Abstract. Extreme hydro-meteorological events can have a substantial impact on vegetation and ecosystems. In particular, with heatwaves and droughts projected to become more frequent due to climate change, understanding their effects on forests is crucial. In this study, we present a novel, large-scale, spatially explicit analysis of forest browning drivers across Europe, using a homogeneous and automated random forest modeling framework. By running independent models at each 0.5° grid point, we enable a region-specific comparison of hydro-meteorological drivers, capturing the diversity of forest responses across the continent. We identify the most relevant hydro-meteorological predictors of low normalized difference vegetation index (NDVI) events at monthly to annual timescales, using NDVI data from the Advanced Very High Resolution Radiometers (AVHRR) and climate variables from ERA5 and ERA5-Land reanalyses. These predictors include maximum temperature at 2 m precipitation, surface latent heat flux, and soil moisture up to 18 months before the observed browning. The random forest model exhibits a high prediction skill over most grid points in Europe, with a critical success index greater than 0.75 for 65 % of grid points. Notably, warm and dry conditions in spring and early summer emerge as essential predictors. We also uncover multi-year influences, with soil moisture and temperature anomalies from the preceding year playing a significant role, especially in Scandinavia and for coniferous forests. The random forest approach further reveals non-linear relationships, such as both positive and negative precipitation anomalies at different lags contributing to browning risk.
Rivoire et al. (2026) studied this question.