Towards high-resolution prediction of drought effects on Switzerland’s Beech forests for improved management

Abstract

Over the past decade, extreme temperature and drought have resulted in widespread early leaf discoloration in European Beech (Fagus sylvatica) forests across central Europe. Discoloration during the particularly hot and dry summer of 2018 was ultimately associated with increased rates of crown dieback and tree mortality. Given the trend towards hotter and drier growing seasons under a changing climate, there is an increasing demand for site-specific recommendations on drought-resilient forest management practices in Switzerland. Making these recommendations requires a robust understanding of empirical forest disturbance and estimates of future forest health under a range of climatic and management conditions. To that end, using 2018 field observations, manual mapping of forest discoloration in aerial imagery, and multispectral Sentinel-2 imagery, we are developing 10 m/pixel estimates of European Beech discoloration across Switzerland during the 2018 to 2023 foliated periods. To date, we have 1) developed a robust interpolated Sentinel-2 time series from 2018 to 2023 for all of Switzerland, 2) trained a random forest model using 2018 ground control data and several vegetation indices from the Sentinel-2 time series to predict 2018 early leaf discoloration across Switzerland’s European Beech forests with c. 90% accuracy and, 3) used the Chlorophyll Red-Edge Index derived from the Sentinel-2 time series to approximate tree phenology and the length of the foliated period. We estimate that the 2018 foliated period was, on average, 45±19 days shorter for discolored sites as compared to sites without discoloration. Our results generally align well with previous studies of the 2018 drought in Switzerland and additional observational data is being compiled to validate the application of 2018 ground truth data across the foliated periods from 2018 to 2023. In combination with high-resolution soil maps, meteorological data, topographic derivatives, and information on Swiss forest structure, we will use empirical discoloration estimates to train ensemble models of site-specific susceptibility to drought. By artificially varying the meteorological and forest structure variables in these models we will have the unique opportunity to better understand European Beech susceptibility to drought and test the influence of a range of future climate scenarios and forest management strategies on Swiss forest health at a high spatial resolution.