A hybrid framework to estimate live fuel moisture content through land surface modelling

Wildfires are among the most powerful natural phenomena that have severe environmental and social impacts. Their increasing frequency and intensity, driven by climate change and unsustainable land-use practices, emphasise the importance of reliable fire risk assessment tools. Live fuel moisture content (LFMC) is a key variable for wildfire risk assessment and management, as it directly influences fire behaviour and propagation. However, retrieving this information is challenging owing to its high spatial and temporal variability, combined with time-consuming and expensive in situ sampling methods. This study introduces a novel approach to estimate the LFMC from a Land Surface model (LSM). This approach was achieved by integrating remote sensing, numerical modelling, and a machine learning (ML) framework. Land surface model simulations were conducted to generate surface and soil conditions, which served as predictors in an ML-based regressor model to estimate the LFMC. The proposed model demonstrated robust performance, with a coefficient of determination (r 2 ) of 0.72 and an absolute root mean square error (RMSE) of 11.6%. This approach produces reliable LFMC estimates with high spatial resolution, which can be used in wildfire propagation models. Finally, this study highlights a novel model to produce LFMC information, which can significantly enhance wildfire management and support more informed decisions in fire prevention and action strategies.