<p>Wildfires are recurrent natural hazards that affect terrestrial ecosystems, the carbon cycle, climate and society. An ignition can lead to a wildfire when there is biomass available for burning, typically in combination with dry and windy conditions. Wildfires are regarded as compound events defined as &#8220;an extreme impact that depends on multiple statistically dependent variables or events&#8221; [1], and dominant drivers include a combination of various meteorological, hydrological and biological conditions. More specifically, wildfires can be regarded preconditioned hazards [2] because the combination of drivers can cause the hazard only in the presence of available and burnable biomass (precondition). The availability of burnable biomass is itself driven by conditions such as soil moisture, temperature, humidity, precipitation, etc. Identifying a selection of dominant controls and their statistical dependence, can ultimately improve predictions and projections of wildfires in both current and future climate. In this study, we apply a data-driven bottom-up statistical learning approach (including random forest and logistic regression) to identify dominant factors determining burned area over northern Europe. Potential explanatory variables include temperature, precipitation, wind, soil moisture and vegetation cover, as well as meteorological drought, soil moisture drought and greenness indices. A monthly 2001-2020 burned area product derived from satellite observations is used as target variable, and multiple hydrometeorological and vegetation metrics stemming from the ERA5 reanalysis and observational datasets (e.g. EOBS) are tested as potential predictors. The derived relationships between wildfires and its compound drivers will further be used to assess the potential changes in such a combination of factors under different climate scenarios using large-ensemble global climate simulations and hydrological models. This new framework will allow us to better quantify the changes in potential wildfire risk in a changing climate using a combination of data driven and physically based models.</p><p>[1] Leonard et al., 2014:&#160;https://doi.org/10.1002/wcc.252<br>[2] Zscheischler et al., 2020: https://doi.org/10.1038/s43017-020-0060-z</p>
Incorporating Sentinel-1 SAR imagery with the MODIS MCD64A1 burned area product to improve burn date estimates and reduce burn date uncertainty in wildland fire mapping
