Development of a statistical model for global burned area simulation within a DGVM-compatible framework

Abstract. Fire-enabled Dynamic Global Vegetation Models (DGVMs) play an essential role in predicting vegetation dynamics and biogeochemical cycles amid climate change, but modelling wildfires has been challenging in process-based biophysics-oriented DGVMs, regarding the role of socioeconomic drivers. In this study, we aimed to build a simple global statistical model that incorporates socioeconomic drivers of wildfire dynamics, together with biophysical drivers, within a DGVM-compatible framework. Using monthly burnt area (BA) data from the latest global burned area product from GFED5 as our response variable, we developed Generalized Linear Models to capture the relationships between potential predictor variables (biophysical and socio-economic) that are simulated by DGVMs and/or available in future scenarios. We used predictors that represent aspects of fire weather, vegetation structure and activity, human land use and behavior and topography. Based on an iterative process of choosing various variable combinations that represent potential key drivers of wildfires, we chose a model with minimum collinearity and maximum model performance in terms of reproducing observations. Our results show that the best performing (deviance explained 56.8 %) and yet parsimonious model includes eight socio-economic and biophysical predictor variables encompassing the Fire Weather Index (FWI), Monthly Ecosystem Productivity Index (MEPI), Human Development Index (HDI), Population Density (PPN), Percentage Tree Cover (PTC), Percentage Non-Tree Cover (PNTC), Number of Dry Days (NDD), and Topographic Positioning Index (TPI). When keeping the other variables constant (partial residual plots), FWI, PTC, TPI and PNTC were positively related to BA, while MEPI, HDI, PPN, and NDD were negatively related to BA. While the model effectively predicted the spatial distribution of BA (Normalized Mean Error = 0.72), its standout performance lay in capturing the seasonal variability, especially in regions often characterized by distinct wet and dry seasons, notably southern Africa (R2= 0.72 to 0.99), Australia (R2= 0.68) and South America (R2= 0.75 to 0.90). Our model reveals the robust predictive power of fire weather and vegetation dynamics emerging as reliable predictors of these seasonal global fire patterns. Finally, simulations with and without dynamically changing HDI revealed HDI as an important driver of the observed global decline in BA.