Large‐scale early‐wilting response of Central European forests to the 2018 extreme drought

Abstract The combination of drought and heat affects forest ecosystems by deteriorating the health of trees, which can lead to large‐scale die‐offs with consequences on biodiversity, the carbon cycle, and wood production. It is thus crucial to understand how drought events affect tree health and which factors determine forest susceptibility and resilience. We analyze the response of Central European forests to the 2018 summer drought with 10 × 10 m satellite observations. By associating time‐series statistics of the Normalized Difference Vegetation Index (NDVI) with visually classified observations of early wilting, we show that the drought led to early leaf‐shedding across 21,500 ± 2,800 km2, in particular in central and eastern Germany and in the Czech Republic. High temperatures and low precipitation, especially in August, mostly explained these large‐scale patterns, with small‐ to medium‐sized trees, steep slopes, and shallow soils being important regional risk factors. Early wilting revealed a lasting impact on forest productivity, with affected trees showing reduced greenness in the following spring. Our approach reliably detects early wilting at the resolution of large individual crowns and links it to key environmental drivers. It provides a sound basis to monitor and forecast early‐wilting responses that may follow the droughts of the coming decades.


Supplementary Tables & Figures
distribution of test data points is shown in panel B. Colors indicate early-wilting absence (green), early-wilting presence (brown), and artifacts, i.e., wilted grass, bare soil, logging activities (yellow). Locations interpreted with Google Earth imagery are shown as circles; locations interpreted with PLANET imagery are shown as triangles. Grey squares represent the Sentinel-2 tiles analyzed. 6     14 Fig. S12 | Pearson correlation coefficients between potential predictors used for the environmental GAM models and indication of removed predictors due to high correlation. Green circles represent absolute Pearson correlation coefficients below 0.7; orange circles represent absolute Pearson correlation coefficients between 0.7 and 0.9; and red circles represent absolute Pearson correlation coefficients higher than 0.9. Variable names in dark red and corresponding black-transparent lines represent variables that were removed due to high collinearity. Data basis were predictors at the Swiss scale. TWI represents terrain wetness index.

Media analysis
We screened a large part of the German-speaking newspaper and magazine articles published in Switzerland during 2018 for keywords related to (1)

Selection of NDVI time-series statistics
The different analyses of NDVI time series resulted in 39 summary statistics in total of which we selected ten. We removed all statistics describing number of observations (five) and intercepts of time-series trends (three) as they had no direct link to early wilting. Due to insufficient coverage, we further removed all statistics describing trends in the sub-series before and after the change points detected in the "Change, Aftereffect, Trend"-analysis of the May-September subset (14 statistics). Similarly, we removed standard deviations for the March/April and the October/November periods due to insufficient coverage. In order to limit multicollinearity, we removed another five predictors (Fig. S9)

Univariate explained deviance of climate predictors for different periods
The explained deviance of the climatic drivers for different time periods varied in particular for climate anomalies and was not always consistent between the Swiss and the Central European scales (Figs S11 and S12). This is likely because the drought varied in its temporal evolution across Central Europe.
However, in order to keep the period consistent for the different predictors and to account for hypothesized ecological relevance, we decided for August 2018 as the most relevant period for all climate predictors, except climatological precipitation, which we considered cumulated for the April-August period.

Correlation between climate, soil, terrain, and vegetation predictors
Among the environmental predictors considered, we found elevated absolute Pearson correlation coefficients particularly between elevation and temperature, as well as between terrain wetness index (TWI), slope, and soil conditions (Fig. S12). In order to reduce multicollinearity among predictors in multivariate analyses, we therefore removed elevation, TWI, coarse soil content, and water storage capacity from the predictor set. We removed elevation as it is a proxy for precipitation and in particular temperature and does not affect trees directly. Terrain wetness index was removed as its effect is slightly less straightforward to interpret than the effect of slope. Water storage capacity was removed due to its similarity to rooting depth.
Among these two factors rooting depth appeared to be the simpler and more robust measure. Finally, we removed coarse soil content because its effect on water stress appeared less direct than the effects of rooting depth and hydraulic conductivity. However, univariate analyses were unaffected by these removals and early-wilting responses to all predictors are shown in Figs. 4 and S7.