Analysis of large fires in European Mediterranean landscapes: Lessons learned and perspectives

Introduction Extreme wildfires are increasingly prevalent worldwide, driving significant forest area loss and severe environmental and socioeconomic impacts (Cunningham et al. 2024). The Mediterranean, in particular, is projected to face heightened fire risks due to climate change-induced drier conditions and lower fuel moisture (de Rivera et al. 2020). However, the drivers of extreme wildfires remain poorly understood. Current fire models, typically calibrated on global fire datasets, are primarily designed to estimate annual total burned areas and struggle to capture the unique behaviours of extreme wildfires (Forrest et al. 2024). Furthermore, correlation-based approaches, which dominate current modelling efforts, may fail to identify the underlying causal drivers of these events and are poorly suited for extrapolation to changing conditions. Causal discovery methods, which aim to identify cause-and-effect relationships from observational data, offer a promising pathway to uncover the mechanisms driving extreme wildfires. While increasingly applied in environmental sciences, their use in wildfire prediction remains limited (de Rivera et al. 2020, Zhang et al. 2024, Zhao et al. 2024).This study will use causal discovery to identify key drivers of extreme wildfire in the Mediterranean, and further integrate the causal graphs into a stand-alone model of wildfire spread. This approach aims to move beyond correlation-based models, improve our understanding of extreme wildfire behaviour and inform more robust mitigation strategies. Study Area and Data We will use the Mesogeos dataset (Kondylatos et al. 2023), designed for wildfire modelling in the Mediterranean region. Spanning 17 years (2006–2022) at a 1 km² spatial and daily temporal resolution, it includes meteorological variables (e.g., temperature, wind speed), vegetation indices (e.g., NDVI, LAI), and human activity indicators (e.g., population density, road proximity). Wildfire data include MODIS fire ignitions and burned areas from EFFIS. Methods Extreme Wildfire Definition and Sampling In this study, we define extreme wildfires as those that are exceptionally large in size. To identify these events, we will first extract the final burned areas associated with each fire ignition recorded in the Mesogeos dataset. Since the classification of large fires is inherently subjective and varies by region, we will adopt a data-driven approach based on an absolute quantitative threshold. Specifically, we will define extreme wildfires as those exceeding the 99th percentile of fire sizes, though this threshold may be adjusted to align with extreme fire events documented in national fire reports. While this method provides a straightforward and reproducible way to define extreme events, we acknowledge its limitations. Future work will refine this approach by incorporating region-specific thresholds and additional contextual factors to improve geographic relevance. Phase I: Causal Discovery Using local variables from Mesogeos, averaged over final burned areas and lagged to time t, we will estimate causal graphs for extreme events via Python’s Tigramite library with the PCMCI method (Runge et al. 2019). PCMCI detects time-lagged causal associations in large nonlinear datasets through iterative conditional independence testing. To ensure robustness, we will assess graph stability across hyperparameters and selected drivers, and validate graphs through expert knowledge. Phase II: Causal Fire Spread Model We will develop a fire spread model incorporating causal mechanisms from Phase I. This model will integrate spatiotemporal fire dynamics, causal dependencies constraining fire spread, and dynamic weather and fuel inputs. By explicitly modeling causal interactions, it aims to improve early warning systems and risk assessments under future climate scenarios. The causal model’s performance will be benchmarked against statistical models to evaluate its predictive accuracy and robustness. Expected Results We expect that the data-driven approach proposed in this study will enhance the predictability of extreme wildfires by reducing confounding effects and capturing key drivers of extreme fire events. Compared to purely statistical approaches, incorporating causal structures should lead to more reliable predictions, particularly in out-of-sample applications or under changing environmental conditions. Furthermore, the causal fire spread model will provide insights into how climate, vegetation, and anthropogenic factors interact to drive fire spread, supporting fire prevention and mitigation strategies.

Forest and wildland fires are a natural part of ecosystems worldwide, but large fires in particular can cause societal, economic and ecological disruption. Fires are an important source of greenhouse gases and black carbon that can further amplify and accelerate climate change. In recent years, large forest fires in Sweden demonstrate that the issue should also be considered in other parts of Fennoscandia. This final report of the project “Forest fires in Fennoscandia under changing climate and forest cover (IBA ForestFires)” funded by the Ministry for Foreign Affairs of Finland, synthesises current knowledge of the occurrence, monitoring, modelling and suppression of forest fires in Fennoscandia. The report also focuses on elaborating the role of forest fires as a source of black carbon (BC) emissions over the Arctic and discussing the importance of international collaboration in tackling forest fires. The report explains the factors regulating fire ignition, spread and intensity in Fennoscandian conditions. It highlights that the climate in Fennoscandia is characterised by large inter-annual variability, which is reflected in forest fire risk. Here, the majority of forest fires are caused by human activities such as careless handling of fire and ignitions related to forest harvesting. In addition to weather and climate, fuel characteristics in forests influence fire ignition, intensity and spread. In the report, long-term fire statistics are presented for Finland, Sweden and the Republic of Karelia. The statistics indicate that the amount of annually burnt forest has decreased in Fennoscandia. However, with the exception of recent large fires in Sweden, during the past 25 years the annually burnt area and number of fires have been fairly stable, which is mainly due to effective fire mitigation. Land surface models were used to investigate how climate change and forest management can influence forest fires in the future. The simulations were conducted using different regional climate models and greenhouse gas emission scenarios. Simulations, extending to 2100, indicate that forest fire risk is likely to increase over the coming decades. The report also highlights that globally, forest fires are a significant source of BC in the Arctic, having adverse health effects and further amplifying climate warming. However, simulations made using an atmospheric dispersion model indicate that the impact of forest fires in Fennoscandia on the environment and air quality is relatively minor and highly seasonal. Efficient forest fire mitigation requires the development of forest fire detection tools including satellites and drones, high spatial resolution modelling of fire risk and fire spreading that account for detailed terrain and weather information. Moreover, increasing the general preparedness and operational efficiency of firefighting is highly important. Forest fires are a large challenge requiring multidisciplinary research and close cooperation between the various administrative operators, e.g. rescue services, weather services, forest organisations and forest owners is required at both the national and international level.

In order to assess fire and wildfire risk one must address various features and analyse the danger and vulnerability aspects. Besides fire ignition probability, one of the most important variables for addressing fire danger is fire propagation potential. Fire propagation potential (FPP) can be described as a quantitative description of the circumstances under which, if fire ignites, it leads towards propagation of fire. This means that not all ignitions cause propagation of significant fires. Some ignitions are easily extinguished and pose no danger to vulnerable assets. On the other hand, some ignitions result in large and mega fires, causing large, burned areas and huge casualties. Fire propagation potential (FPP) provides quantitative distinction between these two different circumstances. Machine learning techniques are more and more applied in fire management tools as they provide us with techniques for learning from the past data and predicting the future outcomes. Majority of previous work is focused on analysis of the large fire events, their causes and development. However, when modelling the FPP, we should consider situations on both ends of the outcome spectrum - situations when fire ignites and propagates and situations when fire ignites and does not propagate. If one uses only data on fires that propagate, without considering the alternative situations data, results that are achieved can be incomplete. In this paper we propose a novel and more full approach to fire danger assessment by analysing situations of both cases - high and low fire danger. We simplify the value of FPP and consider that in cases the fire propagates the value of FPP is one, and zero otherwise. We used data collected from the events of both cases. We obtained a balanced dataset and trained machine learning model with a data set having representatives of both ends of the FPP spectrum. The research is demonstrated in the study area of Split and Dalmatia County. We consider past fires that are sensed by satellite and recorded in the EFFIS system as situations when FPP had value 1. To assess the situations when FPP was 0 we analysed the fire intervention database maintained by fire departments. We filtered fire interventions related to forest fires that lasted less than 2 hours and engaged 2 or less firefighters since these records represent time and place of the fire that did not propagate. For these two cases of events, we collected Sentinel-2 imagery and weather data that consists of temperature and wind speed. Sentinel-2 imagery pixels were extracted for the area associated with both types of events. The dataset was split into train and test datasets, where classifiers were trained by using 80% of data and 20% of remaining data was used for testing the classifier performance. Experiments were conducted by training classifiers using commonly used classifiers - Decision Tree Classifier, K-Nearest Neighbors, Multi-layer perceptron, Random Forest Classifier, Naive Bayes Classifier and Logistic regression. The best performance, according to the R2 score and RMSE is measured on Decision Tree Classifier.

Wildfires are an important disturbance in global ecosystems and are a critical driver of trends in the land carbon budget. Fire is an extreme phenomenon, with the largest burned area often occurring during extreme fire seasons generating large fires. Days with fire conditions conducive to fire ignition and spread are increasing in a warming climate in many regions of the world, contributing to increases in fire occurrence and annual burned area. However, the climate, fuel, and weather conditions that lead to extremely large fires in different biomes are poorly understood.Here, we explore the temporal evolution of extremely large fires in temperate and boreal regions using new satellite-derived fire event tracking datasets optimized to match higher resolution time series of fire progression from aircraft and other sources. We aimed to understand the specific environmental conditions required for the development of a large fire. Our analysis revealed a disproportionate impact of multiple fire ignitions in creating large fires through merging. Our findings suggest that the largest fires in both biomes may be commonly created through multiple fires growing together. We hypothesize that a combination of physical and anthropogenic factors may accelerate merging, making these fires extremely difficult to contain and more robust to environmental controls regulating extinction. In our analysis, we use the Fire Events Database, the Arctic-boreal Fire Atlas, and GOFER, which enable attribution of ignition sources. Our analysis may contribute to an improved understanding of the influence of large-scale lightning storms in creating extremely large and destructive fire events.

Large forest fires have repeatedly affected Europe, in particular the Mediterranean countries. It is now crucial to continue the effort to feed the several layers of prediction models and understand how wildfire spreads in order to develop modern preventative and mitigation methods. The present study evaluates the performance of Sentinel 5-P TROPOMI for the monitoring of carbon monoxide (CO) and methane (CH4) during extreme fire events in Portugal, focusing on the Monchique (2018) and Vila de Rei/Mação (2019) wildfires, which devastated 27,154 ha and 9249 ha, respectively. The spatial distribution and trend of CO and CH4 prior to, during, and following the fire event were accessed and linked with in situ data in a qualitative and quantitative exploration. Large CO plumes were observed with CO columns exceeding 4.5 × 1018 and 6 × 1018 molecules/cm2 on 21 July 2019, and 7 August 2018, respectively. CO distribution profiles after consecutive digital processing steps showed the ability to follow CO fluctuations according to the fire spread. Furthermore, statistically significant differences were found between CO emissions inside and outside the burning area in both fire events. Finally, the CO2 estimated through CO column data presented an emission of 7.6 × 1019 molecules/cm2 for the uppermost emission day on 7 August 2018. Although CH4 monitoring is still unwavering to draw exact conclusions, the CO patterns during extreme fire events show promising and consistent data when compared with in situ data.

Climate is a key driver of spatiotemporal variability in wildfire regimes. Increasingly warmer temperatures and changes in precipitation patterns are linked to an increase in fire danger across the world. There is therefore a need to develop the most reliable projections of future climate-driven fire danger to enable decision-makers and forest managers to prepare for and respond to future fire events. Earth system model (ESM) simulations are the foundation for understanding future changes in fire-conducive weather associated with a warming world. Fire weather projections have typically been expressed by a single model or through a multimodel mean, with the models’ relative strengths and weaknesses rarely taken into consideration. This study presents a novel set of future scenarios in fire-prone conditions, defined by the fire weather index, using a statistical weighting approach that specifically accounts for the performance and interdependence of 26 ESMs from the Coupled Model Intercomparison Project phase 6 (CMIP6). When compared with reanalysis-derived data from 1980 to 2014, the weighting approach substantially reduces multimodel bias, thereby demonstrating its added value in accounting for and reducing model uncertainties. The weighted projections reveal significant increases in seasonal fire weather conditions across 68%–91% of the world’s fire-prone area by the end of the twenty-first century, depending on the emission scenario. Additionally, at least 55% of fire-prone area is expected to be associated with a significant increase in fire weather conditions by 2040. The conclusions highlight the potential benefit to targeted prevention strategies and long-term fire management. Significance Statement This study aims to assess future changes in fire danger driven by climate conditions under different climate change scenarios. The results show an overall increase both in magnitude and geographical expansion in the weather conditions enabling ignition and spread of fires. The value of the results lies in the use of a weighted multimodel ensemble mean, reducing climate model biases with respect to reanalysis data in comparison to previous fire danger studies. This is important to better target areas of the world that are, and will very likely be, at higher fire risk in the coming years.

Reconstruction of fire spread within wildland fire events in Northern Eurasia from the MODIS active fire product

Abstract. Climate-driven changes in the fire regime within boreal forest ecosystems are likely to have important effects on carbon cycling and species composition. In the context of improving fire management options and developing more realistic scenarios of future change, it is important to understand how meteorology regulates different aspects of fire dynamics, including ignition, daily fire spread, and cumulative annual burned area. Here we combined Moderate-Resolution Imaging Spectroradiometer (MODIS) active fires (MCD14ML), MODIS imagery (MOD13A1) and ancillary historic fire perimeter information to produce a data set of daily fire spread maps for Alaska during 2002–2011. This approach provided a spatial and temporally continuous representation of fire progression and a precise identification of ignition and extinction locations and dates for each wildfire. The fire-spread maps were analyzed with daily vapor pressure deficit (VPD) observations from the North American Regional Reanalysis (NARR) and lightning strikes from the Alaska Lightning Detection Network (ALDN). We found a significant relationship between daily VPD and likelihood that a lightning strike would develop into a fire ignition. In the first week after ignition, above average VPD increased the probability that fires would grow to large or very large sizes. Strong relationships also were identified between VPD and burned area at several levels of temporal and spatial aggregation. As a consequence of regional coherence in meteorology, ignition, daily fire spread, and fire extinction events were often synchronized across different fires in interior Alaska. At a regional scale, the sum of positive VPD anomalies during the fire season was positively correlated with annual burned area during the NARR era (1979–2011; R2 = 0.45). Some of the largest fires we mapped had slow initial growth, indicating opportunities may exist for suppression efforts to adaptively manage these forests for climate change. The results of our spatiotemporal analysis provide new information about temporal and spatial dynamics of wildfires and have implications for modeling the terrestrial carbon cycle.

<p>Fire events on a landscape scale are a widespread global phenomenon that influences the interactions between atmosphere and biosphere. Global burned area (BA) products derived from satellite images are used in dynamic vegetation fire modules to estimate greenhouse gas emissions, available fuel biomass and anthropic factors driving fire spread. Fire size and shape complexity from individual fire events can provide better estimates of fuel consumption, fire intensity, post fire vegetation recovery and their effects on landscape changes to better understand regional fire dynamics. Especially in the Brazilian savannas (Cerrado), a mosaic of heterogeneous vegetation where has prevailed an official “zero-fire” policy for decades leading to an increase in large wildfires, intensified also by rapid changes of land use using fire to land clearing in agriculture and livestock purposes. In this way, we aim to assess the fire size and shape patterns in Cerrado from 2013 to 2015, identifying each fire patch event from Landsat BA product and calculating its fire features with landscape metrics. We calculated its surface area to evaluate fire size and the metrics of shape index, core area and eccentricity from an ellipse fitting from burned pixels to estimate the fire shape complexity. The study focused on 48 Landsat path/row scenes and the analysis final compared the fire features of overlapped patches between the years. The total number of coincident fire patches is higher between the years 2013 and 2015 than 2013-2014 and 2014-2015. Large fires are found in the north and east regions for all comparisons. In this region, high core area values are consistent for having large areas of burnt patches and low shape index values and more elongated patches revealed a low fire shape complexity. These results demonstrate a greater burned area in the north, where the remaining native vegetation and less fragmented landscapes allow the fire to spread, when associated with favorable meteorological conditions. However, with the implementation of a new agricultural frontier in 2015, this region is under greater anthropic pressure with positive trends to land use. In the south, the fire shapes are already more complex and smaller because they are from agricultural areas historically developed, and consequently the landscape is more fragmented. Our results demonstrate a distinct spatial pattern of fire shape and size in Cerrado related to fragmentation of landscape and fire use to land cleaning. This information can help the modelling estimates of fire spread processes driven by topography, orientation of watersheds or dominant winds at local level, contributing to understanding the feedback with land cover/use, climate and biophysical characteristics at regional level to develop strategies for fire management.</p><p><strong>Acknowledges:</strong> J.N is funded by the 'Women in Research'-fellowship program (WWU Münster) and within the context of BIOBRAS Project “Research-based learning in neglected biodiverse ecosystems of Brazil”; funding by DAAD (number 57393735); validation dataset was performed under the Andurá project (number 441971/2018–0) funding by CNPq</p>

Background Climate change has increased the occurrence of fire-prone weather and extreme weather events in Europe. Improving resilience of forests to reduce fire hazard has become an imperative challenge to tackle. Aims This study aims to incorporate extreme fire events from future climate projections in developing forest planning and land management scenarios, evaluating fire risk mitigation potential. Methods Using the wildfire–atmosphere coupled modelling system WRF-SFIRE, land management scenarios for the Lombada Forest Intervention Region (ILMA) in northeast Portugal were assessed under a business-as-usual future climate scenario. The Fire Weather Index (FWI) was used as a predictor of fire danger to select two fire events. Results Agricultural areas can function as barrier for wildfire conditions under cooler weather. Management of the existing pine forest yielded a reduction in fire spread speed and intensity, namely in lower wind speed regimes, significantly improving suppressive capabilities. Conclusions Fuel treatment of maritime pine performed best in reducing fire spread rate, intensity and improving suppression capability. Replacing agricultural areas with oak-based land cover can promote higher fire intensity and spread rates in the younger stanges or if left unmanaged. Implications This framework can be used to provide additional tools for forest management across different landscapes reducing fire hazard and vulnerability and improving forest resilience, under climate change.

Modelling the effect of fire-exclusion and prescribed fire on wildfire size in Mediterranean ecosystems

BackgroundCalifornia in the year 2020 experienced a record breaking number of large fires. Here, we place this and other recent years in a historical context by examining records of large fire events in the state back to 1860. Since drought is commonly associated with large fire events, we investigated the relationship of large fire events to droughts over this 160 years period.ResultsThis study shows that extreme fire events such as seen in 2020 are not unknown historically, and what stands out as distinctly new is the increased number of large fires (defined here as > 10,000 ha) in the last couple years, most prominently in 2020. Nevertheless, there have been other periods with even greater numbers of large fires, e.g., 1929 had the second greatest number of large fires. In fact, the 1920’s decade stands out as one with many large fires.ConclusionsIn the last decade, there have been several years with exceptionally large fires. Earlier records show fires of similar size in the nineteenth and early twentieth century. Lengthy droughts, as measured by the Palmer Drought Severity Index (PDSI), were associated with the peaks in large fires in both the 1920s and the early twenty-first century.

Complementing daily fire-danger assessment using a novel metric based on burnt area ranking

Characterizing extreme fire and weather events in the Boreal Shield ecozone of Ontario

The study is a step forward in the characterization of meteorological environments that favour the evolution of large and extreme fires in Southern Portugal. The region has some fire-prone areas which are recognized by the mega fires occurred in 2003, 2005, and 2018. Two numerical simulations were performed using the Meso-NH non-hydrostatic research model and used to investigate in detail the atmospheric environments of two large fires that occurred on 18th July 2012 and 19th June 2020. The simulations were configured using two nested domains with a 375 km × 375 km grid domain (D1) at 2.5 km horizontal resolution and a 150 km × 150 km domain (D2) at 500 m resolution added before the start of the fires. The vertical grid was configured with 50 stretched levels following the terrain. The initial and boundary conditions are provided by the 6-hourly operational ECMWF analyses. The large-scale circulation has been characterised using data obtained from the ECMWF's Meteorological Archival and Retrieval System. In addition to the large-scale circulation, namely the positioning of the Azores anticyclone and the thermal low development over the Iberian Peninsula, the results have shown the important role played by regional orography in creating favourable fire weather conditions. For instance, the high-resolution simulations showed the high daytime temperatures and sometimes overnight, low humidity, and strong wind gusts that favour fire spread. In July 2012, the typical sea breeze circulation affected the fire evolution, whereas the intense downslope winds favoured the fire spread in June 2020. The study brings useful guidelines for interpreting the impact of different mesoscale environments that may produce large fires, namely the orographic effects that can increase the fire susceptibility and vulnerability of some regions. This study was funded by national funds through FCT-Foundation for Science and Technology, I.P. under the PyroC.pt project (Ref. PCIF/MPG/0175/2019).