events. The landscape is shaped by legacies of land use during the past millennium. Mining, smelting and wood overexploitation phases significantly altered the topography and soils leaving widespread and partly hazardous environmental legacies, suggested to interact with modern environments, though the extent of this interaction remains poorly understood. We suggest that this interplay of recent wildfires and legacies, represented by former charcoal production sites, creates diverse fire impacts on soils within a single region. On the one hand, the widespread residues of charcoal kilns persist in the soils and on the other hand, modern wildfire affected soils again.
Our study investigates the influence of the landscape legacies in recently burnt areas by analyzing 16 priority PAHs (Polycyclic aromatic hydrocarbons) listed by the U.S. Environmental Protection Agency in a 1.3 ha site in the Harz Mountains that burnt in 2022 (Jackisch et al., 2023). Samples of organic and mineral horizons were taken in former charcoal kiln and wildfire affected sites mapped by remote sensing. Additionally, control soil profiles were sampled. All samples were analyzed using GC-MS.
We examined the influence of heat on the mineral layer through changes in mineral composition with a focus on thermal transformation of Fe (oxy)hydroxides using SEM (scanning electron microscopy) and XRD (X-ray Diffraction) measurements in mineral layers affected by charcoal production, wildfires and non-affected soils, to improve the mapping of burn severity. We find a high heterogeneity in PAH quantities and composition due to the site’s high soil and micro-relief diversity, with high-molecular weight PAHs dominating in legacy samples. This study contributes to the discussion about post-fire PAH cycling in soils of the Harz Mountains with legacies from past charcoal production.
Jackisch, R., Putzenlechner, B., & Dietze, E. (2023). UAV data of post fire dynamics, Quesenbank, Harz, 2022 (orthomosaics, topography, point clouds) (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7554598
How to cite: Deutsch, L., Yadav, A., Jackisch, R., Kronz, A., and Dietze, E.: Multiple burns affecting post-fire pollution cycling: Legacies of past charcoal production in areas affected by forest fires , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12914, https://doi.org/10.5194/egusphere-egu26-12914, 2026.
Posters virtual: Fri, 8 May, 14:00–18:00 | vPoster spot 3
EGU26-15290 | ECS | Posters virtual | VPS14
The Impact of Radiometric Terrain Normalization (γ⁰) on Burned Area Mapping Accuracy Using Sentinel-1 dataFri, 08 May, 14:12–14:15 (CEST) vPoster spot 3
Wildfires are increasingly destructive events, threatening ecosystems and human infrastructure while contributing significantly to carbon emissions. Accurate and timely burned area mapping is therefore essential for effective mitigation and recovery. Optical satellite sensors are often hindered by clouds and smoke, making Synthetic Aperture Radar (SAR) sensors like Sentinel-1, with their all-weather capability, a crucial tool for monitoring. However, SAR backscatter is significantly influenced by topography, which can distort signals and hinder accurate detection.
This study evaluates the impact of angular-based radiometric terrain normalization (RTN) on burned area mapping using Sentinel-1 SAR data and the Normalized Radar Burn Ratio (NRBR) index. We compare the performance of NRBR calculated with standard sigma nought (σ⁰) and with gamma nought (γ⁰) corrected via an angular-based RTN model implemented in Google Earth Engine. A U-Net deep learning model was used to delineate burned areas in Portugal and California. Results show that NRBR without RTN achieved better accuracy in Portugal, suggesting potential overcorrection effects in moderate terrain. In California, RTN slightly improved overall accuracy and reduced commission errors, although omission errors remained high. These findings indicate that while RTN enhances radiometric consistency, its impact on burned area detection with NRBR is limited, likely because the NRBR formulation itself already mitigates topographic effects through pre/post-fire ratios.
How to cite: Tarazona, Y. and Mantas, V.: The Impact of Radiometric Terrain Normalization (γ⁰) on Burned Area Mapping Accuracy Using Sentinel-1 data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15290, https://doi.org/10.5194/egusphere-egu26-15290, 2026.
