Simulating the recent drought-induced mortality of European beech ( <i>Fagus sylvatica</i> L.) and Norway spruce ( <i>Picea abies</i> L.) in German forests

Abstract. Drought is increasingly recognized as a critical driver of forest dynamics, altering tree species' growth, dominance and survival. To better understand these dynamics, we used a process-based modeling approach to investigate drought-related mortality of European beech (Fagus sylvatica L.) and Norway spruce (Picea abies L.) in German forests. The predisposing-inciting (PI) framework for drought-induced tree mortality incorporated in ForClim v4.1 was combined with a bark beetle module for Norway spruce to account for a key contributing factor, leading to ForClim v4.2. Our study addressed three hypotheses: (1) the PI framework, initially developed for Swiss beech forests, is effective across the broad ecological and climatic gradients found in Germany; (2) Soil properties, i.e. soil water holding capacity (AWC) and its spatial heterogeneity, have a strong influence on drought-related mortality by amplifying mortality risk through limited microsite variability, or dampening it by providing moist refugia, thus complementing climatic drivers; (3) incorporating bark beetle damage ameliorates model performance for simulating drought-related mortality of Norway spruce. Our modelling approach deliberately forgoes calibration to better investigate the underlying mechanisms and drivers of drought-induced tree mortality. We conducted simulations across 149 plots of the ICP Forest Level I network in Germany, covering a wide gradient of climate and soil conditions. ForClim reproduced the general patterns of drought-related mortality, highlighting the ability of the PI framework to capture emergent mortality patterns across a range of environmental conditions. However, mismatches in magnitude and trends highlight areas for improvement. Discrepancies were attributed to sparse mortality data, the drought sensitivity of the bark beetle submodule, and the absence of regional calibration. Our results revealed the critical role of AWC and local soil heterogeneity in modulating drought responses. Sites with low AWC experienced significantly higher mortality rates, while high AWC provided a buffering effect, bringing simulated outcomes closer to observed data. Furthermore, soil heterogeneity played a mitigating role, with sites exhibiting uniform soils showing higher mortality risk, thus emphasizing the importance of the spatial variability of soil properties for dampening drought impacts. Lastly, the new bark beetle submodel, even though highly simplified, considerably improved the simulation of drought-related mortality patterns in Norway spruce-dominated sites. This study underscores the value of process-based models like ForClim for disentangling the mechanisms underlying forest vulnerability and drought-induced mortality. However, improvements such as finer-resolution mortality and crown condition data, as well as regional model calibration, would be useful to enhance its predictive accuracy. Our findings contribute to the better understanding, forecasting and managing of forest resistance under current and future climatic conditions.