A framework for assessing rainfall extremes induced aboveground biomass loss from multisource remote sensing data

Increasingly frequent and intense rainfall extremes can trigger flooding, landslides, and severe soil erosion, leading to substantial losses of vegetation aboveground biomass (AGB). However, event-scale assessments of rainfall extremes induced AGB loss remain largely unexplored. Here, we propose a framework to estimate event-scale AGB loss induced by extreme rainfall events. The framework integrates high-resolution change detection techniques to identify vegetation loss sites with an Extreme Gradient Boosting (XGBoost)-based downscaling approach to derive sub-meter AGB from coarse-resolution AGB products. We evaluated the framework across three extreme rainfall events in China: Beijing (2023), Luochuan (2021), and Zixing (2024). The framework achieved high accuracy across all case studies, with F1 scores of 0.94, 0.92, and 0.94 for three events in identifying loss sites, as well as an R 2 value of 0.88 and RMSE of 4.44 Mg/ha for downscaled high-resolution AGB estimates. We further apply a Random Forest-based residual modeling to investigate model errors, indicating that terrain-related factors play a dominant role in shaping the variability of AGB estimation errors. The average AGB loss densities within identified sites are 75.21 Mg/ha, 67.08 Mg/ha, and 90.94 Mg/ha for three events, respectively. These findings highlight the framework’s potential in assessing event-scale AGB loss, providing a practical tool for ecological monitoring and disaster response under intensifying climate variability.