Advances in field observations, remote sensing, reanalysis products, and high-resolution modelling now provide unprecedented access to meteorological and environmental data. These datasets create new opportunities to study wildfires. A key challenge is combining these diverse data sources into modelling frameworks that support reliable wildfire prediction. Turning data into skilful, actionable forecasts requires ongoing innovation in numerical modelling, statistical methods, and machine learning.
Extreme wildfire events deserve special focus. In these cases, large fire plumes generate their own weather and alter the atmospheric boundary layer. The formation of pyrocumulus and pyrocumulonimbus clouds can dramatically accelerate fire spread. Such fires exhibit new behaviour that we are only beginning to understand and that is not yet captured well in fire spread models.
This session brings together researchers and fire analysts working at the intersection of weather, climate, and wildfire science. We encourage contributions from communities in high-resolution modelling, numerical weather prediction, climate modelling, Earth observation, and climate services, with a focus on understanding fundamental physics, as well as improving how fire-relevant processes and uncertainties are represented in forecasts and projections.
Topics include, but are not limited to:
• Observational and modelling approaches to extreme wildfire dynamics and
• pyroconvective fire spread
• Fire-weather and fire-climate relationships, including extremes and compound events
• Statistical, dynamical, and machine-learning methods for predicting wildfire occurrence and spread
• Integration of satellite data, reanalysis, and in situ observations into forecasting
• frameworks
• Uncertainty assessment of wildfire-relevant variables in weather and climate models
• Transdisciplinary research involving collaboration with operational firefighters
• Applications in climate services, early-warning systems, and risk-based decision support
In a changing climate, we see an increase in weather-related impact on society for a range of different types of impact. Moreover, it was hypothesized by First Responders that we would not only see an increase in risk, but also an increase in risk correlation. When risk correlation increases, we see an increasing number of days per year on which different types of impact show peak risk on the same day.
In the collaboration between First Responders and the Royal Netherlands Meteorological Institute (KNMI) we use quantitative AI/ML methods to build a model of the statistical relationship between weather data and three types of impact data in The Netherlands. However, this is not simply another success story of AI/ML. Rather, we see that the essential aspect is the collaboration between First Responders and the meteorological institute. Impact forecasting is not just about combining data. It is about building trust and communication between specialists, in this particular case as a co-design process between researchers, developers and field-experts.
The three types of impact we have considered are: (i) number of wildfire calls that the Fire Service responds to; (ii) number of Police priority dispatches (i.e. when a patrol car turns on the blue lights); and (iii) response time of Police priority dispatches. These numbers are aggregated over the Netherlands. We have then projected this model on the weather of the past, as well as on four different future climate scenarios. The results show a quite consistent increase in risk and in risk correlation. This implies that we will not only see higher demand for first response services, but are also moving more and more into multi-impact situations, which put higher demands on communication and coordination within and between different branches of first response service providers.
Working on wildfires has been a true and inspiring catalyst for this approach, and is therefore our main focus of this presentation. However, this topic does not live in a vacuum, and we will highlight how wildfire risk is increasingly correlated with other types of impact. The risk correlation hypothesis, as well as its confirmation, provides the rationale for a shift in how we approach collaborative quantitative impact forecasting. We see an immediate need for a shift from a single-hazard single-impact approach to a multi-hazard multi-impact approach.
In addition, given the ‘duty of care’ that the Netherlands Met Office has, we feel the ethical and institutional obligation to not only do impact forecasting research, but also to take this research into operation. As this is an ongoing process, a demonstration of the most current minimum viable product (MVP) will be part of the presentation.
How to cite: de Baar, J., Alfonsi, A., van der Schrier, G., Kok, E., and Verhoeven, B.: The future is now: the need for operational multi-hazard multi-impact collaborative quantitative impact forecasting, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-135, https://doi.org/10.5194/ems2026-135, 2026.
Daily wildfire risk has been monitored in Slovakia since 2002 and it is legally binding within the fire prevention. The original methodology needed an update to improve the accuracy and to bring more robust information about wildfire risk in different parts of country with high diversity of a terrain and land use. Therefore, new wildfire risk forecasting products, which were developed under the project Clim4Cast, were adopted, customized and implemented into existing national platform. The products consist of following indicators: Fire Weather Index (FWI), Fuel Moisture 1H, 10H, 100H and 1000H. The original Clim4Cast wildfire forecasting products are computed using deterministic ECMWF model with the spatial resolution of 9 km. The adopted Slovak products are based on ALADIN NWP model with 2 km resolution (for the first three days, including the day of calculation) and coarser ensemble mean of ECMFW IFS model (day 4 – 7).
As the wildfire risk monitoring products are legally binding for the Firefighting and Rescuing Corps of the Slovak Republic (HaZZ SR), it was necessary to understand the uncertainty of new forecasting products. Therefore, two types of validation assessment was proceeded. The first validation based on the expert assessment (qualitative approach) was done during the wildfire season 2025. This testing of the new products was conducted in selected NUTS4 regions in the co-operation with HaZZ SR. The experts received on daily basis updated forecasting maps of the wildfire risk and evaluated their accuracy with the conditions in their region. The expert assessment as a qualitative validation approach is used on weekly basis within the soil drought monitoring (Intersucho) and shows well reliability on expert assessment. Further, this testing helped us to gather the feedback by the end users and to customize the products for the specific needs of the firefighters.
The second part of validation was based on observed data at 101 meteorological stations in Slovakia. We chose two case study periods – summer 2025 and spring 2026. Both study periods were characterised by diverse weather conditions (dynamically changing weather with convective precipitation vs. long-lasting stable dry and warm weather), which enabled us to test the performance of the forecasting products under different weather conditions and to reach more robust information about uncertainty, which needs to be communicated to the end users, especially firefighters. The validation on objective data showed well accuracy and reliability of the new products. However, the uncertainty rises in the case of forecasted convective precipitation.
Acknowledgment
This work was supported by project Clim4Cast (Central European Alliance for Increasing Climate Change Resilience to Combined Consequences of Drought, Heatwave, and Fire Weather through Regionally-Tuned Forecasting; CE0100059) co-funded by European Union funds (ERDF – Interreg Central Europe).
How to cite: Labudová, L., Holec, J., Štefánik, D., Bálek, J., Ivaňáková, G., Krčová, I., and Mikulová, K.: New gridded daily wildfire risk products for Slovakia and their validation using observed data and expert assessment, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-269, https://doi.org/10.5194/ems2026-269, 2026.
Recent advances in machine learning (ML) are transforming scientific applications, including weather and hazard prediction. In the context of wildfires, these methods enable a fundamental shift from forecasting fire weather conditions to predicting actual fire activity. In this study, we demonstrate that such a transition is not only feasible but can also be implemented in an operational forecasting framework.
Traditional fire danger indices, such as the Fire Weather Index (FWI), often overestimate risk, particularly in fuel-limited ecosystems, resulting in high false-alarm rates. By contrast, our data-driven approach, the probability of fire, integrates information on weather, fuel characteristics, ignitions, and observed fire activity to directly predict the probability of fire occurrence. This leads to substantial improvements in forecast reliability, reducing false alarms while maintaining sensitivity to high-risk conditions.
We show that model performance is driven more by the quality and completeness of input data than by the complexity of the ML architecture itself. In particular, fuel status emerges as the most critical predictor of fire activity. However, the lack of direct, real-time global observations of fuel remains a major limitation. To address this, we rely on physically based models to reconstruct fuel dynamics, highlighting the continued importance of process-based understanding in supporting ML applications.
Our results demonstrate that incorporating all components of the fire triangle, weather, fuel, and ignitions, can improve predictive skill by up to 30% compared to weather-only approaches. Furthermore, the probabilistic nature of the predictions enables direct verification against observed fire activity and opens new opportunities, such as reconstructing missing satellite detections and improving fire emission estimates.
Overall, this work underscores that meaningful progress in ML-based fire forecasting depends not only on algorithmic innovation, but critically on investment in high-quality, physically consistent datasets. Without this, the potential of data-driven approaches cannot be fully realised.
How to cite: Di Giuseppe, F., McNorton, J., and Wetterhall, F.: Data matters most: improving fire activity forecasts through data-driven approaches, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-323, https://doi.org/10.5194/ems2026-323, 2026.
Europe is entering new wildfire regimes, with more frequent extreme events capable of generating strong fire–atmosphere interactions and, in some cases, their own local weather. These events expose important gaps in current early-warning, monitoring, and operational support systems, especially when rapid decisions are needed to characterize fire behaviour and organize targeted observations. To address these scientific and operational challenges, we developed the Southern Europe Biomass Burning Experiment (EUBURN), a meteorological airborne field programme designed to improve the observation, understanding, and prediction of extreme fire–atmosphere interactions in southern Europe. Beyond data collection, the programme also provides a framework for testing and assessing operational tools under realistic field conditions.
Within this context, and to support flight planning during the SILEX campaign, we implemented a prototype end-to-end chain combining near-real-time satellite fire detection, fire-event tracking, and coupled fire–atmosphere forecasting. New fire detections from Meteosat were processed operationally to identify emerging events, which were then used to trigger and initialize coupled ForeFire–MesoNH simulations. These simulations provided rapid short-term guidance on expected fire evolution, plume direction, plume-top height, smoke production, and the potential for extreme propagation or pyroconvective development.
The prototype was deployed operationally in France during summer 2025 and tested during the three-week SILEX campaign in July 2025. Over this period, more than 400 forecast runs were launched automatically following satellite detections, with a direct feedback loop to support scientific aircraft operations. The results show that such a system can deliver useful early information for airborne decision support, in particular for anticipating plume orientation, altitude, and fire development shortly after detection. This first real-scale deployment demonstrates the feasibility of integrating near-real-time detection and coupled fire–atmosphere forecasting into an operational workflow for wildfire observation campaigns, while also highlighting the need for broader benchmark datasets and quantitative evaluation for full forecasting-system assessment.
How to cite: Filippi, J.-B., Denjean, C., and Paugam, R.: Operationalizing Coupled Fire-Atmosphere Forecasting at National Scale: Insights from the EUBURN Programme and SILEX Campaign, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-390, https://doi.org/10.5194/ems2026-390, 2026.
The fires in İzmir that occured in June 2025 and resulted in widespread evacuations and extensive ecological damage, showed that forest fires are becoming more frequent in Türkiye's Aegean and Mediterranean coastal regions, seriously damaging local ecosystems, socioeconomic activity and vital infrastructure. Therefore accurately predicting regional fire hazard using meteorological data has become critically important as climate change intensifies these vulnerabilities. This study uses two methods: long-term climate risk projections using the Fire Weather Index (FWI) and the development of a high-precision, artificial intelligence-based Decision Support System (DSS). Future wildfire dangers (2026–2096) under RCP 2.6, 4.5, and 8.5 scenarios were projected using high-resolution EURO-CORDEX regional climate models and compared to a historical reference period (1971–2005). Quantitative analysis of severe risk thresholds (FWI > 45) reveals a profound geographical expansion of fire hazards. While the historical average stood at 50.48 extreme-risk days, projections indicate an increase to 55.22 (+9.4%) under RCP 2.6 and 61.71 (+22.2%) under the pessimistic RCP 8.5 scenario. Crucially, coastal hotspots are expected to endure up to 234.92 extreme-risk days annually under RCP 8.5. In this projection, the traditional summer fire season shifts into a nearly constant hazard lasting approximately 65% of the year, demanding immediate proactive adaptation strategies. The next stage of this research presents a proactive early warning strategy using the Random Forest machine learning method to handle these extended fire seasons. The system synchronizes dynamic ERA5 meteorological variables with 30m high-resolution National Aeronautics and Space Administration Shuttle Radar Topography Mission (NASA SRTM) topography data (elevation, slope, and aspect). This integration is achieved through bilinear interpolation into a unified 1-km spatial grid. Trained on 199,606 balanced instances from National Aeronautics and Space Administration Fire Information for Resource Management System (NASA FIRMS) (2010–2020), the model’s dependability was rigorously assessed through an out-of-time validation using 34,265 data points from the extreme fire year of 2021. The DSS achieved an outstanding overall accuracy of 95.81%, successfully detecting 15,709 genuine fire events while maintaining a remarkably low false positive count of only 49, effectively eliminating false alarms that strain public resources. Feature relevance ratings identified surface temperature (0.30 weight) and elevation (0.25) as the primary drivers of fire vulnerability. Ultimately, this research serves as a foundational framework for evidence-based policy-making and strategic land-use planning. By identifying non-linear risk patterns and producing real-time, high-resolution vulnerability maps, the DSS enables forestry and emergency directorates to shift from reactive firefighting to proactive governance. This model provides the scientific justification for modernizing national fire-fighting protocols and optimizing the strategic allocation of aircraft and ground resources, ensuring that mitigation policies are dynamically aligned with the emerging reality of a continuous fire season.
How to cite: Aktaş, C. and Özdemir, E. T.: A Random Forest-Based Early Warning System for Wildfire Risk in the Southern Aegean and Western Mediterranean Regions of Türkiye, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-399, https://doi.org/10.5194/ems2026-399, 2026.
Wildfire spread in complex terrain is strongly modulated by fine-scale atmospheric processes that remain poorly represented in operational models. Downslope windstorms are among the most critical drivers of extreme fire behavior, producing hot, dry, and gusty conditions that can rapidly accelerate fire spread. However, their inherently transient and spatially heterogeneous structure poses a major challenge for wildfire prediction systems, particularly in coastal environments where interactions with marine boundary layers further complicate the flow.
This study leverages observations from the Sundowner Winds EXperiment (SWEX, April–May 2022), conducted in Santa Barbara, California, together with high-resolution numerical simulations to investigate the variability of lee-slope jets and hydraulic jumps during wildfire-prone downslope windstorms in Southern California. SWEX provides a unique dataset of in-situ and remote sensing observations, including Doppler lidars and dropsondes, capturing boundary-layer structure and mountain wave dynamics during multiple downslope windstorm events. We focus on a representative event (IOP10) characterized by non-stationary mountain waves and strong spatial variability in downslope winds.
We evaluate the ability of the Weather Research and Forecasting (WRF) model to represent these processes at kilometer-scale resolution and at sub-kilometer Large-Eddy Simulation (LES) scales. While both configurations reproduce key features of lee-slope jets over the mountain slopes, the kilometer-scale simulation overestimates the spatial extent and persistence of strong winds farther downslope. In contrast, sub-kilometer simulations resolve transient mountain wave behavior and associated hydraulic jump variability, more closely resembling the observations. These differences highlight the importance of resolving unsteady wave dynamics and boundary-layer interactions, including coupling with the shallow, stably stratified marine layer. We then use the WRF-FIRE model to understand how hydraulic jump dynamics influence simulated wildfire spread.
Our results demonstrate that wildfire-relevant wind fields are highly sensitive to model resolution and terrain representation. The inability of coarser models to capture transient flow features can lead to systematic biases in near-surface winds that are critical for fire spread prediction. Simplified fire spread models often rely on static or parameterized surface wind inputs and cannot capture the dynamic atmosphere-fire interactions that occur during downslope windstorms. These findings underscore the need to integrate high-resolution modeling, targeted observations, and improved physical parameterizations into wildfire forecasting frameworks.
How to cite: Duine, G.-J., Carvalho, L., De Wekker, S., Jones, C., Seto, D., De Orla-Barile, M., Modjeski, G., Brown, W., Clements, C., Fernando, H., and Desai, J.: The Role of Lee-Slope Jets and Hydraulic Jumps in Wildfire Behavior: Insights from Observations and Modeling , EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-405, https://doi.org/10.5194/ems2026-405, 2026.
In recent years, wildfires have increasingly impacted southern Europe. Although fire is a natural component of Mediterranean ecosystems, the expansion of recreational use of natural and forest areas has increased the number of human-caused ignitions. Climate change further exacerbates this situation by intensifying extreme temperatures and droughts, thereby altering two of the three primary drivers of wildfires: fuel and weather. The Canadian Forest Fire Weather Index (FWI) is a meteorologically based index widely adopted to estimate fire danger. It requires only temperature, wind speed, relative humidity, and precipitation as input, and consists of six components: three fuel-moisture codes and three fire behavior indices. While originally developed for Canadian boreal conditions, the FWI has been successfully applied in many countries, including southern Europe, where studies have demonstrated its ability to capture fire danger in Mediterranean environments, though further evaluation has been recommended, especially in drier landscapes. In this work, we simulate a wildfire in Italy and treat it as a test case to investigate the effects of integrating the FWI into a coupled fire-atmosphere model WRF-SFIRE. WRF-SFIRE leverages an integrated fuel moisture model to account for the influence of spatial and temporal variability of fuel flammability on fire behavior. We compare simulations executed using two approaches: (i) static fuel moisture initialization, and (ii) dynamic fuel moisture modeling. The comparison between these simulations and observational data shows that dynamic fuel moisture modeling improves the fidelity of both the simulated burned area and the meteorological variables, highlighting the importance of accounting for fuel conditions in operational fire modeling. By default, the fuel moisture model simulates the evolution of dead fuel moisture contents according to a time-lag differential equation, and requires a spin-up phase before the fire event for fuel preconditioning. Alternatively, it can be initialized with external operational fuel moisture data, bypassing this requirement - but such observations are sparse and difficult to obtain routinely. In principle, fuel moisture contents can be estimated from FWI codes using empirical relationships. However, a comparison between FWI-based fuel moisture estimates and a dead fuel moisture reanalysis dataset for California reveals that the FWI-based values systematically overestimate the 100-hr fuel moisture content, consistent with findings reported in the literature. This indicates that the empirical relationship linking the Duff Moisture Code to fuel moisture content requires recalibration before it can be reliably applied. To address this, we developed a framework to derive fuel moisture estimates from FWI codes, enabling WRF-SFIRE initialization without the need for in-situ observations or long model spin-up times, and relying only on routinely available fire danger indices. This approach has the potential to enhance the operational applicability of WRF-SFIRE in data-sparse regions, supporting more timely and accurate fire risk assessment across Mediterranean Europe.
How to cite: Rinaldi, S., Kochanski, A., Clements, C. B., Di Sabatino, S., and Leo, L. S.: The Impact of Fuel Moisture Initialization on WRF-SFIRE Simulations of Mediterranean Wildfires, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-465, https://doi.org/10.5194/ems2026-465, 2026.
Wildfires are increasingly showing pyro-cloud formation, thereby transitioning into a state with unpredictable fire behaviour. Although a multitude of factors determine the ability of any plume to create pyro-clouds, one of the essential components is the ability of dry convective plumes to reach altitudes where condensation can occur. Hence, to predict the onset of pyro-cloud formation, it is essential to understand what factors govern the plume top height of dry convective plumes.
However, previous research predominantly focused on the injection height to improve smoke pollution predictions. Our research extends beyond the plume injection height, investigating the physics that govern the plume top height. As observations of convective wildfires are scarce due to the dangerous measurement conditions, we use MicroHH to create high-resolution (~20 m) large eddy simulations of convective wildfire plumes. Our preliminary analyses, in which we varied the fire intensity, revealed three distinct plume regimes dictated by the interaction between heating by the fire and atmospheric stratification:
- ABL-Plumes: plumes that cannot overshoot the atmospheric boundary layer (ABL), resulting in a plume top equal to the ABL top.
- Overshooting Plumes: Plumes that overshoot into the free troposphere, but with an injection height equal to the ABL top.
- Free Tropospheric Plumes: Plumes that both overshoot and inject into the free troposphere, meaning that both the plume top height and the injection height exceed the ABL top.
With our study, we aim to answer two questions. First, how do the scaling relationships for maximum plume top height evolve as a plume transitions from an Overshooting to a Free Tropospheric plume? Second, for the Free Tropospheric plumes, do the plume top height and injection height share the same scaling relationships, or do distinct physical processes govern the plume top height?
For example, we know from a previous study that the injection height of free tropospheric plumes scales with fire intensity to the power of 0.35. If both the plume top height and injection height follow the same scaling, this provides a unified physical explanation for plume rise. Alternatively, a different scaling suggests that additional physical processes govern the overshoot beyond the injection height. To explore these physics, we will vary the fire intensity across a range of realistic atmospheric conditions by modifying the boundary-layer height, ambient wind speed, capping inversion strength, and free-tropospheric lapse rate. Ultimately, our goal is to condense the plume dynamics derived from our 3D large eddy simulations into a simplified (adiabatic) parcel model to explain the scaling behaviours and regime transitions of dry convective wildfire plumes.
How to cite: Roelofs, T., Janssens, M., Vilà-Guerau de Arellano, J., and Van Heerwaarden, C.: Beyond the injection height: Understanding the plume top height behaviour of wildfire-induced plumes. , EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-492, https://doi.org/10.5194/ems2026-492, 2026.
The Mediterranean is a climate change hotspot where increasing temperatures and changing hydroclimate are intensifying wildfire risk. In the summer of 2025, the fire season was extremely destructive, with human casualties, homes destroyed, and large areas of forest and agricultural land burned. While previous studies have primarily relied on probabilistic or trend-based approaches to understand how climate change affects fire weather conducive to such fire seasons, process-based conditional attribution remains limited. Here, we apply a spectrally nudged storyline approach to conditionally attribute the thermodynamic contribution of climate change to the Fire Weather Index (FWI) in the Mediterranean during July–August 2025.
We find a robust increase in fire weather across the Mediterranean under climate change, with higher FWI values in all regions when comparing present-day to pre-industrial conditions. This increase becomes even more pronounced under +3 and +4 °C warming levels, although with greater regional variability in robustness. The strongest and most spatially coherent signal is driven by temperature increases, while reductions in relative humidity and precipitation further amplify fire-conducive conditions in specific regions. The magnitude and robustness of the signal vary spatially, with clearer responses in western Mediterranean countries and more variability in regions with low precipitation.
These results demonstrate that climate change has intensified the atmospheric conditions conducive to wildfires during the Mediterranean summer of 2025. More broadly, this study highlights the value of conditional storyline attribution for providing an event-based understanding of extreme events. As an intermediate metric, the Fire Weather Index offers a bridge between atmospheric conditions and wildfire impacts, suggesting a pathway to extend attribution frameworks towards impact-based assessments.
How to cite: van Garderen, L., León-FonFay, D., Dietz, V., van de Berg, W. J., and Feser, F.: Storyline Attribution of the 2025 Mediterranean Summer Fire Weather, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-517, https://doi.org/10.5194/ems2026-517, 2026.
Wildfires represent a growing challenge for weather and climate science, with significant impacts on ecosystems, air quality, infrastructure, and human safety. In the Basque Country (CAE), wildfire ignition is predominantly driven by human activity, but understanding spatial and temporal patterns provides crucial insight into the interaction between environmental conditions and anthropogenic behavior. Characterizing these patterns supports predictive modelling, early-warning systems, and risk management strategies under both current and changing climate conditions.
Wildfires in the CAE show complex spatial and temporal patterns driven by both environmental and human factors. This study analyzes a dataset of 5,053 wildfire incidents recorded between 1995 and 2022, including small fires (<1 ha), medium fires (1–50 ha), and large fires (>50 ha), providing a comprehensive overview of wildfire dynamics over more than two decades.
Results indicate that small fires are the most frequent type of incident (54.9% of all events) but contribute minimally to the total burned area (3.8%), while medium (43.7% of events; 51.9% of burned area) and large fires (1.4% of events; 44.2% of burned area) dominate landscape-level impacts. Monthly patterns reveal that fire frequency peaks in March (829 events), whereas burned area reaches its maximum in February (~3,200 ha). Spatial analysis across historical territories shows broadly similar patterns, with small fires dominating in all provinces (52–58%). However, Araba exhibits a slightly higher proportion of small fires (57.7%), while Bizkaia shows a greater relative contribution of medium fires (47.0%), and Gipuzkoa presents an intermediate distribution.
Temporal analysis reveals clear weekly and hourly patterns: ignitions peak during weekends and during the late afternoon, with a maximum around 16:00, reflecting the predominance of human-driven fires. Most fires (98.3%) are controlled within 24 hours, although large fires show significantly longer durations (mean 27.3 hours) compared to medium (5.2 hours) and small fires (2.6 hours). Fire size distributions indicate that a very small proportion of large events contributes disproportionately to total burned area, consistent with heavy-tailed behavior observed in other temperate fire regimes.
These results provide an updated and comprehensive overview of wildfire dynamics in the Basque region, corroborating patterns strongly influenced by human activity. Spatial and temporal variability further highlights the value of territory-specific management strategies
How to cite: Gaztelumendi, S.: Spatial and Temporal Patterns of Wildfires in the Basque Country (1995-2022), EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-593, https://doi.org/10.5194/ems2026-593, 2026.
Here we present a simple entraining‑plume model designed to represent the essential dynamic and thermodynamic behaviour that develops above an active wildfire. The model combines widely used theory describing how wildfire heat is injected into the atmosphere, with plume parameterisations commonly employed in weather and climate models, into a new representation of convection above a wildfire. We tune the model’s entrainment and detrainment rates to align with direct numerical simulation experiments, and we validate its performance using in‑situ observations collected across a broad range of wildfires (through surface rates of spread and fuel models) and their associated plumes (through in-plume and environmental radiosondes). In this presentation, we describe how a simple model of this kind can support predictions of regime transitions during wildfire events, for example by indicating the potential onset of moist pyroconvection. We also outline how the model can serve as an instructional tool for fire analysts or other practitioners: It allows them to explore fundamental plume processes by testing sensitivities to a small set of surface and atmospheric control parameters, such as fuel moisture content, fire size, fire intensity, free tropospheric stability, inversions, boundary-layer and free-tropospheric humidity, and boundary layer height. Ultimately, despite its deliberate simplicity, we conclude that the model retains enough physical realism to provide meaningful insight into wildfire plumes. As such, we believe that in an age of increasingly complex, data-driven models, simple tools such as this one remain a valuable resource for both wildfire analysis in the field, but especially for educational activities.
How to cite: Janssens, M., Castellnou Ribau, M., Bachfischer, M., Miralles Bover, M., Verhoeven, B., Roelofs, T., Eyken, J., Van Heerwaarden, C., and Vila, J.: The didactic and practical value of a simple wildfire plume model, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-562, https://doi.org/10.5194/ems2026-562, 2026.
Extreme wildfires are strongly modulated by their interaction with the atmosphere, yet this coupling is rarely accounted for in operational wildfire management. The Extreme Wildfire Event Data Hub for Improved Decision Making (EWED) project brought together firefighters and atmospheric scientists from four European countries to bridge this gap: raising awareness of wildfire-atmosphere coupling and translating scientific understanding into tools that improve the management and suppression of extreme wildfires. In this contribution, we present the lessons learned from EWED.
Central to the EWED approach is the deployment of balloon soundings into and near wildfire plumes, a novel observation strategy that provides decision makers with information on the vertical structure of the atmosphere. Such profiles are critical for anticipating dramatic changes in fire behaviour and rate of spread, and simultaneously serve as a unique dataset for advancing plume science. During 2024 and 2025, firefighting teams across the project collected balloon observations during live wildfire events. These data are made publicly available through the Wildfire Data Portal (wildfiredataportal.eu), a platform developed within EWED to serve both the research and operational communities.
On the modelling side, EWED has advanced the understanding of wildfire plume dynamics through two complementary lines of work. First, three-dimensional simulations informed by the balloon soundings have revealed important new insights into near-surface inflow patterns and circulations downstream of the fire head. Second, coupled fire-spread simulations operating at the turbulence time scale reproduce poorly understood fire-spread acceleration as an emergent effect of deep plume formation, offering a promising pathway to address the rate-of-speed underestimation commonly observed in extreme wildfires.
Building on these observations and simulations, we developed a conceptual weather model combined with an entraining plume model that enables firefighters to estimate plume rise as a function of fire properties and the vertical thermodynamic structure of the lower atmosphere. By incorporating forecast atmospheric boundary layer profiles a few hours ahead of time, the model allows decision makers to anticipate changes in plume behaviour before they materialise. The potential of combining balloon observations with these modelling tools was demonstrated during a three-day training event for operational firefighters.
By revealing the potential for substantial safety improvements, EWED has provided a roadmap for better integration of meteorological information into wildfire firefighting. This integration will be carried forward and deepened in the recently started EU-funded Open Decision-making system for enhancing Europe's preparedness and response capacities to Extreme wildfires (ODET) project.
How to cite: van Heerwaarden, C., Castellnou, M., Roelofs, T., Vilà-Guera de Arellano, J., Janssens, M., Verhoeven, B., Ntasiou, Z., Stokkeland, O., Troncho, J., Guarque, P., Miralles, M., Estevil, L., Ruiz, B., Pages, J., and Prat, N.: Lessons learned from the EWED project: integrating atmospheric observations and modelling into extreme wildfire decision making, EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-790, https://doi.org/10.5194/ems2026-790, 2026.
Understanding the drivers of wildfire ignition is essential for improving predictive models and early-warning systems, particularly in regions with high human activity such as the Basque Country. While natural factors like lightning and drought contribute to fire risk, anthropogenic causes dominate, shaping both the frequency and the area burned. Detailed knowledge of human-driven ignition, motivations, and seasonal patterns is critical to link historical wildfire behavior with environmental and meteorological conditions, and to improve strategies for risk management under changing climate scenarios.
The causes and motivations behind wildfires are central to effective prevention and management strategies. In the Basque Country (CAE), human activity is the dominant driver of wildfire ignitions, accounting for more than 80% of all incidents between 1995 and 2022, while natural causes such as lightning contribute less than 2% of events. This study examines a long-term dataset of 4,054 wildfire incidents, providing a detailed characterization of ignition causes, motivations, seasonal variability, and territorial patterns. Analysis of human-caused fires reveals that negligence, including uncontrolled agricultural burns, disposal of yard waste, and mishandling of fires, is the most prevalent source, followed by deliberate ignitions, accidents, and infrastructure-related causes such as electrical lines or machinery use. Among intentional fires, 48% have unknown motivations, while identified motivations include agricultural or livestock practices, political acts, pyromania, and land-use changes.
Seasonal analysis shows that negligent fires peak during the spring and summer months, coinciding with high fire risk, whereas deliberate fires are more common in spring and winter and natural fires are strongly concentrated in summer. This pattern varies across historical territories: Araba exhibits a stronger winter intentional-fire peak, while Bizkaia and Gipuzkoa show less pronounced seasonal patterns. In terms of burned area, more than 83% of total hectares affected were caused by human activity, highlighting the disproportionate influence of anthropogenic factors on fire severity. Fires attributed to negligence account for the majority of burned area in small to medium events, while intentional fires dominate in large-scale events. Temporal trends indicate a reduction in unknown causes over recent decades, reflecting improved reporting and monitoring practices.
These facts highlight the importance of integrating anthropogenic drivers into wildfire predictive models alongside meteorological and climate variables, as historical fire data can support the calibration of statistical, machine-learning, and numerical fire-spread models, improving their ability to forecast both ignition probability and burned area. Recognizing spatial, seasonal, and motivational patterns is key to developing targeted prevention measures and early-warning systems that incorporate human behavior as a central component of wildfire risk.
How to cite: Gaztelumendi, S.: Human and Natural Drivers of Wildfires in the Basque Country: Causes and Motivations (1995–2022), EMS Annual Meeting 2026, Utrecht, Netherlands, 6–11 Sep 2026, EMS2026-586, https://doi.org/10.5194/ems2026-586, 2026.
Supporters & sponsors
