Predicting drought-induced tree mortality in Swiss beech forests hinges upon predisposing and inciting factors

The increase in the frequency and severity of drought -induced tree mortality in many Euro- pean low-elevation forests poses considerable challenges to forest management and requires an understanding of its causes. We propose a novel framework for integrating the factors un- derlying drought-induced tree mortality in a dynamic vegetation model. We evaluate whether this framework accurately reproduces drought-related mortality in six mesic beech-dominated stands in 2018-2020, and over multiple years in a xeric Scots pine- dominated stand in Switzerland. Additionally, we investigate its behavior along a large cli- matic gradient in central Europe. We employ a three-step approach. First, we evaluate multiple drought indices for capturing tree growth responses to ex- treme drought. We find that in contrast to widespread indices such as SPI and SPEI, the For- Clim drought index captures growth responses to drought intensity during summer, the grow- ing period, and annually. Second, we assess in detail the capability of the ForClim soil moisture model to simulate soil water dynamics, comparing it to the mechanistic soil -vegetation-atmosphere model LWFBrook90. The ForClim soil moisture model adequately simulates soil water dynamics, particularly in extreme drought years. Third, based on Manion’s Decline Disease Theory, we develop a novel mortality sub- model that combines predisposing and inciting factors. Its integration in ForClim captures drought-induced mortality events in the mesic beech forests as well as the multi-year mortality at the xeric Scots pine-dominated stand. Along a climatic gradient in central Europe, the model provides good quantifications of Potential Natural Vegetation. The novel framework to capture drought -related tree mortality is simple yet produces accurate results. The underlying hypothesis regarding the factors leading to drought-induced tree mortality is promising but requires further tests for its generality. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 3

Tree mortality, Beech decline, Soil water deficit , ForClim, Dynamic Vegetation Model .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 4 1 | Introduction 1 Over the last decades, climate-driven disturbances have progressively exposed temperate for-2 ests to stress, often compromising ecosystem functioning and the provision of ecosystem ser-3 vices (Breshears et al., 2005; Gilliam, 2016; Senf & Seidl, 2021). An increase in the frequency, 4 intensity, and duration of droughts (Buras et al., 2019; Jia et al., 2019; Rakovec et al., 2022) 5 has already led to substantial changes of forest structure in Europe (Schuldt et al. 2020) and 6 worldwide (Allen et al. 2015 , Millar et al. 2015). Droughts, which span from weekly to sea-7 sonal to multi-decadal, are important drivers of forest dynamics (Cook et al., 2015), triggering 8 declines in productivity and eventually increasing the likelihood of mortality of both decidu-9 ous and conifer trees (Allen et al. 2010, Anderegg et al., 2013). 10 While spruce (Picea abies) dieback has been observed for a long time in low-elevation 11 forests in Europe (Klimo et al., 2000 ; Bosela et al. 2021 ), the extreme droughts since 2015 12 have caused unprecedented mortality events also in beech (Fagus sylvatica) forests, including 13 areas previously not considered to be drought -prone (Schuldt et al., 2020 ; Frei et al., 2022; 14 Neycken et al., 2022 ). It is not clear whether this drought-induced tree mortality, often pre-15 ceded by canopy decline, is the result of prolonged or rapid stress , or both (Herguido et al., 16 2016; Neycken et al., 2022; Klesse et al., 2022; Schmied et al., 2023), and to what extent it is 17 the result of the interplay of abiotic and biotic stressors (Manion, 1981). 18 Declines in tree productivity are commonly assessed based on tree -ring data and drought in-19 dices (Bigler et al. 2006, Bose et al. 2021). However, the most frequently employed drought 20 indices neglect the role of soil water, although it is a major driver of productivity decline, at 21 least in more recent drought events (Liu et al., 2020, Carminati & Javaux, 2020). To predict 22 forest responses to climate change, process-based models are needed because they allow to 23 assess the interactions among the multiple factors that shape forest dynamics (Manusch et al., 24 2014; Huber et al., 2020; Bugmann & Seidl, 2022). Yet, to date dynamic vegetation models 25 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 5 have failed to reproduce drought-induced tree mortality (McDowell et al., 2013; Steinkamp & 26 Hickler, 2015; Hendrik & Cailleret, 2017; De Kauwe et al., 2022). This may be due to i) non-27 linear interactions among processes (Wang et al., 2012; Xu et al. 2013; Camarero, 2021), ii) 28 high process complexity that is difficult to capture (Orth et al., 2015; Lin & Yang, 2022), and 29 iii) scarcity of long-term, stand-scale observations with at least an annual resolution (Parolari 30 et al. 2014; Hartmann et al., 2018; Camarero, 2021). 31 Drought-induced tree mortality is generally thought to be due to the synergistic action of 32 multiple factors rather than isolated causes (McDowell et al., 2008-2011; Adams et al., 2017; 33 Gazol & Camarero, 2022). Three common hypotheses in this context are i) hydraulic failure 34 (Martínez-Vilalta et al., 2002 ; Brodribb & Cochard, 2008 ), ii) carbon starvation (Sevanto et 35 al., 2014), and iii) biotic agents (Neely & Manion, 1991). The interplay among these and likely 36 further factors may explain drought-induced tree mortality (“Decline Disease” framework, 37 Manion, 1981), including non-linear interactions among multiple stressors and specifically 38 distinguishing predisposing, inciting, and contributing factors (hereafter termed the “predis-39 posing and inciting factor ” (PIC) scheme). The PIC scheme posits that predisposing factors 40 (chronic stressors) enhance the vulnerability of trees to inciting (short-term acute stress) and 41 contributing factors ( secondary, long-term stress agents), eventually resulting in tree death 42 (Manion, 1981; Pedersen, 1998; Wang et al., 2012). However, in dynamic vegetation models, 43 the focus has predominantly been on either predisposing factors (e.g., low growth rate, growth 44 efficiency) or inciting factors (e.g., hydraulic failure), rather than their combination (Wang et 45 al., 2012). A few models that incorporated both processes (TRIPLEX by Liu et al . 2021 , 46 ED(X) by Moorcroft et al. 2001, GOTILWA+ by Nadal-Sala et al. 2017) still were limited in 47 their ability to capture observed tree mortality patterns. In these studies, indicators of hydraulic 48 failure and carbon starvation indicators were derived from the same data to both calibrate the 49 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 6 mortality formulation and determine its causes —without independent validation or incorpo-50 ration of temporal dynamics—thus potentially involving circular reasoning. 51 In this study, we developed a model based on the PIC scheme and tested whether it cap-52 tures drought-induced mortality. We then compared its performance to an earlier model ver-53 sion that does not distinguish between predisposing and inciting factors. Specifically, we ad-54 dressed the following questions: 55 1. Which drought index is best suited for capturing tree growth response to drought under 56 non-extreme climatic conditions ? We posit that drought indices lack ing information on 57 soil water content provide inferior performance, regardless of their temporal resolution. 58 2. What complexity is required to simulate the soil water balance in extremely dry years 59 accurately? We hypothesize that a low -complexity approach, e.g. a single-layer soil 60 bucket model at monthly temporal resolution, provides a good compromise between ease 61 of parameterization and process representation. 62 3. Do trees succumb to reduced growth alone, or is it due to compound events such as pro-63 longed stress followed by short -intense events? According to the PIC scheme, trees ex-64 posed to prolonged stress are most susceptible to mortality when a short -term, intense 65 event such as a seasonal drought occurs. We suggest that the PIC scheme helps to eluci-66 date the drought-induced mortality observed at many low-elevation sites in Europe during 67 2018-2020. 68 2. | Materials and Methods 69 2.1. | Site selection 70 We selected six even-aged mature beech-dominated forests in Switzerland that had experi-71 enced varying degrees of canopy decline and tree mortality in 2018-2020 (Neycken et al., 72 2022) (Figure 1 , Table 1 ). The y are located on the Swiss Plateau (Blattenberg, 73 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 7 Grosszinggibrunn, Rossberg, Tüeliboden, Vogtacher) and in the Jura mountains (Usserholz), 74 featuring highly varying soil properties as evident from the Available Water holding Capacity 75 (hereafter AWC, 86 to 155 mm), and annual precipitation sum (914 to 1325 mm). Despite the 76 observed mortality events, these sites are characteristic of the mesic climate of central Europe, 77 where drought is usually just mellow and drought-related mortality is quite rare (Supplemen-78 tary Material, Figure SM A1). 79 Additionally, we included the Lägeren site, which is part of the European ICP Level II 80 network (Etzold et al., 2010), is also dominated by Fagus sylvatica and has comparable eco-81 logical characteristics, e.g. an average annual precipitation of 1051 mm (MeteoSwiss, 2023). 82 Importantly, measured data on actual and potential evapotranspiration rates have been col-83 lected at Lägeren—data unavailable for the other beech -dominated sites. Including Lägeren 84 thus allows for an assessment of the water balance during recent drought years under environ-85 mental conditions analogous to those of the six beech stands. 86 We also selected a xeric forest dominated by Scots pine (Pinus sylvestris) in Visp, located 87 at 650 m a.s.l on a steep north -facing slope in the Valais (Switzerland) , to test the generality 88 of our framework. This site is also part of the ICP Level II network (Rebetez & Dobbertin, 89 2004). At Visp, drought-induced mortality events occurred regularly in the past (Bigler et al., 90 2006) and have been documented in detail since 1996 (Rigling & Cherubini, 1999; Hunziker 91 et al., 2022), thus offering a test bed for our framework in the context of the impacts of fre-92 quent and severe water stress on forests in the south-central Alps. 93 Lastly, to encompass a still broader climatic and edaphic spectrum, we studied multi-94 species forest dynamics at twelve sites across Switzerland and Germany. They span a gradient 95 from cold-mesic to warm-xeric climates, with mean annual precipitation ranging from 600 to 96 1350 mm and AWC from 100 to 240 mm (Bugmann & Solomon, 2000). 97 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 8 2.2 | Tree-ring data and drought indices 98 We retrieved cross -dated tree-ring width data from Neycken et al. (2022) for the six beech 99 sites. We standardized and detrended the individual tree-ring chronologies using a power-100 transformed Age-Dependent Spline (Obladen et al., 2021; Neycken et al., 2022). All detrended 101 series were averaged into a site chronology using a bi -weight robust mean for each site (cf. 102 SM B1). We assessed the tree growth response to extreme events using pointer year analysis 103 after calculating z-transformations of the detrended site chronolog ies based on a bi -weight 104 robust mean to detect extreme events at the stand level ( SM B2, Jetschke et al., 2019) . All 105 analyses were performed in R (v4.2, R Core Team, 2023) using the packages pointRes (van 106 der Maaten-Theunissen et al., 2021) and dplR (Bunn, 2008, 2010). 107 We selected a range of drought indices with varying levels of complexity. From the most 108 widely and commonly used single -variable indices, we used the Standardized Precipitation 109 Index (SPI, McKee et al., 1993) , which is defined as the difference of precipitation from its 110 mean over a specific time scale divided by its standard deviation . Moving towards higher 111 complexity, we included the Climatic Water Balance , defined as the difference between 112 monthly precipitation and monthly potential evapotranspiration (Thornthwaite, 1948; 113 Thornthwaite & Mather, 1954) . We furthermore selected the Standardized Precipitation 114 Evapotranspiration Index, SPEI (Beguería et al., 2014), computed as the standardized differ-115 ence between precipitation and potential evapotranspiration. We also used the self-calibrated 116 Palmer Drought Severity Index, scPDSI (Wells et al., 2004), which takes into account AWC. 117 Lastly, we evaluated the ForClim drought index (hereafter termed ‘ForClim’, Bugmann & 118 Solomon, 2000), which additionally distinguishes between tree water supply and demand (see 119 section ForClim model below). It is computed at a monthly time step during either the growing 120 season (for deciduous species) or across the year (for evergreen species ), provided that the 121 temperature is sufficiently high for tree growth. 122 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 9 The R package spei (Beguería et al., 2014) was used to calculate SPI, CWB and SPEI . A 123 modified version of the pdsi R wrapper function was used to calculate scPDSI (Zang, 2018). 124 The ForClim drought index was calculated using the R package ForDrought (cf. Data Avail-125 ability). 126 To consistently compare these drought indices, we computed all indices at a temporal resolu-127 tion of one month and averaged the monthly values for different periods: autumn (Sept through 128 Nov), winter (Dec through Feb), spring (March through May), summer (June through Aug), 129 the complete growing season (April to October), and the entire year (annual). We standardized 130 the scPDSI and ForClim drought indices using a z-transformation to facilitate direct compar-131 ison with the indices that are standardized by definition (i.e., SPI and SPEI) . We further re-132 scaled all indices according to McKee et al. (1993) to assign drought severity classes within 133 the following ranges: mild (-0 – -0.99), moderate (-1 – -1.49), severe (-1.5 – -1.99) and extreme 134 (≤ -2). 135 To assess the strength of the relationship between tree growth and drought intensity, we com-136 puted the bootstrapped Spearman correlation between the six detrended residual tree -ring 137 chronologies (ring width indices, hereafter RWI, Figure SM B2) and the detrended drought 138 indices (Table SM B1) in the common period 1980 –2020 for the selected temporal extents 139 and both the previous and the current year. To identify autocorrelation, we utilized residual 140 tree-ring chronologies and conducted a Mann-Kendall test to evaluate the extent of long-term 141 trends in the predictors ( i.e., the drought indices). When autocorrelation was found , we 142 detrended the series using Seasonal Trend Decomposition based on Loess (hereafter STL, 143 Cleveland et al., 1990 ; for full details cf. SM B3), which effectively separates a time series 144 into its seasonal trend and residual components. 145 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 10 2.3. | Soil moisture models of contrasting complexity: LWFBrook90 and ForClim 146 We evaluated the complexity required to simulate soil water dynamics regarding atmospheric 147 water demand (potential evapotranspiration, hereafter PET) and supply (soil moisture and ac-148 tual evapotranspiration, hereafter SM and AET, respectively) by comparing the performance 149 of the single-layer soil moisture balance model in ForClim (for details, see below) and the 150 detailed multi-layer Soil -Vegetation-Atmosphere Transfer (SVAT) model LWFBrook90 151 (Hammel & Kennel, 2001). LWFBrook90, an extension of Brook90 (Federer, 2002; Federer 152 & Lash, 1978), simulates daily transpiration and interception from the canopy using a single 153 plant layer ("big leaf" approach), as well as soil water fluxes by numerically solving the Rich-154 ards equation. In contrast to ForClim, LWFB rook90 does not impose a field capacity con-155 straint, allowing soil moisture to exceed this threshold and enabling a more detailed simulation 156 of soil water dynamics. This fundamental difference presents a considerable challenge when 157 comparing the behavior of the two models (Guswa et al., 2022). 158 In a first step, we compared the measured above-canopy AET at the site Lägeren with simu-159 lated AET. Simulated AET data for LWFBrook90 were obtained from Meusburger et al. 160 (2022) for the period 2013-2019. For ForClim, we simulated AET by forcing the model with 161 measured temperature and precipitation obtained from MeteoTest (2020). 162 In a second step, we simulated soil moisture, PET and AET at the six beech -dominated sites 163 and compared model behavior. LWFBrook90 was forced with daily weather time series (Me-164 teoTest, 2020) ; elevation and slope angle were provided from the digital elevation model 165 DHM25 (spatial resolution of 25 m ; Swisstopo, 2004). Additionally, Leaf Area Index (here-166 after LAI) data were obtained from the MODIS global Leaf Area Index and Fraction of Pho-167 tosynthetically Active Radiation (FPAR) product (Myneni et al., 2021) , while stand height 168 was retrieved from the global forest canopy height model by Potapov et al. (2020). Full details 169 on the simulation setup and parameters are provided in SM C (cf. Table C2.1). We calculated 170 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 11 AWC from soil texture properties (Baltensweiler et al., 2021) using the pedotransfer function 171 of Wessolek et al. (2009) (cf. sections D1 and D2 in SM, Eq. 10-12). 172 Model performance was evaluated as follows: at Lägeren, we computed the root mean squared 173 error (RMSE) and the mean absolute error (MAE ; cf. Eq. 1-2), while for the six beech sites 174 where no soil moisture data were available, we assessed the similarity of the simulated water 175 balance variables using cumulative density functions (CDFs) , in particular cumulated soil 176 moisture (CSM), and the Bland Altman method (BAM), which is commonly applied in med-177 icine and genomics (Giavarina, 2015; Bland & Altman, 1986) . BAM aims to describe the 178 degree of agreement between two methods or data sources. The y -axis of BAM plots repre-179 sents the difference between each pair of simulated variables from the two methods, whereas 180 the x-axis shows the mean of each pair of values among the two models (Eq. 3). The mean 181 bias (β, Eq. 4) and the limits of agreement (Eq. 5) were also calculated, where σ is the standard 182 deviation of the differences in each pair of values. According to BAM, 95% of the scatter 183 points should reside within the limits of agreement , representing ±1.96σ from the mean dif-184 ference between the data of the two models. If the mean difference of the data is not signifi-185 cantly different from zero (based on a one-sample t-test), this indicates good agreement be-186 tween the two methods. 187 𝑅𝑀𝑆𝐸 = √1 𝑛 ∑ (𝑌𝑝𝑖 − 𝑋𝑜𝑏𝑠𝑖)2 𝑛 𝑖=1 Eq. 1 𝑀𝐴𝐸 = ∑ |𝑌𝑝𝑖 − 𝑋𝑜𝑏𝑠𝑖|𝑛 𝑖=1 𝑛 Eq. 2 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 12 𝜎𝑑 = √∑(𝑋𝑖 − 𝑌𝑖) (𝑁 − 1) Eq. 3 𝛽 = 1 𝑛 ∑ 𝑌𝑝𝑖 − 𝑋𝑜𝑏𝑠𝑖 𝑛 𝑖=1 Eq. 4 𝐶𝐼𝑢𝑝𝑝𝑒𝑟 = 𝛽 − 1.96𝜎𝑑 𝐶𝐼𝑙𝑜𝑤𝑒𝑟 = 𝛽 + 1.96𝜎𝑑 Eq. 5 2.4 | ForClim Model 188 2.4.1. | Overview 189 ForClim is a forest gap model (Botkin et al., 1972) originally designed to capture the long-190 term (i.e., decades to centuries) growth, mortality and regeneration of trees in temperate for-191 ests of central Europe and account for climate change effects on forest dynamics (Bugmann, 192 1994, 1996). In gap models, forest dynamics are simulated on small areas (‘patches’), usually 193 with a size of 400-1000 m2, each representing one out of many stochastic realizations that are 194 spatially independent of each other. The soil water balance is calculated using a monolayer 195 “bucket” model that stores all incident water until its capacity (“bucket size”, kBS) is reached, 196 which corresponds to AWC. The bucket model has a fixed field capacity, beyond which soil 197 moisture cannot increase. This simplification facilitates computations but limits the represen-198 tation of hydrological processes such as percolation and soil moisture variability above field 199 capacity. For details, cf. Bugmann & Cramer (1998) and Bugmann & Solomon (2000). 200 This bucket model is directly connected to plant dynamics, as tree growth is determined as a 201 species-specific potential (i.e., under optimum conditions) that is reduced via growth -reduc-202 tion factors (GRF) accounting for light availability, crown condition, temperature, nitrogen 203 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 13 and soil water supply vs. demand (Bugmann & Solomon, 2000). Tree mortality from stress is 204 assumed to occur when diameter growth is below a specific stem diameter increment or a 205 fraction of the maximum increment for several years ( Solomon, 1986). Thus, predisposing 206 stress is assumed to increase the mortality rate , but there is no formulation of inciting stress 207 (cf. the PIC scheme). 208 2.4.2. | Integrating predisposing and inciting stress factors 209 To enhance the response of tree growth to environmental extremes (temperature and soil water 210 dynamics), we modified the original growth reduction formulation (Huber et al. 2020) by ap-211 plying Liebig's “law of the minimum” for the temperature (DDGF) and soil moisture (SMGF) 212 growth factors (Eq. 6; cf. Liebig et al., 1842), rather than multiplying them: 213 𝐺𝑅𝐹 = √𝐶𝐿𝐺𝐹 · SNGF · ALGF · min (𝑆𝑀𝐺𝐹, 𝐷𝐷𝐺𝐹) Eq. 6 where GRF is the overall growth reduction factor, and CLGF, SNGF and ALGF are the crown-214 related, nitrogen -related and light -related growth factors, respectively. The underlying ra-215 tionale is that in the temperate and boreal regions, conditions are typically either dry or cold, 216 but not dry and cold (cf. Bugmann, 1996b). To preserve the species-specific response to en-217 vironmental stress although the level of GRF is changing due to the use of Eq. 6, we had to 218 re-estimate a temperature-related species-specific parameter (for details cf. SM D6), and we 219 also modified the dynamic formulation of site index, which reflects the response of maximum 220 tree height to temperature and drought and their changes over time (for details cf. SM D7). 221 To disentangle predisposing and inciting factors leading to drought-related mortality, we 222 identified short-term (within a year) and long -term (multi-year) stressors linked to drought 223 duration and intensity as well as carbon starvation, as explained below. 224 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 14 First, we accounted for the effect of long-lasting droughts by a predisposing factor using 225 a drought memory term (DrM; Wang et al., 2012; Eq. 7): 226 𝐷𝑟𝑀 = {𝐷𝑟𝑀 + 1, 𝑔𝐷𝑟 > 𝑘𝐷𝑟𝑇ℎ · 𝑘𝐷𝑟𝑇𝑜𝑙𝑠 0, 𝑒𝑙𝑠𝑒 Eq. 7 This formulation counts all continuous years in which drought intensity, represented by the 227 ForClim drought index (gDr), exceeds a fraction (kDrTh) of the species-specific drought tol-228 erance (kDrTols). In this manner, we account for the species -specific resistance to multi -an-229 nual drought stress. To be consistent, we also modified the “slow growth” factor 𝑆𝐺𝑟 to better 230 mimic the impact of carbon reserves on mortality risk (for details see SM D3). 231 Second, to deal with the inciting factor we defined drought duration within any given year 232 (gDrD, Eq. 8) as the ratio of the number of dry months relative to the total number of months 233 m of the growing period (for deciduous species, 𝑚𝑔𝑝; annual for evergreen species, 𝑚𝑎𝑛). The 234 algorithm selects those months in which average temperature (Tm) is above a threshold kJ (5.5 235 °C) and water supply (i.e., transpiration, gEm, cm ) relative to water demand from the soil 236 (gDm, cm) is below a threshold kEg; i.e., gEm/gDm < kEg; and SMs < kBS). The term 𝟙 repre-237 sents an indicator function, which equals 1 if the condition is true for a given month , and 0 238 otherwise. 239 𝑔𝐷𝑟𝐷𝑔𝑝 = 1 𝑚𝑔𝑝 ∑ 𝟙(𝑇𝑚 ≥ kJ)∙ 𝟙( 𝑆𝑀𝑚 < 𝑘𝐵𝑆)∙ 𝟙 ( 𝑔𝐸𝑚 𝑔𝐷𝑚 < kEg) 10 𝑚=4 Eq. 8 𝑔𝐷𝑟𝐷𝑎𝑛 = 1 𝑚𝑎𝑛 ∑ 𝟙(𝑇𝑚 ≥ kJ)∙ 𝟙( 𝑆𝑀𝑚 < 𝑘𝐵𝑆)∙ 𝟙 ( 𝑔𝐸𝑚 𝑔𝐷𝑚 < kEg) 12 𝑚=1 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 15 Third, under conditions of low demand (i.e., in spring and fall) this index would not record 240 any drought, although soil moisture (SMm) may be limiting e.g. for bud break in spring. Thus, 241 we used two variables (Eq. 9) to capture limiting soil moisture levels in spring and fall along 242 with a threshold (kDu) for the duration of the drought to define an inciting factor for drought 243 stress (IncFDr; Eq. 9). 244 𝐼𝑛𝑐𝐹𝐷𝑟 = { 1, (𝑆𝑀𝑠𝑝𝑟𝑖𝑛𝑔 < 𝑘𝑅𝐸𝑊𝑠𝑝𝑟𝑖𝑛𝑔 · 𝑘𝐵𝑆) ∧ (𝑆𝑀𝑓𝑎𝑙𝑙 𝑘𝐷𝑢 0, 𝑒𝑙𝑠𝑒 Eq. 9 This formulation for the seasonal water deficit in spring and fall is based on the concept of 245 ‘relative extractable water’ (REW; Bréda et al., 2006; Granier et al., 1999 ; SM D4). Specifi-246 cally, the spring component of IncFDr reflects the need of trees to mobilize water for bud 247 break and cell division and elongation, while the fall component reflects the need of accumu-248 lating carbon reserves for the subsequent year (Figure 3). The threshold kDu was set to 0.28, 249 corresponding to two months out of a seven -month growing period (2/7 ≈ 0.28 ) for broad-250 leaves, and three to four months out of the whole year (3.5 /12 ≈ 0.29) for evergreen species, 251 provided that winters are warm enough ( cf. Hidy et al., 2021; Merganičová, 2023) . The sea-252 sonal soil moisture levels (SMfall, SMspring) are calculated for the fall (September to November) 253 and spring (March to May) periods for both evergreen and deciduous species. 254 Lastly, t he overall stress -induced mortality probability (gPStr), including the carbon 255 memory and integrating predisposing as well as inciting factors, is formulated as follows: 256 𝑔𝑃𝑆𝑡𝑟 = { 𝑘𝑆𝑡𝑟𝑒𝑠𝑠𝑃, 𝑆𝐺𝑟 > 𝑘𝑆𝐺𝑟𝑇 ∨ ( 𝐷𝑟𝑀 > 𝑘𝑆𝐺𝑟𝑇 ∧ 𝐼𝑛𝑐𝐹𝐷𝑟 = 1 ) 0, 𝑒𝑙𝑠𝑒 Eq. 10 where 𝑘𝑆𝑡𝑟𝑒𝑠𝑠𝑃 is the stress-induced enhanced mortality probability, 𝑆𝐺𝑟 is the slow-growth 257 counter, and 𝑘𝑆𝐺𝑟𝑇 indicates the number of stress years that are tolerated until mortality prob-258 ability is enhanced (Peltier et al., 2023). The first term of Eq. 10 (SGr condition) captures the 259 probability that a tree may die due to slow growth induced by whatever cause (e.g., insufficient 260 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 16 light), whereas the second term (DrM and IncFDr conditions) reflects that a string of dry years 261 can enhance mortality under a particularly prolonged summer drought coupled to early- and/or 262 late-season soil moisture depletion. 263 The ensemble of these features gives rise to ForClim v4.1. Since contributing factors 264 (Manion, 1981) such as insect damage are currently not included, we refer to the concept 265 underlying ForClim v4.1 as the “PI framework”, rather than PIC . Note that none of the pa-266 rameters of the new drought-related mortality framework (Eqs. 7-10) were calibrated against 267 any data sources, least those used for the simulation studies that are explained below. 268 2.4.3 | Model initialization 269 Six beech sites – We compiled forest stand data using relascopic sampling (Grosenbaugh, 270 1958, Bitterlich, 1948) in combination with the single -tree inventory conducted by Neycken 271 et al. (2022) in 2019 and 2020, resulting in a total of 14 small sampling plots of 10 m radius 272 for Blattenberg, 35 for Grosszinggibrunn, 14 for Rossberg, 10 for Tüeliboden, 11 for Usser-273 holz, and 36 for Vogtacher. For each site, we calculated the mean basal area and its standard 274 deviation for the respective years (cf. Table A1, SM A). Individual DBH data were collected 275 for the target trees, defined as those sampled by Neycken et al. (2022) for dendrochronological 276 analysis and crown vitality assessment. Additionally, DBH measurements for neighboring 277 trees within a 10 m radius of each target tree were included in the dataset (Neycken et al., 278 2022). To further enhance the inventory data, we incorporated information from trees felled 279 following observed mortality at each site. These records were used to retrospectively recon-280 struct stand structure from 2018 back to 2000, which was employed to initialize the model for 281 the year 2000. This process was carried out by reintroducing the dead trees into each of the 282 sampled plots at the start of the simulation time , and the same process was applied for the 283 neighboring trees. 284 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 17 For model initialization, t he inventory data of vital, declining, and dead/cut trees were ran-285 domly distributed on n = 200 patches, each with a size of 0.085 ha, as simulations need to be 286 replicated to represent n stochastic realizations of the processes simulated in ForClim (cf. 287 Bugmann, 2001). The simulations were forced by interpolated, site-specific monthly meteor-288 ological time series of precipitation and temperature (MeteoTest, 2020), while soil properties 289 were obtained from Baltensweiler et al. (2021). For each site, AWC ( kBS) was calculated 290 using the Wessolek pedotransfer function (cf. Eqs. 10-12 in SM D1 and D2). These stands are 291 experiencing high atmospheric nitrogen (N) deposition (Roy et al., 2021), and thus we set N 292 availability to a high value of 180 kg ha -1, assuming N not to be limiting (cf. Table S7, SM 293 D7). 294 ICP Level II sites – For both Lägeren and Visp, the state of the forest was determined by 295 simulating Potential Natural Vegetation from bare ground (i.e., no stand initialization) to equi-296 librium under current climate conditions, assuming no management. At Lägeren, this state 297 was used to initialize the simulations for the period 2011-2020. Similarly, for Visp we forced 298 the model with precipitation and temperature time series from the LWF Visp meteorological 299 station for the period 1997-2004 (Haeni et al., 2019). 300 European climate gradient (large-scale Potential Natural Vegetation) – We performed sim-301 ulations starting from bare ground over 200 patches (patch size 0.08 ha) for 1500 years to 302 determine the equilibrium between forest vegetation and climate. We averaged the simulated 303 species-specific basal area between the simulation years 1300 and 1500 ( for details cf. SM 304 D8). The weather time series were obtained by randomly sampling years from the site-specific 305 climatology (Bugmann, 1994, 1996; cf. SM D8). 306 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 18 3. | Results 307 3.1 | Selecting the most suitable Drought Index for tree growth response 308 As expected, we found only weak relationships between tree growth and drought intensity in 309 the mesic mature beech forests: correlations were low to medium, yet with distinctive seasonal 310 differences (Figure 3). Still, the drought indices featured rather distinct behavior across the six 311 sites and seasons, suggesting that it is possible to identify an index that best captures extreme 312 events. 313 Specifically, there was a tendency for indices of low complexity, particularly SPI, to have 314 lower correlations across all seasons and sites, while indices of higher complexity performed 315 better, yet with notable exceptions. Surprisingly, t he scPDSI demonstrated poorer perfor-316 mance compared to simpler indices during spring, the growing period, and the entire year. 317 Similarly, the ForClim drought index exhibited low performance in spring, but outperformed 318 the other indices during summer, the entire growing period and throughout the year. In the 319 autumn period, the ForClim drought index showed the highest, yet non -significant, correla-320 tions at three out of six sites. 321 Non-parametric correlations between the current tree -ring index and the previous year ’s 322 drought index exhibited consistent patterns across sites, yet they were non-significant for all 323 seasons (Figure SM B3). Notably, ForClim showed the highest correlations during the grow-324 ing period and at the annual time scale at most sites for the previous year . It also featured 325 significant correlations in autumn for Grosszinggibrunn and Usserholz, followed by Blatten-326 berg, and in spring except for Usserholz. However, due to the zero variance of the ForClim 327 index at three of the beech sites—Grosszinggibrunn, Tüeliboden, and Vogtacher —correla-328 tions could not be calculated for the current and previous year at these sites. 329 Overall, our analysis indicate d that ForClim is capable of picking up the relatively weak 330 drought signals at many of these mesic beech sites, being the most suitable index for the entire 331 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 19 growing period and the entire year; all indices perform ed similarly for the summer, and – 332 importantly – the use of any of these indices is discouraged for spring and autumn. 333 3.2 | Simulating the water balance: how much complexity is needed? 334 At the site Lägeren, observed versus simulated AET indicate d that both ForClim and 335 LWFBrook90 captured the seasonal cycle of AET well (Figure 4A,B), yet with distinct dif-336 ferences, surprisingly indicating a somewhat lower performance of the more complex model, 337 LWFBrook90, compared to ForClim, regardless of the metric being used (Table 2). Meus-338 burger et al. (2022) attributed the performance problems of LWFBrook90 to the omission of 339 lateral water fluxes at Lägeren, which resulted in lower simulated AET (Figure 4A). The For-340 Clim model almost consistently underestimated AET, especially during winter. However, it 341 performed well during the dry years of 2015 and 2018 -2019, showing a reduction in AET, 342 particularly in the summer months. In contrast, the LWF-Brook90 model featured both under- 343 and overestimations of AET (Figure 4B). Notably, it significantly overestimated AET during 344 the summer months of 2015-2016 and 2018-2019, followed by a substantial reduction in fall. 345 Soil moisture simulated by ForClim at the six beech sites was consistently higher than for 346 LWFBrook90 (Figure 5 A) except for Blattenberg (cf. Fig.SM C2.1), where simulated soil 347 moisture was larger in LWFBrook90 ( at average + 16%). For Tüeliboden, ForClim yielded 348 higher values (+17%), and considerably higher values for Grosszinggibrunn, Usserholz and 349 Vogtacher (+28 to +30%; cf. Table 3). The higher soil water availability simulated by ForClim 350 occurred mostly during the cold seasons, i.e. autumn and winter (cf. SM C, Fig. C2.2). 351 The cumulative density functions (CDFs) for AET showed good agreement between the 352 models at Grosszinggibrunn, Rossberg and Vogtacher, but ForClim produced notably lower 353 values at Usserholz, Blattenberg, and Tüeliboden. For PET, slight ly lower values were ob-354 served from ForClim across all sites, ranging from -2.93 cm (Vogtacher) to -5.83 cm (Usser-355 holz) (cf. Fig. SM C2.3). 356 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 20 The Bland-Altmann analyses (Figure 5B) revealed varying degrees of agreement between 357 ForClim and LWFBrook90 in simulating cumulative soil moisture (CSM , cm) across the six 358 beech sites and across the years. The models showed closer alignment during dry years (2003, 359 2015, and 2018 ; Figure 5B ) compared to wetter years, albeit with year -specific differences 360 (cf. Fig. SM C2.4). Notably, in 2018 —the driest year — the points were more tightly clustered 361 around the zero-bias line compared to 2003 and 2015, indicating higher consistency between 362 the models during this extreme drought. Moreover, no significant difference was found be-363 tween the mean CSM simulated by the two models for 2018 (t = 1.02, p = 0.21), although 364 some site-specific discrepancies were evident, particularly at Tüeliboden. 365 These results suggest that both models capture the same trend of soil moisture depletion 366 during extreme droughts, and the ForClim model tends to yield higher soil moisture in wetter 367 periods and lower values in drier periods at some sites compared to the more complex 368 LWFBrook90. Overall, LWFBrook90 featured a less conservative behavior with soil mois-369 ture, which likely is a result of its daily resolution, thus better capturing extreme situations. 370 ForClim also captured the summer drying of the soil and the extreme summer droughts in the 371 years 2003, 2015, and 2018, while simulated soil moisture remained close to saturation for the 372 remainder of the year (cf. Fig. SM C2.2). 373 3.3 | Is tree mortality driven by compound stress events? 374 The previous model version, ForClim v4.0.1, lacked the capability to simulate drought-in-375 duced mortality at the six beech-dominated sites, whereas the implementation of the PI frame-376 work in ForClim v4.1 enabled us to quantify this phenomenon well, except for the site Blat-377 tenberg (Figure 6). In both ForClim versions, t he simulated basal area show ed no drought-378 related mortality events in the period 2000 -2017, not even in the very dry year s 2003 and 379 2015, whereas ForClim v4.1 featured a notable reduction starting in 2018 (Figure 6). These 380 simulated mortality events are consistent with observations, yet with some site-specific over- 381 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 21 and underestimation. The simulations for Rossberg featured the lowest MAE and RMSE 382 among sites (Table 4), indicating the highest accuracy, followed by Usserholz and Tüeliboden, 383 suggesting an overall good to medium level of consistency in prediction errors across sites. 384 Grosszinggibrunn and Vogtacher featured a higher variability (dispersion of errors) compared 385 to the other sites. Blattenberg featured the lowest accuracy and only a slight decrease in basal 386 area, i.e. only little mortality. 387 The reduction of simulated basal area in ForClim v4.1 across all six beech sites indicated a 388 consistent trend over the three driest years (Table 3). Tüeliboden had the strongest decrease 389 compared to the previous years, followed by Usserholz and Grosszinggibrunn. Vogtacher and 390 Rossberg showed a smaller decrease than the other sites, while an increase of simulated basal 391 area was observed in Blattenberg in the year 2018. 392 When simulating stand dynamics at the ICP-Level II site of Visp (Figure 7), basal area and 393 stem number featured a strong reduction in the years 1999 and 2003. This pattern is in line 394 with the mortality observed by Dobbertin et al. (2004) and Hunziker et al. (2022) for that same 395 period, amounting to ~75.6% from 1997 to 2004. 396 When analyzing model behavior across an extended environmental gradient in Europe, For-397 Clim v4.1 tended to feature lower basal area compared to ForClim 4.0.1 (Figure 8) . The dif-398 ferences were particularly revealing when analyzing the species composition along the eleva-399 tional and climatic gradient, as explained below. 400 In ForClim v4.1. the major features of forest basal area and species composition along 401 the extended climatic gradient from cold -wet to warm -dry conditions (Figure 8A) were re-402 tained compared to the predecessor version 4.0.1 (Figure 8B): Conifers dominate the high -403 elevation cold-wet sites (Grande Dixence to Adelboden), mixed conifer-deciduous forests are 404 found at mid-elevations, which also feature the highest basal area along the gradient . Beech 405 (Fagus sylvatica) dominance extends from Huttwil to Basel, and towards the dry edge of the 406 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 22 gradient drought-adapted species like oaks ( Quercus spp.) and lime ( Tilia spp.) are gaining 407 dominance, ending up with pine dominance towards the dry treeline (Sion). 408 Yet, there are marked difference in the performance of ForClim v4.1 compared to v4.0.1. 409 Specifically, we found a consistent reduction in the share of Picea abies (Grande Dixence, 410 Davos, Bern), and the share of Abies alba decreased slightly at the cool -wet sites. Also, a 411 substantial decrease of Tilia cordata was found (cf. Figure SM D7). On the contrary, an in-412 crease in the share of Pinus sylvestris (Schaffhausen, Basel) was evident from ForClim v4.1. 413 At the south-facing site Bever in the continental high Alps, the share of Pinus sylvestris was 414 reduced substantially in favor of Pinus cembra. Importantly, the share of Fagus sylvatica in-415 creased moderately in Adelboden, Schwerin and Potsdam, while it decreased somewhat at the 416 drier sites of the Swiss Plateau (Schaffhausen and Basel). 417 4. | Discussion 418 4.1 | Seasonal drought periods modulate growth responses to drought 419 Drought indices are often used to capture the relationship between drought intensity and tree 420 growth. Also, they are important for monitoring forest health and resilience as they help to 421 identify areas under drought stress (Zargar et al., 2011; Vicente-Serrano et al., 2012). As ex-422 pected, we found that different indices perform quite differently . T he performance of a 423 drought index is contingent on its complexity, expressed among others by the number of pre-424 dictors and the temporal extent. In our study, particularly those indices that are used widely in 425 dendrochronology, i.e. SPI and SPEI (Speich, 2019; Schwarz et al., 2020) performed less well 426 at all sites and across different temporal extents. Surprisingly, the Palmer Drought Severity 427 Index performed rather poorly as well, particularly in summer, despite being used as a stand-428 ard index to capture drought patterns (e.g. as a component of the US Drought Monitor, 2019). 429 Thus, while these indices may be suitable to capture strong droughts in semi-arid or arid areas, 430 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 23 they are less suitable to characterize drought occurrence in mesic areas, as shown in the present 431 study. 432 In contrast, the ForClim drought index, which embeds a simple single-layer soil moisture 433 model, was able to capture the distinct drought intensity signals in summer, across the vege-434 tation period, and at the annual level. These results underline that water availability is key for 435 shaping the response of trees to drought (Speich et al., 2018; Gessler et al., 2022; Klesse et 436 al., 2022; Meusburger et al., 2022). Thus, the ForClim Drought Index has notable predictive 437 capability in detecting drought signals during the growing season even under mesic condi-438 tions. This underscores its robustness in regions with moderate drought, highlighting its utility 439 for assessing drought impacts on tree growth across a wide range of conditions. 440 The rather poor performance of the ForClim Drought Index in spring and autumn suggests 441 that it cannot capture early- or late-season water deficits, as it is tied to the ratio of water 442 supply to demand alone. Yet, recent studies highlight that forests in northern as well as south-443 ern Europe have experienced widespread increases in drought sensitivit y in the spring to au-444 tumn periods (Jin et al., 2023), underscoring the importance of detecting such early-season 445 droughts more effectively. To this end, an alternative approach to the ratio of supply and de-446 mand needs to be found for capturing drought intensity , because in spring and fall demand 447 (Potential Evapotranspiration) tends to be low, but trees may still suffer if there is insufficient 448 soil moisture to supply their needs (cf. Granier et al. 1980-2006, Breda et al. 1999). 449 4.2 | A simple soil water dynamics model captures drought stress in extreme years 450 Model selection for simulating soil water dynamics must be made with care, as simple bucket 451 models (Koster and Suarez, 1994) and Soil-Vegetation-Atmosphere Transfer (SVAT) models 452 are built on fundamentally different principles. Each model type is tailored to address specific 453 research objectives, justifying varying temporal and spatial resolutions (Ohrt et al., 2015). 454 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 24 On the one hand, the SVAT LWFBrook90 incorporates detailed soil hydraulic properties 455 through the Mualem-van Genuchten parameterization , which enhances the spatial (vertical 456 stratification) and temporal (sub-monthly) resolution of the simulated variables. However, this 457 comes with the challenge of high data demand (cf. Table SM C2.2) and the need for extensive 458 model calibration. This becomes a particular challenge for forest sites where meteorological 459 data of high spatial and temporal resolution are scarce and long records of daily soil moisture 460 data are typically absent. This can lead to high uncertainty in parameter estimation and, ulti-461 mately, model projections (Schmidt-Walter et al. 2020, Franks et al. 1997). 462 On the other hand, ForClim’s simple soil moisture scheme with its monthly time step and 463 a single soil bucket layer was found to perform well for detecting strong droughts, as shown 464 in the comparison with (1) tree-ring series at the six beech sites and (2) evapotranspiration 465 dynamics at Lägeren. However, the simplicity of a bucket model may have limitations during 466 wet seasons and in moist years (cf. Romano et al. 2011, Fischer et al. 2011), where ForClim 467 tends to simulate lower water demand and higher water supply compared to LWFBrook90 (cf. 468 Hagemann and Stacke, 2015). This behavior depends on the structural simplifications under-469 lying bucket models that, while aimed at reducing computational complexity, tend to ignore 470 processes such as water retention and movement across soil layers (Kondo, 1993). 471 Despite these differences, which are least pronounced during dry spells, both models are 472 highly suitable for simulating the soil moisture balance, particularly in the context of long -473 term forest growth and development, i.e. when the focus is on seasonal patterns and inter-474 annual variability rather than short periods of time (e.g., days to weeks) , where only 475 LWFBrook90 would provide the required level of detail. 476 Overall, despite their structural and methodological differences, ForClim and 477 LWFBrook90 feature strong agreement in simulat ed soil moisture depletion and declining 478 actual evapotranspiration during dry years, in line with previous studies (e.g., Meusburger et 479 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 25 al., 2022). Th us, although ForClim is based on a very simple framework compared to the 480 complex SVAT approach of LWFBrook90, it is still effective in capturing strong drought 481 events even under mesic conditions. In addition, its simplicity in terms of parameter require-482 ments and temporal resolution render it a viable tool for studies where ease of use and low 483 input requirements are important. 484 4.3 | Predisposing and inciting factors allow to capture drought-induced tree mortality 485 The combination of predisposing and inciting factors (Manion, 1981) in ForClim v4.1 enabled 486 to capture the exceptional mortality events induced by prolonged and intense drought s in 487 Swiss temperate forests that are not usually prone to drought. Specifically, the PI framework 488 revealed the importance of separately treating predisposing stress factors that operate over 489 longer periods (Neycken et al. 2022) , and short -term, intense stress factors that affect trees 490 already under stress (Etzold et al., 2019). A long-term reduction in growth alone is insufficient 491 to capture the mortality probability (Steinkamp & Hickler, 2015, DeSoto et al., 2020), and the 492 same goes for a short -term stressor such as hydraulic failure ( Körner, 2019). Our analysis 493 suggests that a simple framework combining both aspects is suitable for predicting drought -494 related tree mortality , even though – or perhaps exactly because – plant-physiological pro-495 cesses are not modeled explicitly. 496 Regarding the predisposing factors, the capability to recover from stress (Schwalm et al., 497 2017; Ovenden et al., 2021 ) was captured well by a growth memory term in our PI scheme 498 (cf. Cailleret et al., 2017; Bottero et al., 2021; Zamora-Pereira et al., 2021). By incorporating 499 a drought-induced reduction in diameter increment over the long term, our approach closely 500 aligns with the concept of carbon starvation (Peltier et al. 2023, Gessler et al. 2018 ). Our 501 approach assesses tree vitality without engaging with the highly complex and uncertain as-502 pects of tree carbon balance and allocation (Sala et al. 2010 ; Hartmann 2015; Fatichi et al. 503 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 26 2014). Instead, we rely on an integrative measure, i.e. tree-ring width, as a proxy of tree vitality 504 (cf. Waring, 1980, Cherubini et al., 2021). Yet, the inclusion of tree recovery mechanisms in 505 dynamic forest models is essential to reflect the adaptive mechanisms of trees and assess the 506 impacts of droughts on the long-term stability of ecosystems as well as future trajectories 507 under changing climatic conditions (Gessler, 2020). 508 Regarding the inciting factors, the combination of seasonal drought duration, limitations 509 to the evapotranspiration rate along with the early- and late-seasonal soil water deficit turned 510 out to be highly suitable for approximating the combination of low precipitation and pro-511 nounced soil moisture deficits that were found to be characteristic of the hot -dry droughts of 512 the last decades (Breshears et al., 2005; Adams et al., 2009) . Note that high Vapor Pressure 513 Deficit (VPD), which was found to be particularly pronounced in the 2018 drought in low -514 elevation forests (e.g., Gharun et al., 2020) is closely correlated with a low ratio of water 515 supply to demand, which is at the core of our formulation of the inciting factor. Soil moisture 516 in spring and fall is critical for budburst and reserve building, respectively (Michelot et al. 517 2012; Massonet et al., 2021), and its integration via Relative Extractable Water (Breda et al., 518 1999) clearly improved the model’s capability of capturing the prolonged water deficits expe-519 rienced during extreme droughts such as in 2018 (Brun et al., 2020; Gharun et al., 2020; 520 Meusburger et al., 2022), which are projected to increase under future climatic conditions 521 (Ruosteenoja et al., 2018). 522 Modeling frameworks that incorporated both carbon starvation and hydraulic failure often 523 have failed to capture tree mortality well (e.g., Hajek et al., 2022; Fischer et al., 2024). Thus, 524 even dynamic forest models that feature detailed physiological process representations may 525 still lack methods, processes or feedback s that account for multiple causes of tree mortality, 526 thereby missing key elements of tree death (Anderegg et al. 2012 , Bugmann & Seidl 2022 ). 527 Our framework suggests that mortality arises from a combination of prolonged, chronic stress 528 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 27 (which may include carbon starvation, but is not restricted to it) that is weakening trees over 529 multiple years, followed by acute, severe stress (which may or may not include hydraulic 530 failure) within a growing season. This integrat ive approach accounts for the compounding 531 effects of long-term physiological decline and sudden environmental extremes, offering a po-532 tentially holistic depiction how trees succumb to drought. By focusing on the compound effect 533 of chronic and acute stress, our model adds to the conventional view of a dichotomy between 534 carbon starvation and hydraulic failure and highlights the importance of capturing stress dy-535 namics across different scales and time frames. 536 The simulations for a xeric site dominated by Scots pine (site Visp) demonstrate the gen-537 erality of our approach beyond the mesic beech sites. It should be noted that we did not adjust 538 any parameters for these simulations, although the climatic setting as well as the dominant 539 species are entirely different. To our own surprise, the simulated increase of the mortality rate 540 under these xeric conditions was consistent with previous research (Dobbertin et al., 2004; 541 Bigler et al., 2006; Wohlgemuth et al., 2018), suggesting that our framework may not be re-542 stricted to beech and highlighting the general importance of short, intense droughts in addition 543 to less severe, longer-lasting droughts in the context of tree death (cf. Rigling and Cherubini, 544 1999). 545 Ultimately, our approach shows that strong model simplifications may not be a problem 546 but a virtue, rendering it possible to assess drought-related tree mortality via the combination 547 of growth-related and growth -independent limitation s, not necessitating a physiologically 548 based approach (Fatichi et al., 2014; Körner, 2015). Our choice reflects the trade-off between 549 a small number of ecological assumptions while still accounting for the critical role of limiting 550 factors, such as temperature and soil moisture, in shaping long-term tree growth responses 551 under drought (cf. Huber et al., 2020, 2021). 552 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 28 4.4 | Model behavior along a large gradient of temperature and precipitation 553 Since its early development (versions ≤4.0.1), ForClim featured a distinct species distribution 554 pattern across an extended gradient’s elevational belts and climatic regions in Europe, which 555 was attributable to the main underlying ecological assumptions (Bugmann, 1994). Our study 556 demonstrates the continued realism of the simulations of Potential Natural Vegetation by For-557 Clim v4.1 along the very same gradient of temperature and precipitation (Bugmann and Sol-558 omon, 2000) , thus corroborating the robustness of the new formulation of drought -related 559 mortality across sites and species. Yet, more detailed studies are required to further substanti-560 ate this, using data e.g. from the ICP Level-I and Level-II network (Bussotti et al. 2024) re-561 garding widespread, recently drought-inflicted species such as beech (Fagus sylvatica), spruce 562 (Picea abies) or Scots pine ( Pinus sylvestris), to generalize from our case studies to larger 563 areas (Knapp et al. 2024). 564 The simulations with the new model version (v4.1) along this gradient featured not only 565 a reduction of total basal area to more plausible values (cf. Idoate et al., 2024), but also distinct 566 differences at the level of the abundance of individual species. The reduction in the abundance 567 of spruce (Picea abies) in low-elevation areas is consistent with descriptions by Frehner et al. 568 (2005), and the same goes for the decrease of beech (Fagus sylvatica ) towards drier sites 569 (Bohn & Gollub, 2006) . Yet, in ForClim v4.1 beech continued to dominate forests on the 570 Swiss Plateau and strongly suppressed other species via light competition , which is also in 571 line with previous studies (cf. Heiri et al. 2009). Towards the dry end of the gradient, beech 572 lost dominance in favor of lime (Tilia spp.) and Scots pine (Pinus sylvestris), which is con-573 firmed by studies e.g. from the northern German lowlands (Diers et al., 2023) as well as gen-574 eral vegetation mapping efforts (Bohn & Gollub, 2006 ). Thus, the PI scheme that we intro-575 duced did not only improve model behavior under drought conditions, but it generally im-576 proved the depiction of competitive relationships along this very wide environmental gradient. 577 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 29 4.5 | Limitations and outlook 578 The quantitative testing of our PI framework relied on data from a small set of temperate, 579 mesic beech-dominated sites and one Scots-pine dominated stand, which may limit the general 580 applicability of our findings. Additionally, the absence of comprehensive inventory data and 581 long-term observations required us to reconstruct stand structure partly based on circumstan-582 tial evidence, which may have reduced the accuracy of our results. While drought events like 583 the one in 2018 seemed rare or even erratic at the time of observation , their repeated occur-584 rence over the last years highlights the need for long-term monitoring, including annual, tree-585 level mortality assessments, to better capture the ongoing effects of these extreme events. 586 To simulate Potential Natural Vegetation, we decided to use a gradient of sites across a 587 wide range of climatic conditions in Switzerland and Germany. This was primarily a modeling 588 choice rather than one rooted in the reality of the landscape, which strongly limits the quanti-589 tative comparison of simulated vs. ‘real’ data, as there are few if any primeval forests in Eu-590 rope that could be considered to be in equilibrium with climate. A way forward would be to 591 validate the model against data from strict forest reserves (cf. Käber et al. 2024 for tree regen-592 eration), thus providing a more tangible baseline for quantitative comparisons (cf. Brang and 593 Bolliger, 2015). 594 A clear limitation of our study concerns soil water dynamics and the comparison of the 595 ForClim and LWFBrook90 models, which was difficult due to the scarcity of long-term data 596 on key components of the water balance in forests beyond precipitation, i.e. soil moisture, 597 evapotranspiration, and runoff. The analyses of simulated soil water dynamics would benefit 598 from an in-depth data-model comparison using long-term time series data from multiple sites. 599 Lastly, the lack of measured soil data ( i.e., quantitative analyses of samples from soil 600 profiles) introduced some degree of uncertainty in the simulated soil moisture dynamics as 601 well as the simulated forest responses to drought . As soil properties vary tremendously in 602 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 30 space, particularly in forest stands, high-resolution soil maps are key tools to derive estimates 603 of AWC (cf. Baltensweiler et al. 2022). These products, however, are characterized by sub-604 stantial uncertainty, particularly when it comes to soil depth and rock fraction, which play a 605 major role in determining AWC. Constraining this uncertainty and enhancing the quality of 606 soil maps is key to better constrain and assess the impacts of drought stress on forests (Wal-607 thert et al. 2020, Klesse et al., 2022). 608 5. | Conclusions 609 We demonstrated that widely applied drought indices like the SPI or SPEI, although showing 610 significant responses in the summer months, are overall not capable of detecting drought sig-611 nals at mesic sites across multiple seasons, even though these droughts have led to tree mor-612 tality. By contrast, indices of higher complexity, particularly the ForClim drought index, are 613 suitable for identifying such dry years and capturing their effect on tree growth and demogra-614 phy. We therefore recommend using indices such as the ForClim drought index; the essential 615 feature is that such indices must consider soil water storage in addition to mere climatic data. 616 Our comparison of the highly detailed LWFBrook90 model with the simple ForClim soil 617 moisture balance model, which is based on a monthly resolution, suggests that annual drought 618 severity can be captured well with a highly simplified scheme, although it is clear that for 619 capturing any higher -resolution features in either time or space , particularly under milder 620 drought conditions, SVAT models may be preferred. In the context of drought effects on 621 growth and mortality of forest trees, we posit that simple soil moisture balance models are 622 sufficient, bringing the distinct advantage of ease of parameterization and low runtime. 623 The novel framework for drought-related tree mortality that we developed suggests that 624 the combination of predisposing and inciting factors is pivotal. It furthermore indicates that 625 trees do not normally die due to carbon starvation or hydraulic failure alone, but due to the 626 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 31 combination of prolonged stress with sudden (within a year) severe stress that covers the entire 627 growing period. We even posit that it is possible to model these ecophysiological processes in 628 a highly simplified way, specifically that they can be formulated at an aggregate, general level. 629 Even if simplified, our framework permits to (1) isolate the multiple factors as drivers of tree 630 responses to drought stress and (2) accurately simulate beech mortality in response to the 2018 631 drought, even at sites not usually prone to drought-related mortality. The new mortality model 632 also features high robustness when applied to a xeric system with another species (Scots pine) 633 as well as along an extended climatic gradient and mixed species stands. 634

635 We are thankful to Dr. Anna Neycken and Dr. Mathieu Lévesque (Chair of Silviculture, Dept. 636 Environmental Systems Sciences, ETH Zürich) for providing the inventory and tree-ring data 637 of the six beech sites. We thank Prof. Dr. Christof Bigler (Chair of Forest Ecology, Dept. 638 Environmental Systems Sciences, ETH Zürich) for the discussions on detrending predictors 639 and response variables altogether as good scientific practice in dendrochronology. We are 640 thankful to Dr. Fabian Bernard (Oeschger Centre for Climate Change Research, University of 641 Bern; WSL, Birmensdorf) for advice on the selection of LWFBrook90 outputs and how to 642 aggregate them, and to Dr. Lorenz Walthert (WSL, Birmensdorf) for references on the rela-643 tionship between soil water and drought stress. The ForClim source code was curated by 644 Hussain Abbas and Dr. Thomas Oliver Hands (Forest Ecology, ETH Zurich), and they deserve 645 a special acknowledgment for their excellent support. 646 Gina Marano was supported by the project entitled "Embracing structural uncertainty in mod-647 els of forest dynamics", under Grant Agreement N. 188882 which has received funding by the 648 Swiss National Science Foundation (SNSF). 649 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 32 Credit authorship contribution statement 650 Gina Marano, Harald Bugmann and Ulrike Hiltner conceived the idea, developed the concept 651 and framing of the paper; Gina Marano conducted the analyses, created the graphics, and 652 drafted the main manuscript; Katrin Meusburger supported the work with LWFBrook90 sim-653 ulations, input data and provided advice regarding the analyses ; Gina Marano and T homas 654 Oliver Hands curated the ForClim source code . All authors reviewed the manuscript and ap-655 proved its submission. 656 Data availability 657 All R scripts, R packages and codes are archived in a Zenodo repository and can be accessed 658 at: https://zenodo.org/records/14234972 659 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 33

Adams, H. D., Guardiola-Claramonte, M., Barron-Gafford, G. A., Villegas, J. C., Breshears, D. D., Zou, C. B., Troch, P. A., & Huxman, T. E. (2009). Temperature sensitivity of drought-induced tree mortality portends increased regional die-off under global-change-type drought. Proceedings of the National Academy of Sciences of the United States of America, 106(17), 7063–7066. https://doi.org/10.1073/pnas.0901438106 Adams, H. D., Zeppel, M. J. B., Anderegg, W. R. L., Hartmann, H., Landhäusser, S. M., Tissue, D. T., Hux- man, T. E., Hudson, P. J., Franz, T. E., Allen, C. D., Anderegg, L. D. L., Barron-Gafford, G. A., Beer- ling, D. J., Breshears, D. D., Brodribb, T. J., Bugmann, H., Cobb, R. C., Collins, A. D., Dickman, L. T., Duan, H., Ewers, B. E., Galiano, L., Galvez, D. A., Garcia-Forner, N., Gaylord, M. L., Germino, M. J., Gessler, A., Hacke, U. G., Hakamada, R., Hector, A., Jenkins, M. W., Kane, J. M., Kolb, T. E., Law, D. J., Lewis, J. D., Limousin, J.-M., Love, D. M., Macalady, A. K., Martínez-Vilalta, J., Mencuccini, M., Mitchell, P. J., Muss, J. D., O’Brien, M. J., O’Grady, A. P., Pangle, R. E., Pinkard, E. A., Piper, F. I., Plaut, J. A., Pockman, W. T., Quirk, J., Reinhardt, K., Ripullone, F., Ryan, M. G., Sala, A., Sevanto, S., Sperry, J. S., Vargas, R., Vennetier, M., Way, D. A., Xu, C., Yepez, E. A., & McDowell, N. G. (2017). A multi-species synthesis of physiological mechanisms in drought-induced tree mortality. Nature Ecol- ogy & Evolution, 1(9), 1285–1291. https://doi.org/10.1038/s41559-017-0248-x Allen, C. D., Macalady, A. K., Chenchouni, H., Bachelet, D., McDowell, N., Vennetier, M., ... & Cobb, N. (2010). A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. Forest ecology and management, 259(4), 660-684. Allen, C. D., Breshears, D. D., & McDowell, N. G. (2015). On underestimation of global vulnerability to tree mortality and forest die‐off from hotter drought in the Anthropocene. Ecosphere, 6(8), 1-55. Anderegg, L. D. L., Anderegg, W. R. L., Abatzoglou, J., Hausladen, A. M., & Berry, J. A. (2013). Drought characteristics’ role in widespread aspen forest mortality across Colorado, USA. Global Change Biol- ogy, 19(5), 1526–1537. https://doi.org/10.1111/gcb.12146 Anderegg, L. D. L., Anderegg, W. R. L., & Berry, J. A. (2013). Not all droughts are created equal: Translat- ing meteorological drought into woody plant mortality. In Tree Physiology (Vol. 33, Issue 7, pp. 701– 712). https://doi.org/10.1093/treephys/tpt044 Baltensweiler, A., Walthert, L., Hanewinkel, M., Zimmermann, S., & Nussbaum, M. (2021). Machine learn- ing based soil maps for a wide range of soil properties for the forested area of Switzerland. Geoderma Regional, 27. https://doi.org/10.1016/j.geodrs.2021.e00437 Beguería, S., Vicente-Serrano, S. M., Reig, F., & Latorre, B. (2014). Standardized precipitation evapotranspi- ration index (SPEI) revisited: Parameter fitting, evapotranspiration models, tools, datasets and drought monitoring. International Journal of Climatology, 34(10). https://doi.org/10.1002/joc.3887 Bhuyan, U., Zang, C., & Menzel, A. (2017). Different responses of multispecies tree ring growth to various drought indices across Europe. Dendrochronologia, 44, 1–8. https://doi.org/10.1016/j.dendro.2017.02.002 Bigler, C., Bräker, O. U., Bugmann, H., Dobbertin, M., & Rigling, A. (2006). Drought as an Inciting Mortal- ity Factor in Scots Pine Stands of the Valais, Switzerland. Ecosystems, 9(3), 330–343. https://doi.org/10.1007/s10021-005-0126-2 Bitterlich, W. (1948). Die Winkelzghlprobe. Allgemeine Forst- und Holzwirtschaftliche Zeitung 59( l/2):4-5. Bohn, U., & Gollub, G. (2006). The use and application of the map of the natural vegetation of Europe with particular reference to Germany. In Biology and Environment: Proceedings of the Royal Irish Acad- emy (pp. 199-213). Royal Irish Academy. Bose, A., Scherrer, D., Camarero, J. J., Ziche, D., Babst, F., Bigler, C., Bolte, A., Dorado-Liñán, I., Etzold, S., Fonti, P., Bose, A. K., Forrester, D. I., Gavinet, J., Gazol, A., González De Andrés, E., Nikolaus Karger, D., Lebourgeois, F., Lévesque, M., Martínez-Sancho, E., … Rigling, A. (2021). Climate sensi- tivity and drought seasonality determine post-drought growth recovery of Quercus petraea and Quercus robur in Europe. Science of the Total Environment, 784, 147222. https://doi.org/10.1016/j.sci- totenv.2021.147222ï Bosela, M., Tumajer, J., Cienciala, E., Dobor, L., Kulla, L., Marčiš, P., Popa, I., Sedmák, R., Sedmáková, D., Sitko, R., Šebeň, V., Štěpánek, P., & Büntgen, U. (2021). Climate warming induced synchronous .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 34 growth decline in Norway spruce populations across biogeographical gradients since 2000. Science of the Total Environment, 752. https://doi.org/10.1016/j.scitotenv.2020.141794 Botkin, D. B., Janak, J. F., & Wallis, J. R. (1972). Some ecological consequences of a computer model of for- est growth. The Journal of ecology, 849-872. Bottero, A., Forrester, D. I., Cailleret, M., Kohnle, U., Gessler, A., Michel, D., Bose, A. K., Bauhus, J., Bug- mann, H., Cuntz, M., Gillerot, L., Hanewinkel, M., Lévesque, M., Ryder, J., Sainte-Marie, J., Schwarz, J., Yousefpour, R., Zamora-Pereira, J. C., & Rigling, A. (2021). Growth resistance and resilience of mixed silver fir and Norway spruce forests in central Europe: Contrasting responses to mild and severe droughts. Global Change Biology, 27(18), 4403–4419. https://doi.org/10.1111/gcb.15737 Brang, P., & Bolliger, M. (2015). Waldreservate. In A. Rigling & H. P. Schaffer (Eds.), Waldbericht 2015. Zustand und Nutzung des Schweizer Waldes (pp. 88-89). Bundesamt für Umwelt BAFU; Eidg. For- schungsanstalt für Wald, Schnee und Landschaft WSL. Bréda, N., Huc, R., Granier, A., & Dreyer, E. (1980). Temperate forest trees and stands under severe drought: a review of ecophysiological responses, adaptation processes and long-term consequences Nathalie. 28(3), 389–390. https://doi.org/10.1016/s0034-5288(18)32731-0 Bréda, N., Huc, R., Granier, A., & Dreyer, E. (2006). Temperate forest trees and stands under severe drought: A review of ecophysiological responses, adaptation processes and long-term consequences. In Annals of Forest Science (Vol. 63, Issue 6, pp. 625–644). https://doi.org/10.1051/forest:2006042 Breshears, D. D., Cobb, N. S., Rich, P. M., Price, K. P., Allen, C. D., Balice, R. G., Romme, W. H., Kastens, J. H., Floyd, M. L., Belnap, J., Anderson, J. J., Myers, O. B., & Meyer, C. W. (2005). Regional vegeta- tion die-off in response to global-change-type drought. Proceedings of the National Academy of Sci- ences of the United States of America, 102(42), 15144–15148. https://doi.org/10.1073/pnas.0505734102 Brodribb, T. J., & Cochard, H. (2009). Hydraulic failure defines the recovery and point of death in water- stressed conifers. Plant Physiology, 149(1), 575–584. https://doi.org/10.1104/pp.108.129783 Brun, P., Psomas, A., Ginzler, C., Thuiller, W., Zappa, M., & Zimmermann, N. E. (2020). Large-scale early- wilting response of Central European forests to the 2018 extreme drought. Global Change Biology, 26(12), 7021–7035. https://doi.org/10.1111/gcb.15360 Bugmann, H. (1994). On the Ecology of Mountainous Forests in a Changing Climate: A Simulation Study. Bugmann, H. (1996a). A Simplified Forest Model to Study Species Composition Along Climate Gradients. In Source: Ecology (Vol. 77, Issue 7). Bugmann, H. (1996b). Functional types of trees in temperate and boreal forests: Classification and testing. Journal of Vegetation Science 7: 359-370. Bugmann, H., & Cramer, W. (1998). Improving the behaviour of forest gap models along drought gradients. Forest Ecology and Management, 103(2–3), 247–263. https://doi.org/10.1016/S0378-1127(97)00217-X Bugmann, H. K. M., & Solomon, A. M. (2000). Explaining forest composition and biomass across multiple biogeographical regions. Ecological Applications, 10(1), 95–114. https://doi.org/https://doi.org/10.1890/1051-0761(2000)010[0095:EFCABA]2.0.CO;2 Bugmann, H. (2001). A review of forest gap models. Climatic Change, 52, 259–305. https://doi.org/10.1017/S0022112088002617 Bugmann, H., & Seidl, R. (2022). The evolution, complexity and diversity of models of long‐term forest dy- namics. Journal of Ecology, 110(10), 2288-2307. Bunn, A. G. (2008). A dendrochronology program library in R (dplR). Dendrochronologia, 26(2), 115–124. https://doi.org/10.1016/j.dendro.2008.01.002 Bunn, A. G. (2010). Statistical and visual crossdating in R using the dplR library. Dendrochronologia, 28(4), 251–258. https://doi.org/10.1016/j.dendro.2009.12.001 Buras, A., Rammig, A., & Zang, C. S. (2019). Quantifying impacts of the drought 2018 on European ecosys- tems in comparison to 2003. ArXiv, 1655–1672. https://doi.org/10.5194/bg-2019-286 Bussotti, F., Potočić, N., Timmermann, V., Lehmann, M. M., & Pollastrini, M. (2024). Tree crown defolia- tion in forest monitoring: concepts, findings, and new perspectives for a physiological approach in the face of climate change. Forestry: An International Journal of Forest Research, 97(2), 194-212. Cailleret, M., Jansen, S., Robert, E. M. R., Desoto, L., Aakala, T., Antos, J. A., Beikircher, B., Bigler, C., Bugmann, H., Caccianiga, M., Čada, V., Camarero, J. J., Cherubini, P., Cochard, H., Coyea, M. R., Čufar, K., Das, A. J., Davi, H., Delzon, S., … Martínez-Vilalta, J. (2017). A synthesis of radial growth patterns preceding tree mortality. Global Change Biology, 23(4), 1675–1690. https://doi.org/10.1111/gcb.13535 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 35 Camarero, J. J. (2021). The drought‒dieback‒death conundrum in trees and forests. In Plant Ecology and Di- versity (Vol. 14, Issues 1–2, pp. 1–12). Taylor and Francis Ltd. https://doi.org/10.1080/17550874.2021.1961172 Carminati, A., & Javaux, M. (2020). Soil Rather Than Xylem Vulnerability Controls Stomatal Response to Drought. Trends in Plant Science, xx(xx), 1–13. https://doi.org/10.1016/j.tplants.2020.04.003 Cleveland R. B., Cleveland W. S., McRae J. E., Terpenning & I. (1990). STL: A Seasonal-Trend Decomposi- tion Procedure Based on Loess. Journal of Official Statistics, 6(1), 3–73. Cherubini, P., Battipaglia, G., & Innes, J. L. (2021). Tree vitality and forest health: can tree-ring stable iso- topes be used as indicators? Current Forestry Reports, 7, 69-80. Cook, E. R., Seager, R., Kushnir, Y., Briffa, K. R., Büntgen, U., Frank, D., Krusic, P. J., Tegel, W., van der Schrier, G., Andreu-Hayles, L., Baillie, M., Baittinger, C., Bleicher, N., Bonde, N., Brown, D., Carrer, M., Cooper, R., Čufar, K., Dittmar, C., Esper, J., Griggs, C., Gunnarson, B., Günther, B., Gutierrez, E., Haneca, K., Helama, S., Herzig, F., Heussner, K.-U., Hofmann, J., Janda, P., Kontic, R., Köse, N., Kyncl, T., Levanič, T., Linderholm, H., Manning, S., Melvin, T. M., Miles, D., Neuwirth, B., Nicolussi, K., Nola, P., Panayotov, M., Popa, I., Rothe, A., Seftigen, K., Seim, A., Svarva, H., Svoboda, M., Thun, T., Timonen, M., Touchan, R., Trotsiuk, V., Trouet, V., Walder, F., Ważny, T., Wilson, R., & Zang, C. (2015). Old World megadroughts and pluvials during the Common Era. Science Advances, 1(10), e1500561. https://doi.org/10.1126/sciadv.1500561 De Kauwe, M. G., Sabot, M. E., Medlyn, B. E., Pitman, A. J., Meir, P., Cernusak, L. A., Gallagher R.V., Uk- kola A. M, Sami W.R., Choat, B. (2022). Towards species‐level forecasts of drought‐induced tree mor- tality risk. New Phytologist, 235(1), 94-110. DeSoto, L., Cailleret, M., Sterck, F., Jansen, S., Kramer, K., Robert, E. M. R., Aakala, T., Amoroso, M. M., Bigler, C., Camarero, J. J., Čufar, K., Gea-Izquierdo, G., Gillner, S., Haavik, L. J., Hereş, A.-M., Kane, J. M., Kharuk, V. I., Kitzberger, T., Klein, T., Levanič, T., Linares, J. C., Mäkinen, H., Oberhuber, W., Papadopoulos, A., Rohner, B., Sangüesa-Barreda, G., Stojanovic, D. B., Suárez, M. L., Villalba, R., & Martínez-Vilalta, J. (2020). Low growth resilience to drought is related to future mortality risk in trees. Nature Communications, 11(1), 545. https://doi.org/10.1038/s41467-020-14300-5 Diers, M., Weigel, R., & Leuschner, C. (2023). Both climate sensitivity and growth trend of European beech decrease in the North German Lowlands, while Scots pine still thrives, despite growing sensitivity. Trees - Structure and Function, 37(2), 523–543. https://doi.org/10.1007/s00468-022-02369-y Dobbertin, M., Mayer, P., Wohlgemut, T., Feldmeyer-Christe, E., Graf, U., Zimmermann, N. E., & Rigling, & A. (2004). The Decline of Pinus sylvestris L. Forests in the Swiss Rhone Valley-a Result of Drought Stress? In APGC (Vol. 45). www.biologiezentrum.at Etzold, S., Buchmann, N., & Eugster, W. (2010). Contribution of advection to the carbon budget measured by eddy covariance at a steep mountain slope forest in Switzerland. Biogeosciences, 7(8), 2461–2475. https://doi.org/10.5194/bg-7-2461-2010 Etzold, S., Ziemińska, K., Rohner, B., Bottero, A., Bose, A. K., Ruehr, N. K., Zingg, A., & Rigling, A. (2019). One century of forest monitoring data in Switzerland reveals species- And site-specific trends of climate-induced tree mortality. Frontiers in Plant Science, 10(March). https://doi.org/10.3389/fpls.2019.00307 Fatichi, S., Leuzinger, S., & Körner, C. (2014). Moving beyond photosynthesis: From carbon source to sink- driven vegetation modeling. New Phytologist, 201(4), 1086–1095. https://doi.org/10.1111/nph.12614 Federer, C. A. (2002). BROOK 90: A simulation model for evaporation, soil water, and streamflow. [WWW Document]. http://www.ecoshift.net/brook/brook90.html Federer, C. A., & Lash, D. (1978). Brook: A Hydrologic Simulation Model for Eastern Forests. https://scholars.unh.edu/nh_wrrc_scholarship Fischer R., Anders T., Bugmann H., Djahangard M., Gunnar Dreßler G., Hetzer J., Hickler T., Hiltner U., Marano G., Sperlich D., Yousefpour R., Knapp N. (2024). Perspectives for forest modeling to improve the representation of extensive tree mortality after extreme events (under revision, Journal of Cultivated Plants) Franks, S. W., Beven, K. J., Quinn, P. F., & Wright, I. R. (1997). On the sensitivity of soil-vegetation-atmos- phere transfer (SVAT) schemes: equifinality and the problem of robust calibration. Agricultural and Forest Meteorology, 86(1-2), 63-75. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 36 Frehner, M., Wasser, B., & Schwitter, R. (2005). Nachhaltigkeit und Erfolgskontrolle im Schutzwald. Weg- leitung für Pflegemassnahmen in Wäldern mit Schutzfunktion. Bundesamt für Umwelt, Wald und Land- schaft, Bern, 564. Gazol, A., & Camarero, J. J. (2022). Compound climate events increase tree drought mortality across Euro- pean forests. Science of the Total Environment, 816. https://doi.org/10.1016/j.scitotenv.2021.151604 Gessler, A., Cailleret, M., Joseph, J., Schönbeck, L., Schaub, M., Lehmann, M., Treydte K, Rigling A., Timo- feeva G., Saurer, M. (2018). Drought induced tree mortality–a tree-ring isotope based conceptual model to assess mechanisms and predispositions. New Phytologist, 219(2), 485-490. Gessler, A., Bächli, L., Rouholahnejad Freund, E., Treydte, K., Schaub, M., Haeni, M., Weiler, M., Seeger, S., Marshall, J., Hug, C., Zweifel, R., Hagedorn, F., Rigling, A., Saurer, M., & Meusburger, K. (2022). Drought reduces water uptake in beech from the drying topsoil, but no compensatory uptake occurs from deeper soil layers. New Phytologist, 233(1), 194–206. https://doi.org/10.1111/nph.17767 Gharun, M., Hörtnagl, L., Paul-Limoges, E., Ghiasi, S., Feigenwinter, I., Burri, S., Marquardt, K., Etzold, S., Zweifel, R., Eugster, W., & Buchmann, N. (2020). Physiological response of Swiss ecosystems to 2018 drought across plant types and elevation: Summer 2018 drought in Switzerland. Philosophical Transac- tions of the Royal Society B: Biological Sciences, 375(1810). https://doi.org/10.1098/rstb.2019.0521 Giavarina, D. (2015). Understanding Bland Altman analysis. Biochemia Medica, 25(2), 141–151. https://doi.org/10.11613/BM.2015.015 Gilliam, F. S. (2016). Forest ecosystems of temperate climatic regions: from ancient use to climate change. New Phytologist, 212(4), 871–887. https://doi.org/10.1111/nph.14255 Gillner, S., Rüger, N., Roloff, A., & Berger, U. (2013). Low relative growth rates predict future mortality of common beech (Fagus sylvatica L.). Forest Ecology and Management, 302, 372–378. https://doi.org/10.1016/j.foreco.2013.03.032 Granier, A., Bréda, N., Biron, P., & Villette, S. (1999). A lumped water balance model to evaluate duration and intensity of drought constraints in forest stands. Ecological Modelling, 116(2–3), 269–283. https://doi.org/10.1016/S0304-3800(98)00205-1 Grimm, V., Revilla, E., Berger, U., Jeltsch, F., Mooij, W. M., Railsback, S. F., Thulke, H. H., Weiner, J., Wiegand, T., & DeAngelis, D. L. (2005). Pattern-oriented modeling of agent-based complex systems: Lessons from ecology. Science, 310(5750), 987–991. https://doi.org/10.1126/science.1116681 Grosenbaugh, L. H. (1958). Point-sampling and line-sampling: Probability theory, geometric implications, synthesis. USDA Forest Service, Southern Forest Experiment Station, Occasional Paper 160, 34p. Hammel, K. , & Kennel, M. (2001). Charakterisierung und Analyse der Wasserverfügbarkeit und des Was- serhaushalts von Waldstandorten in Bayern mit dem Simulationsmodell. Forstliche Forschungsberichte Muenchen Hajek, P., Link, R. M., Nock, C. A., Bauhus, J., Gebauer, T., Gessler, A., Kovach K., Messier C., Paquette A., Saurer M., Scherer-Lorenzen M., Rose L., Schuldt, B. (2022). Mutually inclusive mechanisms of drought‐induced tree mortality. Global Change Biology, 28(10), 3365-3378. Hartmann, H., Moura, C. F., Anderegg, W. R. L., Ruehr, N. K., Salmon, Y., Allen, C. D., Arndt, S. K., Breshears, D. D., Davi, H., Galbraith, D., Ruthrof, K. X., Wunder, J., Adams, H. D., Bloemen, J., Cailleret, M., Cobb, R., Gessler, A., Grams, T. E. E., Jansen, S., … O’Brien, M. (2018a). Research frontiers for improving our understanding of drought-induced tree and forest mortality. New Phytolo- gist, 218(1), 15–28. https://doi.org/10.1111/nph.15048 Haeni, Matthias; Gessler, Arthur; Rebetez, Martine (2019). Visp, Switzerland: Long-term forest meteorologi- cal data from the Long-term Forest Ecosystem Research Programme (LWF), from 1997 onwards. En- viDat. doi:10.16904/envidat.108. Hagemann, S., & Stacke, T. (2015). Impact of the soil hydrology scheme on simulated soil moisture memory. Climate Dynamics, 44, 1731-1750. Hartmann, H. (2015). Carbon starvation during drought-induced tree mortality–are we chasing a myth?. Jour- nal of Plant Hydraulics, 2, e005. Heiri, C., Wolf, A., Rohrer, L., & Bugmann, H. (2009). Forty years of natural dynamics in Swiss beech for- ests: structure, composition, and the influence of former management. Ecological Applications, 19(7), 1920-1934. Hendrik, David., & Maxime, Cailleret. (2017). Assessing drought-driven mortality trees with physiological process-based models. Agricultural and Forest Meteorology, 232, 279–290. https://doi.org/10.1016/j.agrformet.2016.08.019 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 37 Herguido, E., Granda, E., Benavides, R., García-Cervigón, A. I., Camarero, J. J., & Valladares, F. (2016). Contrasting growth and mortality responses to climate warming of two pine species in a continental Mediterranean ecosystem. Forest Ecology and Management, 363, 149–158. https://doi.org/10.1016/j.foreco.2015.12.038 Hidy, D., Barcza, Z., Hollós, R., Thornton, P. E., Running, S. W., & Fodor, N. (2021). User’s Guide for Biome-BGCMuSo 6.1. http://www.ntsg.umt.edu/project/biome-bgc Huber, N., Bugmann, H., & Lafond, V. (2020). Capturing ecological processes in dynamic forest models: why there is no silver bullet to cope with complexity. Ecosphere, 11(5). https://doi.org/10.1002/ecs2.3109 Huber, N., Bugmann, H., Cailleret, M., Bircher, N., & Lafond, V. (2021). Stand-scale climate change impacts on forests over large areas: transient responses and projection uncertainties. Ecological Applications, 31(4). https://doi.org/10.1002/eap.2313 Hunziker, S., Begert, M., Scherrer, S. C., Rigling, A., & Gessler, A. (2022). Below average midsummer to early autumn precipitation evolved into the main driver of sudden Scots pine vitality decline in the Swiss Rhône valley. Frontiers in Forests and Global Change, 5, 874100. Idoate-Lacasia, J., Stillhard, J., Portier, J., Brang, P., Zimmermann, S., Bigler, C., Bugmann H.,Hobi, M. L. (2024). Long-term biomass dynamics of temperate forests in Europe after cessation of manage- ment. Forest Ecology and Management, 554, 121697. Jetschke, G., van der Maaten, E., & van der Maaten-Theunissen, M. (2019). Towards the extremes: A critical analysis of pointer year detection methods. In Dendrochronologia (Vol. 53, pp. 55–62). Elsevier GmbH. https://doi.org/10.1016/j.dendro.2018.11.004 Jia, G., Shevliakova, E., Artaxo, P., Noblet-Ducoudré, N. De, Houghton, R., House, J., Kitajima, K., Len- nard, C., Popp, A., Sirin, A., Sukumar, R., & Verchot, L. (2019). Climate Change and Land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems. In Climate Change and Land: an IPCC special report on climate change, desertification, land degradation, sustainable land management, food security, and greenhouse gas fluxes in terrestrial ecosystems. Jin, H., Vicente-Serrano, S. M., Tian, F., Cai, Z., Conradt, T., Boincean, B., Murphy, C., Farizo, B. A., Grainger, S., López-Moreno, J. I., & Eklundh, L. (2023). Higher vegetation sensitivity to meteorologi- cal drought in autumn than spring across European biomes. Communications Earth and Environment, 4(1). https://doi.org/10.1038/s43247-023-00960-w Käber, Y., Hartig, F., & Bugmann, H. (2024). Inferring the tree regeneration niche from inventory data using a dynamic forest model. Geoscientific Model Development, 17(7), 2727-2753. Klesse, S., Wohlgemuth, T., Meusburger, K., Vitasse, Y., von Arx, G., Lévesque, M., Neycken, A., Braun, S., Dubach, V., Gessler, A., Ginzler, C., Gossner, M. M., Hagedorn, F., Queloz, V., Samblás Vives, E., Rigling, A., & Frei, E. R. (2022). Long-term soil water limitation and previous tree vigor drive local variability of drought-induced crown dieback in Fagus sylvatica. Science of the Total Environment, 851. https://doi.org/10.1016/j.scitotenv.2022.157926 Klimo, E., Hager, H., & Kulhavý, J. (2000). Spruce monocultures in Central Europe: Problems and prospects (Vol. 33). Joensuu, Finland: European Forest Institute. Knapp, N., Wellbrock, N., Bielefeldt, J., Dühnelt, P., Hentschel, R., & Bolte, A. (2024). From single trees to country-wide maps: Modeling mortality rates in Germany based on the Crown Condition Survey. Forest Ecology and Management, 568, 122081. ISSN 0378-1127. https://doi.org/10.1016/j.foreco.2024.122081. Kobayashi, T., Matsuda, S., Nagai, H., & Tesima, J. (2001). A bucket with a bottom hole (BBH) model of soil hydrology. IAHS-AISH Publication, 270, 41–46. Koster, R. D., & Suarez, M. J. (1994). The components of a ‘SVAT’ scheme and their effects on a GCM's hydrological cycle. Advances in Water Resources, 17(1-2), 61-78. DOI: 10.1016/0309-1708(94)90024- 8 Körner, C. (2015). Paradigm shift in plant growth control. In Current Opinion in Plant Biology (Vol. 25, pp. 107–114). Elsevier Ltd. https://doi.org/10.1016/j.pbi.2015.05.003 Körner, C. (2019). No need for pipes when the well is dry—a comment on hydraulic failure in trees. Tree Physiology, 39(5), 695–700. https://doi.org/10.1093/treephys/tpz030 Liebig, J., Playfair, L. P., & Webster, J. W. (1842). Chemistry in its application to agriculture and physiol- ogy. J. Owen. https://doi.org/10.5962/bhl.title.30425 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 38 Lin, C. Y., & Yang, Y. C. E. (2022). The Effects of Model Complexity on Model Output Uncertainty in Co- Evolved Coupled Natural-Human Systems. Earth’s Future, 10(6). https://doi.org/10.1029/2021EF002403 Lindroth, A., Holst, J., Heliasz, M., Vestin, P., Lagergren, F., Biermann, T., Cai, Z., & Mölder, M. (2018). Agricultural and Forest Meteorology E ff ects of low thinning on carbon dioxide fl uxes in a mixed hem- iboreal forest. 262(June), 59–70. https://doi.org/10.1016/j.agrformet.2018.06.021 Liu, L., Gudmundsson, L., Hauser, M., Qin, D., Li, S., & Seneviratne, S. I. (2020). Soil moisture dominates dryness stress on ecosystem production globally. Nature Communications, 11(1). https://doi.org/10.1038/s41467-020-18631-1 Liu, Q., Peng, C., Schneider, R., Cyr, D., Liu, Z., Zhou, X., & Kneeshaw, D. (2021). TRIPLEX-Mortality model for simulating drought-induced tree mortality in boreal forests: Model development and evalua- tion. Ecological Modelling, 455, 109652. Manion, P. D. (1981). Tree disease concepts. Prentice-Hall, Inc. Manusch, C., Bugmann, H., & Wolf, A. (2014). Sensitivity of simulated productivity to soil characteristics and plant water uptake along drought gradients in the Swiss Alps. Ecological Modelling, 282, 25–34. https://doi.org/10.1016/j.ecolmodel.2014.03.006 Bland, J. M., & Altman, D. (1986). Statistical methods for assessing agreement between two methods of clin- ical measurement. The Lancet, 327(8476), 307-310. Martínez-Vilalta, J., Piñol, J., & Beven, K. (2002). A hydraulic model to predict drought-induced mortality in woody plants: An application to climate change in the Mediterranean. Ecological Modelling, 155(2–3), 127–147. https://doi.org/10.1016/S0304-3800(02)00025-X Massonnet, C., Chuste, P.-A., Levillain, J., Gérémia, F., Silva, D.E., Maillard, P., Dreyer, E., Dupouey, J.-L., Bréda, N. (2021). Leafy season length is reduced by a prolonged soil water deficit but not by repeated defoliation in beech trees (Fagus sylvatica L.): Comparison of response among regional populations grown in a common garden. Agricultural and Forest Meteorology 297, 108228. https://doi.org/10.1016/j.agrformet.2020.108228. McDowell, N., Pockman, W. T., Allen, C. D., Breshears, D. D., Cobb, N., Kolb, T., Plaut, J., Sperry, J., West, A., Williams, D. G., & Yepez, E. A. (2008). Mechanisms of plant survival and mortality during drought: Why do some plants survive while others succumb to drought? In New Phytologist (Vol. 178, Issue 4, pp. 719–739). https://doi.org/10.1111/j.1469-8137.2008.02436.x McDowell, N. G. (2011). Mechanisms linking drought, hydraulics, carbon metabolism, and vegetation mor- tality. Plant Physiology, 155(3), 1051–1059. https://doi.org/10.1104/pp.110.170704 McDowell, N. G., Fisher, R. A., Xu, C., Domec, J. C., Hölttä, T., Mackay, D. S., Sperry, J. S., Boutz, A., Dickman, L., Gehres, N., Limousin, J. M., Macalady, A., Martínez-Vilalta, J., Mencuccini, M., Plaut, J. A., Ogée, J., Pangle, R. E., Rasse, D. P., Ryan, M. G., … Pockman, W. T. (2013). Evaluating theories of drought-induced vegetation mortality using a multimodel-experiment framework. New Phytologist, 200(2), 304–321. https://doi.org/10.1111/nph.12465 McKee, B. T., Dorsken, N., & Kleist, J. (1993). The relationship of drought frequency and duration to time scales. Eighth Conference on Applied Climatology, 179--183. https://doi.org/10.1002/jso.23002 Meir, P., Mencuccini, M., & Dewar, R. C. (2015). Drought-related tree mortality: Addressing the gaps in un- derstanding and prediction. In New Phytologist (Vol. 207, Issue 1, pp. 28–33). https://doi.org/10.1111/nph.13382 Merganičová, K. (2023). Personal communication, 22.09.2023 MeteoSwiss. (2023). IDAWEB portal, http://www.meteoschweiz. admin.ch/web/en/services/data_por- tal/idaweb.html. MeteoTest. (2020). Klimadaten Waldforschung: Berechnung von historischen Meteo- und Trocken-heitsindi- zes für die Waldforschung auf der Basis von SwissMetNet-Stationsdaten. Meusburger, K., Trotsiuk, V., Schmidt-Walter, P., Baltensweiler, A., Brun, P., Bernhard, F., Gharun, M., Ha- bel, R., Hagedorn, F., Köchli, R., Psomas, A., Puhlmann, H., Thimonier, A., Waldner, P., Zimmermann, S., & Walthert, L. (2022). Soil–plant interactions modulated water availability of Swiss forests during the 2015 and 2018 droughts. Global Change Biology, 28(20), 5928–5944. https://doi.org/10.1111/gcb.16332 Michelot, A., Bréda, N., Damesin, C., Dufrêne, E. (2012). Differing growth responses to climatic variations and soil water deficits of Fagus sylvatica, Quercus petraea and Pinus sylvestris in a temperate forest. Forest Ecology and Management 265, 161–171. https://doi.org/10.1016/j.foreco.2011.10.024 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 39 Millar, C. I., & Stephenson, N. L. (2015). Temperate forest health in an era of emerging megadisturbance. Science, 349(6250), 823–826. https://doi.org/10.1126/science.aaa9933 Moorcroft, P. R., Hurtt, G. C., & Pacala, S. W. (2001). A method for scaling vegetation dynamics: the eco- system demography model (ED). Ecological monographs, 71(4), 557-586. Myneni, R., Knyazikhin, Y., & Park, T. (2021). MODIS/Terra Leaf Area Index/FPAR 8-Day L4 Global 500m SIN Grid V061 [Data set]. NASA EOSDIS Land Processes Distributed Active Archive Center. Nadal-Sala, D., Keenan, T. F., Sabaté, S., & Gracia, C. (2017). Forest eco-physiological models: water use and carbon sequestration. Managing forest ecosystems: the challenge of climate change, 81-102. Neely, D., & Manion, P. D. (1991). Tree Disease Concepts. Mycologia, 83(5). https://doi.org/10.2307/3760234 Neycken, A., Scheggia, M., Bigler, C., & Lévesque, M. (2022). Long-term growth decline precedes sudden crown dieback of European beech. Agricultural and Forest Meteorology, 324(July). https://doi.org/10.1016/j.agrformet.2022.109103 Obladen, N., Dechering, P., Skiadaresis, G., Tegel, W., Keßler, J., Höllerl, S., Kaps, S., Hertel, M., Dulamsu- ren, C., Seifert, T., Hirsch, M., & Seim, A. (2021). Tree mortality of European beech and Norway spruce induced by 2018-2019 hot droughts in central Germany. Agricultural and Forest Meteorology, 307. https://doi.org/10.1016/j.agrformet.2021.108482 Orth, R., Staudinger, M., Seneviratne, S. I., Seibert, J., & Zappa, M. (2015). Does model performance im- prove with complexity? A case study with three hydrological models. Journal of Hydrology, 523, 147– 159. https://doi.org/10.1016/j.jhydrol.2015.01.044 Ovenden, T. S., Perks, M. P., Clarke, T. K., Mencuccini, M., & Jump, A. S. (2021). Life after recovery: In- creased resolution of forest resilience assessment sheds new light on post-drought compensatory growth and recovery dynamics. Journal of Ecology, 109(9), 3157–3170. https://doi.org/10.1111/1365- 2745.13576 Parolari, A. J., Katul, G. G., & Porporato, A. (2014). An ecohydrological perspective on drought-induced for- est mortality. Journal of Geophysical Research: Biogeosciences, 119(5), 965–981. https://doi.org/10.1002/2013JG002592 Pedersen, B. S. (1998). Modeling tree mortality in response to short-and long-term environmental stresses. In Ecological Modelling (Vol. 105). Peltier, D. M. P., Carbone, M. S., McIntire, C. D., Robertson, N., Thompson, R. A., Malone, S., LeMoine, J., Richardson, A. D., McDowell, N. G., Adams, H. D., Pockman, W. T., & Trowbridge, A. M. (2023). Carbon starvation following a decade of experimental drought consumes old reserves in Pinus edulis. New Phytologist. https://doi.org/10.1111/nph.19119 Potapov, P., Li, X., Hernandez-Serna, A., Tyukavina, A., Hansen, M. C., Kommareddy, A., Pickens, A., Turubanova, S., Tang, H., Silva, C. E., Armston, J., Dubayah, R., Blair, J. B., & Hofton, M. (2020). Mapping global forest canopy height through integration of GEDI and Landsat data. Remote Sensing of Environment. https://doi.org/10.1016/j.rse.2020.112165 Rakovec, O., Samaniego, L., Hari, V., Markonis, Y., Moravec, V., Thober, S., Hanel, M., & Kumar, R. (2022). The 2018–2020 Multi-Year Drought Sets a New Benchmark in Europe. Earth’s Future, 10(3). https://doi.org/10.1029/2021EF002394 Rebetez, M., & Dobbertin, M. (2004). Climate change may already threaten Scots pine stands in the Swiss Alps. Theoretical and Applied Climatology, 79(1–2), 1–9. https://doi.org/10.1007/s00704-004-0058-3 Rigling, A., & Cherubini, P. (1999). Wieso sterben die Waldföhren im «TeIwald» bei Visp? Eine Zusammen- fassung bisheriger Studien und eine dendroökologische Untersuchung | What is the Cause of the High Mortality Rates of the Scots Pines in the «Telwald» near Visp (Switzerland)? A Summary of Previous Studies and a Dendroecological Study. Schweizerische Zeitschrift Fur Forstwesen, 150(4), 113–131. https://doi.org/10.3188/szf.1999.0113 Romano, N., Palladino, M., & Chirico, G. B. (2011). Parameterization of a bucket model for soil-vegetation- atmosphere modeling under seasonal climatic regimes. Hydrology and Earth System Sciences, 15(12), 3877–3893. https://doi.org/10.5194/hess-15-3877-2011 Roy, M. È., Surget-Groba, Y., Delagrange, S., & Rivest, D. (2021). Legacies of forest harvesting on soil properties along a chronosequence in a hardwood temperate forest. Forest Ecology and Management, 496. https://doi.org/10.1016/j.foreco.2021.119437 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 40 Ruehr, N. K., Martin, J. G., & Law, B. E. (2012). Effects of water availability on carbon and water exchange in a young ponderosa pine forest: Above- and belowground responses. Agricultural and Forest Meteor- ology, 164, 136–148. https://doi.org/https://doi.org/10.1016/j.agrformet.2012.05.015 Ruosteenoja, K., Markkanen, T., Venäläinen, A., Peltola, H. (2018). Seasonal soil moisture and drought oc- currence in Europe in CMIP5 projections for the 21st century. Climate Dynamics 50, 1177–1192. https://doi.org/10.1007/s00382-017-3671-4. Sala, A., Woodruff, D. R., & Meinzer, F. C. (2012). Carbon dynamics in trees: feast or famine?. Tree physio- logy, 32(6), 764-775. Schmidt-Walter, P., Trotsiuk, V., Meusburger, K., Zacios, M., & Meesenburg, H. (2020). Advancing simula- tions of water fluxes, soil moisture and drought stress by using the LWF-Brook90 hydrological model in R. Agricultural and Forest Meteorology, 291. https://doi.org/10.1016/j.agrformet.2020.108023 Schmied, G., Pretzsch, H., Ambs, D., Uhl, E., Schmucker, J., Fäth, J., Biber, P., Hoffmann, Y. D., Šeho, M., Mellert, K. H., & Hilmers, T. (2023). Rapid beech decline under recurrent drought stress: Individual neighborhood structure and soil properties matter. Forest Ecology and Management, 545. https://doi.org/10.1016/j.foreco.2023.121305 Schwalm, C. R., Anderegg, W. R. L., Michalak, A. M., Fisher, J. B., Biondi, F., Koch, G., Litvak, M., Ogle, K., Shaw, J. D., Wolf, A., Huntzinger, D. N., Schaefer, K., Cook, R., Wei, Y., Fang, Y., Hayes, D., Huang, M., Jain, A., & Tian, H. (2017). Global patterns of drought recovery. Nature. https://doi.org/10.1038/nature23021 Schwarz, J., Skiadaresis, G., Kohler, M., Kunz, J., Schnabel, F., Vitali, V., & Bauhus, J. (2020). Quantifying Growth Responses of Trees to Drought—a Critique of Commonly Used Resilience Indices and Recom- mendations for Future Studies. In Current Forestry Reports (Vol. 6, Issue 3, pp. 185–200). Springer. https://doi.org/10.1007/s40725-020-00119-2 Senf, C., & Seidl, R. (2021). Persistent impacts of the 2018 drought on forest disturbance regimes in Europe. Biogeosciences, 18(18), 5223–5230. https://doi.org/10.5194/bg-18-5223-2021 Sevanto, S., McDowell, N., Dickman, L., Pangle, R., & Pockman, W. (2014). How do trees die? A test of the hydraulic failure and carbon starvation hypotheses. Plant, Cell & Environment, 37, 153–161. https://doi.org/10.1111/pce.12141 Speich, M. (2019). Quantifying and modeling water availability in temperate forests: a review of drought and aridity indices. IForest - Biogeosciences and Forestry, 12(1), 1–16. https://doi.org/10.3832/ifor2934- 011 Speich, M. J. R., Zappa, M., & Lischke, H. (2018). Sensitivity of forest water balance and physiological drought predictions to soil and vegetation parameters – A model-based study. Environmental Modelling and Software, 102, 213–232. https://doi.org/10.1016/j.envsoft.2018.01.016 Steinkamp, J., & Hickler, T. (2015). Is drought-induced forest dieback globally increasing? Journal of Ecol- ogy, 103(1), 31–43. https://doi.org/10.1111/1365-2745.12335 Swisstopo. (2004). The digital height model of Switzerland Product information. http://www.swis- stopo.ch/en/pro Thornthwaite, C. W. (1948). An Approach toward a Rational Classification of Climate. Geographical Re- view, 38(1). https://doi.org/10.2307/210739 Thornthwaite, C. W., & Mather, J. R. (1954). Climate in Relation to Crops. In Recent Studies in Bioclimatol- ogy. https://doi.org/10.1007/978-1-940033-11-2_1 Trugman, A. T., Anderegg, L. D. L., Anderegg, W. R. L., Das, A. J., & Stephenson, N. L. (2021). Why is Tree Drought Mortality so Hard to Predict? Trends in Ecology and Evolution (Vol. 36, Issue 6, pp. 520– 532). Elsevier Ltd. https://doi.org/10.1016/j.tree.2021.02.001 NOA (2019). United States Drought Monitor. University of Nebraska-Lincoln, National Drought Mitigation Center; U.S. Department of Agriculture van der Maaten-Theunissen, M., Trouillier, M., Schwarz, J., Skiadaresis, G., Thurm, E. A., & van der Maaten, E. (2021). pointRes 2.0: New functions to describe tree resilience. Dendrochronologia, 70. https://doi.org/10.1016/j.dendro.2021.125899 Vicente-Serrano, S. M., Beguería, S., Lorenzo-Lacruz, J., Camarero, J. J., López-Moreno, J. I., Azorin-Mo- lina, C., Revuelto, J., Morán-Tejeda, E., & Sanchez-Lorenzo, A. (2012). Performance of drought indi- ces for ecological, agricultural, and hydrological applications. Earth Interactions, 16(10). https://doi.org/10.1175/2012EI000434.1 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 41 Vilhar, U. (2016). Comparison of drought stress indices in beech forests: A modelling study. IForest, 9(4), 635–642. https://doi.org/10.3832/ifor1630-008 Vitali, V., Büntgen, U., & Bauhus, J. (2017). Silver fir and Douglas fir are more tolerant to extreme droughts than Norway spruce in south-western Germany. Global Change Biology, 23(12), 5108– 5119. https://doi.org/10.1111/gcb.13774 Vitasse, Y., Bottero, A., Cailleret, M., Bigler, C., Fonti, P., Gessler, A., Lévesque, M., Rohner, B., Weber, P., Rigling, A., & Wohlgemuth, T. (2019). Contrasting resistance and resilience to extreme drought and late spring frost in five major European tree species. Global Change Biology, 25(11), 3781–3792. https://doi.org/10.1111/gcb.14803 Wang, W., Peng, C., Kneeshaw, D. D., Larocque, G. R., & Luo, Z. (2012). Drought-induced tree mortality: Ecological consequences, causes, and modeling. In Environmental Reviews (Vol. 20, Issue 2, pp. 109– 121). https://doi.org/10.1139/a2012-004 Wells, N., Goddard, S., & Hayes, M. J. (2004). A Self-Calibrating Palmer Drought Severity Index. Journal of Climate, 17(12), 2335–2351. https://doi.org/10.1175/1520-0442(2004)0172.0.CO;2 Wessolek, G., Kaupenjohann, M., & Renger, M. (2009). Bodenökologie und Bodengenese. Bodenphysikali- sche Kennwerte und Berechnungsverfahren für die Praxis. Bodenökologie und Bodengenese, 40. Wohlgemuth, T., Doublet, V., Nussbaumer, C., Feichtinger, L., & Rigling, A. (2018). Vegetation shift in Scots pine forests in the Valais accelerated by large disturbances. In Epidemiologia e Prevenzione (Vol. 169, Issue 5, pp. 260–268). Inferenze Scarl. https://doi.org/10.3188/szf.2018.0260 Wohlgemuth, T., Kistler, M., Aymon, C., Hagedorn, F., Gessler, A., Gossner, M. M., Queloz, V., Vögtli, I., Wasem, U., Vitasse, Y., & Rigling, A. (2020). Early leaf fall of european beech during the hot and dry summer of 2018: Resistance or sign of weakness? Schweizerische Zeitschrift Fur Forstwesen, 171(5), 257–269. https://doi.org/10.3188/szf.2020.0257 Zamora-Pereira, J. C., Yousefpour, R., Cailleret, M., Bugmann, H., & Hanewinkel, M. (2021). Magnitude and timing of density reduction are key for the resilience to severe drought in conifer-broadleaf mixed forests in Central Europe. Annals of Forest Science, 78(3), 68. https://doi.org/10.1007/s13595-021- 01085-w Zang, C. (2018). (sc)pdsi R wrapper function. https://github.com/cszang/pdsi Zargar, A., Sadiq, R., Naser, B., & Khan, F. I. (2011). A review of drought indices. In Environmental Re- views (Vol. 19, Issue 1, pp. 333–349). National Research Council of Canada. https://doi.org/10.1139/a11-013 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 42 Tables Table 1 | Summary of site -specific properties for the six beech sites (site acronym, S Id) on the Swiss Plateau: annual mean temperature (T, °C), annual precipitation sum (P, mm) for the period 1981 -2022 (data source: MeteoTest) and available water capacity (AWC, cm) estimated using pedotransfer functions. Site SId T (mm) P (°C) AWC (mm) Blattenberg BB 9.3 1325.8 86 Grosszinggibrunn GZ 9.6 980.5 154 Rossberg RB 9.3 914.3 155 Tüeliboden TB 9.7 1041.9 140 Usserholz UH 9.4 1076.2 112 Vogtacher VA 9.7 996.1 137 Table 2 | Summary statistics of observed vs. simulated AET at the site Lägeren across the years 2012 - 2020 (observational data from Meusburger et al., 2022). MSE = Mean Squared Error; MAE = Mean ab- solute Error; RMSE = Root Mean Squared Error. Model MSE MAE RMSE ForClim 3.33 1.39 1.82 LWFBrook90 5.74 1.61 2.4 Table 3 | Ratio of Cumulative Soil Moisture (CSM) between the ForClim and LWFBrook90 models (ForClim:LWFBrook90) for each year until the month of December, averaged across the years 2000 - 2020 along with standard deviations. Site CSMForClim:LWFBrook90 Blattenberg 0.86 ± 0.09 Grosszinggibrunn 1.28 ± 0.12 Rossberg 1.24 ± 0.12 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 43 Tüeliboden 1.17 ± 0.12 Usserholz 1.3 ± 0.14 Vogtacher 1.3 ± 0.11 Table 4 | Model statistics for the years 2018, 2019 and 2020 (data source for inventory: Neycken et al., 2022) and simulated basal area losses at the six beech sites expressed as the percent changes from the long-term mean (2000-2015 for Rossberg and Tüeliboden, 2000-2016 for Grosszinggibrunn, Vogtacher and Usserholz, 2000-2017 for Blattenberg). Site MAE RMSE Year ΔBAsim (%) Blattenberg 5.97 6.05 2018 7.9 2019 8.8 2020 9.7 Usserholz 2.34 2.37 2018 -14.4 2019 -12.3 2020 -11.0 Tüeliboden 2.55 2.62 2018 -25.8 2019 -24.7 2020 -24.2 Rossberg 0.44 0.62 2018 -7.9 2019 -7.0 2020 -6.4 Grosszinggi- brunn 4.23 4.27 2018 -9.5 2019 -7.6 2020 -4.9 Vogtacher 4.75 4.78 2018 -7.8 2019 -6.1 2020 -5.1 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 44 Figures Figure 1: Study sites classified as main sites (red dots) where drought-induced mortality in 2018- 2020 was observed, the ICP-Level II plots (abbr. ICP, blue triangles) and sites where Potential Natural Vegetation was simulated along an elevational and climatic gradient (Bugmann & Solo- mon, 2000 ) in Switzerland and Germany (black dots). Elevational map of Switzerland from Swisstopo (2014). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 45 Figure 2: Inciting factor scheme showing the stress factors operating in the short term (one year): during the growing period, the formulation selects only those months in which (1) water demand (ratio of Evapotranspiration E on the water Demand D, unitless) is sub-optimal (<0.9), and (2) soil moisture falls below a threshold in spring (0.9) and fall (0.5). Both axes are unitless. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 46 Figure 3: Seasonal Spearman rank correlation between detrended tree-ring indices and detrended drought indices of increasing complexity (from indices including precipitation (P) only (SPI) to precipitation and temperature (CWB, SPEI) to precipitation, temperature and available water .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 47 capacity (AWC; scPSDSI, ForClim). The significance levels (p-values) are indicated by symbols ("***", "**", "*", "ns" from 10-3 to >0.05). When no variance was observed, an intercept y=0 was used (i.e. at Grosszinggibrunn, Tüeliboden and Vogtacher for the ForClim drought index). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 48 Figure 4: (A) Simulated and observed monthly actual evapotranspiration (AET) at the ICP Level II site Lägeren (Switzerland). The trend lines in panel A indicate a progressive reduction in simulated and measured AET from 2015. Both models showed a slight negative temporal trend in simulated AET. (B) Differences between simulated and observed AET across years. Observed data are indicated with a black line (A) and a dashed line (B), respectively. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 49 Figure 5:Simulated soil moisture with LWFBrook90 and ForClim. (A) Monthly variation of simulated soil moisture at the site of Tüeliboden for the model ForClim (black line and black dots ) and LWFBrook90 (grey lines and grey dots). The highlighted red areas indicate the driest years 2003, 2015 and 2018. (B) Differences of Cumulative Soil Moisture (CSM, cm on the y-axis) in Bland-Altmann plots for the driest years (2003, .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 50 2015, 2018) at the six beech-dominated sites (dots with different shapes) simulated with ForClim and LWFBrook90. The x-axis shows the mean of each pair of values among the two model s. The limits of agreement are indicated with the blue dashed intercepts. The intercept on the y-axis indicated by the red dashed line in Figure 5B shows the mean bias. Positive deviations from the mean bias between ForClim and LWFBrook90 indicate a higher simulated soil moisture for ForClim compared to LWFBrook90. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 51 Figure 6: Simulated basal area of Fagus sylvatica over time and observed basal area (black dots) in the years 2018-2020 (Neycken et al., 2022) with the newly developed predisposing and incit- ing factor scheme (ForClim 4.1) and the earlier model version (ForClim 4.0.1). The dashed lines indicate the drought years 2003, 2015, 2018 and 2019. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 52 Figure 7: Simulated basal area (A) and mortality rate (B) over the years 1997-2004 at the site Visp for Scots pine and Quercus pubescens. The black dots indicate the observed mortality rates according to previous studies (Hunziker et al. 2022). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 53 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 54 Figure 8: ForClim model version comparison along the EU gradient of sites. (A) Basal area (BA) at the equilibrium of potential natural vegetation (PNV) simulated with ForClim version 4.1 and (B) and with the older ForClim version 4.0.1 across 12 sites in Switzerland and Germany (Bug- mann, 1996 b). Species with BA ≤ 2 m2 ha-1 are summarized in the “various sp.” category. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 55 Supplementary Material SM A: Climatic anomalies and site inventories Figure A1: Precipitation and temperature anomalies for the years 2003, 2015 and 2018 relative to the reference period 1980- 2010 across sites (left panel) and averaged across the six beech sites (right panel). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 56 Table A1: Measured basal area (m2 ha-1) at the six beech sites from 2015 to 2020, as retrieved from Neycken et al. (2022). Site Time (yr) Basal area (m2 ha-1) Usserholz 2016 25.8 ± 1.7 2018 20.0 ± 5.6 2019 20.2 ± 5.5 2020 19.9 ± 5.6 Vogtacher 2016 23.0 ± 1.4 2018 15.0 ± 4.6 2019 14.2 ± 4.2 2020 14.5 ± 4.4 Tüeliboden 2015 31.5 ± 1.2 2018 20.7 ± 6.4 2019 20.9 ± 7.8 2020 19.9 ± 7.4 Rossberg 2015 28.1 ± 1.0 2018 25.0 ± 9.2 2019 25.0 ± 8.7 2020 24.3 ± 9.6 Grosszinggibrunn 2016 19.7 ± 1.5 2018 12.4 ± 4.6 2019 11.6 ± 4.1 2020 12.0 ± 3.6 Blattenberg 2017 33.0 ± 1.0 2018 27.7 ± 6.0 2019 28.6 ± 6.2 2020 26.5 ± 5.7 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 57 SM B: Tree-ring data and drought indices B1: Ring-width indices Fig B1.1: Ring-width Index (RWI) standard and residual chronologies of the six sites. Red dashed lines indicate the drought years 2003, 2011, 2015, 2018-2019. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 58 Fig B1.2: Tree-ring chronologies for the six beech sites. RES stands for residual chronology, which was obtained using auto-regressive models (AR) of the individual tree series, while STD indicates the standard chronology. The shaded area represents the sample depth. The main spline is indicated by the red line (i.e., low-pass filter). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 59 B2: Z-chronologies for the identification of pointer years Fig B2: Calculation of pointer years for the six sites using a z-transformation of the site chronologies (Tukey’s biweight robust mean site chronology). The dashed lines indicated dry years 2003, 2011, 2015, 2018 and 2019. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 60 B3: Tree-rings indices and drought indices Table B3.1: Mann-Kendall seasonal test for the five drought indices at the six beech sites. Site p-value τ Drought Index significant Blattenberg 0.073690743 -0.053947096 CWB FALSE Grosszinggibrunn 0.218507857 -0.037115382 CWB FALSE Rossberg 0.464476155 -0.022063907 CWB FALSE Tüeliboden 0.105823985 -0.048780892 CWB FALSE Usserholz 0.156277559 -0.042762109 CWB FALSE Vogtacher 0.313238929 -0.030417431 CWB FALSE Blattenberg 0.215672487 -0.037384032 SPI FALSE Grosszinggibrunn 0.527478963 -0.019058972 SPI FALSE Rossberg 0.8687562 -0.004984186 SPI FALSE Tüeliboden 0.180862792 -0.040362473 SPI FALSE Usserholz 0.223478476 -0.036718589 SPI FALSE Vogtacher 0.532513542 -0.018827465 SPI FALSE Blattenberg 0.062978216 -0.056083123 SPEI FALSE Grosszinggibrunn 0.06721257 -0.05520507 SPEI FALSE Rossberg 0.308520949 -0.030715478 SPEI FALSE Tüeliboden 0.053881759 -0.063989469 SPEI FALSE Usserholz 0.040411808 -0.059270359 SPEI FALSE Vogtacher 0.117173918 -0.047257129 SPEI FALSE Blattenberg 0 -0.348273502 sc-PDSI TRUE Grosszinggibrunn 1.97373E-09 -0.183332548 sc-PDSI TRUE Rossberg 6.24286E-06 -0.138075743 sc-PDSI TRUE Tüeliboden 3.21965E-14 -0.231929032 sc-PDSI TRUE Usserholz 1.3468E-10 -0.19622966 sc-PDSI TRUE Vogtacher 1.91588E-08 -0.171724536 sc-PDSI TRUE Blattenberg 0.271524187 0.039440173 ForClim FALSE Grosszinggibrunn 0.251986335 0.041176087 ForClim FALSE Rossberg 0.526524962 0.022411501 ForClim FALSE Tüeliboden 0.313488755 0.036149312 ForClim FALSE Usserholz 0.828500556 -0.007769694 ForClim FALSE Vogtacher 0.354662213 0.03323285 ForClim FALSE .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 61 Table B3.2: Mann-Kendall test on the detrended sc-PDSI index by applying the STL method Site p-value τ Drought Index significant Blattenberg 0.058372779 -0.059543321 sc-PDSI FALSE Grosszinggibrunn 0.88910471 -0.004205785 sc-PDSI FALSE Rossberg 0.119632891 -0.046942523 sc-PDSI FALSE Tüeliboden 0.543383497 0.018329938 sc-PDSI FALSE Usserholz 0.548127808 -0.018114681 sc-PDSI FALSE Vogtacher 0.686588218 -0.012170284 sc-PDSI FALSE .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 62 Figure B3: Seasonal non-parametric correlations between tree-ring width indices and drought indices of previous year (lag=1). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 63 SM C: LWFBrook90 and ForClim model comparison C1: Lägeren simulated and observed soil moisture Fig C1: Difference between observed and simulated (ForClim) soil moisture in the year 2011. The dashed black line indicates the reference (observed data) indicating no difference between observed and simulated soil moisture. Observed data were re- trieved from Pastorello, G., Trotta, C., Canfora, E. et al. The FLUXNET2015 dataset and the ONEFlux processing pipeline for eddy covariance data. Sci Data 7, 225 (2020). https://doi.org/10.1038/s41597-020-0534-3 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 64 C2: Simulation setup for LWFBrook90 Table C2.1 | Parameter set LWFBrook90 for the Lägeren. Sid: site id, Sp: species, El: elevation (m a.s.l), Sl: slope, As: aspect, Bas: basal area (m2 ha-1), Ag: age (years), He: height (m), Lm: maximum LAI (m2 m-2). SId Lat Lon Sp El Sl As Bas Ag He Lm GZ 47.51 7.67 Fagus sylvatica 407.3 0.26 0.37 11.64 99 22.3 5 VA 47.51 7.7 Fagus sylvatica 467.2 0.24 6.03 14.22 119 24.7 5 TB 47.54 8.21 Fagus sylvatica 468.3 0.28 2.67 19.9 128 23.8 5 UH 47.48 7.48 Fagus sylvatica 570.8 0.12 2.72 19.91 132 22 6 BB 47.31 9.55 Fagus sylvatica 539.49 0.39 1.48 26.57 122 24.3 5 RB 47.67 8.57 Fagus sylvatica 515.1 0.09 2.55 24.29 122 24.6 5 Table C2.2: Parameter set for LWFBrook90 simulations for the six beech -dominated sites ( site id, SId ). For a detailed description of each parameter consult Schmidt-Walter et al. (2020). Parameter SId RB GZ BB UH VA TB maxlai 5 5 5 6 5 5 sai 1 1 1 1 1 1 sai_ini 1 1 1 1 1 1 height 25 22.3 24.3 22 24.7 23.8 densef 1 1 1 1 1 1 densef_ini 1 1 1 1 1 1 age_ini 122 99 122 132 119 128 winlaifrac 0 0 0 0 0 0 budburst_spe- cies Fagus syl- vatica Fagus syl- vatica Fagus syl- vatica Fagus syl- vatica Fagus syl- vatica Fagus syl- vatica budburstdoy 121 121 121 121 121 121 leaffalldoy 279 279 279 279 279 279 shp_budburst 0.3 0.3 0.3 0.3 0.3 0.3 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 65 shp_leaffall 3 3 3 3 3 3 shp_optdoy 210 210 210 210 210 210 emergedur 28 28 28 28 28 28 leaffalldur 58 58 58 58 58 58 alb 0.2 0.2 0.2 0.2 0.2 0.2 albsn 0.5 0.5 0.5 0.5 0.5 0.5 ksnvp 0.3 0.3 0.3 0.3 0.3 0.3 fxylem 0.5 0.5 0.5 0.5 0.5 0.5 mxkpl 8 8 8 8 8 8 lwidth 0.1 0.1 0.1 0.1 0.1 0.1 psicr -2.8 -2.8 -2.8 -2.8 -2.8 -2.8 nooutf 1 1 1 1 1 1 lpc 4 4 4 4 4 4 cs 0.035 0.035 0.035 0.035 0.035 0.035 czs 0.13 0.13 0.13 0.13 0.13 0.13 czr 0.05 0.05 0.05 0.05 0.05 0.05 hs 1 1 1 1 1 1 hr 10 10 10 10 10 10 rhotp 2 2 2 2 2 2 nn 2.5 2.5 2.5 2.5 2.5 2.5 maxrlen 3000 3000 3000 3000 3000 3000 initrlen 12 12 12 12 12 12 initrdep 0.25 0.25 0.25 0.25 0.25 0.25 rrad 0.35 0.35 0.35 0.35 0.35 0.35 rgrorate 0.03 0.03 0.03 0.03 0.03 0.03 rgroper 0 0 0 0 0 0 maxrootdepth -1.5 -1.5 -1.5 -1.5 -1.5 -1.5 betaroot 0.97 0.97 0.97 0.97 0.97 0.97 radex 0.5 0.5 0.5 0.5 0.5 0.5 glmax 0.0053 0.0053 0.0053 0.0053 0.0053 0.0053 glmin 3.00E-04 3.00E-04 3.00E-04 3.00E-04 3.00E-04 3.00E-04 rm 1000 1000 1000 1000 1000 1000 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 66 r5 100 100 100 100 100 100 cvpd 2 2 2 2 2 2 tl 0 0 0 0 0 0 t1 10 10 10 10 10 10 t2 30 30 30 30 30 30 th 40 40 40 40 40 40 frintlai 0.06 0.06 0.06 0.06 0.06 0.06 frintsai 0.06 0.06 0.06 0.06 0.06 0.06 fsintlai 0.04 0.04 0.04 0.04 0.04 0.04 fsintsai 0.04 0.04 0.04 0.04 0.04 0.04 cintrl 0.15 0.15 0.15 0.15 0.15 0.15 cintrs 0.15 0.15 0.15 0.15 0.15 0.15 cintsl 0.6 0.6 0.6 0.6 0.6 0.6 cintss 0.6 0.6 0.6 0.6 0.6 0.6 infexp 0 0 0 0 0 0 bypar 0 0 0 0 0 0 qfpar 1 1 1 1 1 1 qffc 0 0 0 0 0 0 imperv 0 0 0 0 0 0 drain 1 1 1 1 1 1 gsc 0 0 0 0 0 0 gsp 0 0 0 0 0 0 ilayer 1 1 1 1 1 1 qlayer 0 0 0 0 0 0 z0s 0.001 0.001 0.001 0.001 0.001 0.001 rstemp -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 melfac 1.5 1.5 1.5 1.5 1.5 1.5 ccfac 0.3 0.3 0.3 0.3 0.3 0.3 laimlt 0.2 0.2 0.2 0.2 0.2 0.2 saimlt 0.5 0.5 0.5 0.5 0.5 0.5 grdmlt 0.35 0.35 0.35 0.35 0.35 0.35 maxlqf 0.05 0.05 0.05 0.05 0.05 0.05 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 67 snoden 0.3 0.3 0.3 0.3 0.3 0.3 obsheight 0.025 0.025 0.025 0.025 0.025 0.025 cor- rect_prec_sta- texp mg mg mg mg mg mg rssa 100 100 100 100 100 100 rssb 0 0 0 0 0 0 dtimax 0.5 0.5 0.5 0.5 0.5 0.5 dswmax 0.05 0.05 0.05 0.05 0.05 0.05 dpsimax 5.00E-04 5.00E-04 5.00E-04 5.00E-04 5.00E-04 5.00E-04 wndrat 0.3 0.3 0.3 0.3 0.3 0.3 fetch 5000 5000 5000 5000 5000 5000 z0w 0.005 0.005 0.005 0.005 0.005 0.005 zw 2 2 2 2 2 2 zminh 2 2 2 2 2 2 coords_x 9.9095 9.9095 9.9095 9.9095 9.9095 9.9095 coords_y 47.66597 47.51021 47.30874 47.4789 47.50791 47.53836 c1 0.25 0.25 0.25 0.25 0.25 0.25 c2 0.5 0.5 0.5 0.5 0.5 0.5 c3 0.2 0.2 0.2 0.2 0.2 0.2 pdur1 4 4 4 4 4 4 pdur2 4 4 4 4 4 4 pdur3 4 4 4 4 4 4 pdur4 4 4 4 4 4 4 pdur5 4 4 4 4 4 4 pdur6 4 4 4 4 4 4 pdur7 4 4 4 4 4 4 pdur8 4 4 4 4 4 4 pdur9 4 4 4 4 4 4 pdur10 4 4 4 4 4 4 pdur11 4 4 4 4 4 4 pdur12 4 4 4 4 4 4 eslope 0.0934714 0.2566277 0.3871424 0.1191926 0.2372652 0.2798152 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 68 7 7 8 6 aspect 2.5535829 1 0.3663539 1.4802918 4 2.7246344 1 6.0281777 4 2.6662714 5 dslope 0 0 0 0 0 0 slopelen 200 200 200 200 200 200 intrainini 0 0 0 0 0 0 intsnowini 0 0 0 0 0 0 gwatini 0 0 0 0 0 0 snowini 0 0 0 0 0 0 psiini -6.3 -6.3 -6.3 -6.3 -6.3 -6.3 Fig C2.1: Simulated monthly soil moisture at the six beech sites. Drought years 2003, 2015 and 2018 are highlighted in red. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 69 Fig C2.2: Soil moisture differences between ForClim and LWFBrook90 across seasons (upper panel) and yearly means (lower panel) at the six beech sites. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 70 Fig C2.3: Annual cumulative AET, PET and DI (AET/PET) from the two models at the six beech sites. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 71 Figure C2.4: Bland-Altmann plots of yearly cumulated soil moisture (CSM) across the six beech sites of the two models for the for the distinct years of the simulation period. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 72 SM D: Description of new features in ForClim 4.1 D1 Distributed soil properties per patch (kBS distribution) Soils feature high spatial variability in forests as the result of the interplay of geology, geomorphology, vegetation, and time (Binkley & Fisher, 2013; Boyle, 2005; Osman, 2013) . Notably, one of the most influential factors in this context is soil depth, determined by weathering and affecting nutrient availabil- ity. Within this mosaic of soil properties, hydraulic properties pose a notable source of uncertainty, inher- ently linked to physical attributes. This uncertainty can be interpreted as being stochastic because there is no soil data at meter resolution across entire forest stands. Understanding and quantifying this apparent stochasticity is pivotal for accurately characterizing soil properties, with implications for ecosystem func- tioning, hydrological processes, and sustainable land management practices in forested environments. To accurately capture the influence of drought within specific time frames, such as the drought events of 2003, 2015, and 2018, time series of precipitation and temperature need to be used. Such time series provide a detailed account of weather conditions, allowing us to accurately model and analyze the drought signals provided that we can capture the spatial variability of soil physical properties. To achieve this, we attributed distinct soil properties, particularly soil water availability, to each forest patch of ForClim, thereby mimicking the heterogeneity present in natural ecosystems. This approach ensured that the sim- ulation experiments appropriately reflected the intricate interplay between drought, weather patterns, and spatial heterogeneity in soil properties across forest stands. The lognormal probability distribution is often used to model skewed and left-truncated variables (such as the “bucket size” of ForClim). It was in this case selected to attribute to each forest patch its own soil properties. In the ForClim site file, values for kBSmean and kBSmin must be provided, and the standard deviation (σ) and mean (μ) of the lognormal distribution are calculated using these values. The standard deviation (σ) is determined as: 𝜎 = √𝑙𝑛( 1 + (𝑘𝐵𝑆𝑚𝑒𝑎𝑛 − 𝑘𝐵𝑆𝑚𝑖𝑛 𝑘𝐵𝑆𝑚𝑖𝑛 )2 ) Eq. SM 1 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 73 𝜇 = 𝑙𝑛(𝑘𝐵𝑆𝑚𝑒𝑎𝑛) − 0.5 × 𝜎2 Eq. SM 2 A loop is initiated to sample kBS values from the lognormal distribution until a valid value greater than or equal to the minimum is obtained. The sampling is performed using the Box-Muller transform to gen- erate a lognormally distributed random variable based on the calculated μ and σ. As the water balance module in ForClim did not undergo changes compared to version 4.0.1 (Huber et al., 2020), the mathematical formulation of E (transpiration), D (water demand), PET (potential evap- otranspitation), AET (actual evapotranspiration) and SM (soil moisture) are not reported here but can be found in the ForClim documentation (cf. Data Availability). D2 Parameter derivation: pedotransfer functions The determination of kBSmean and kBSmin is based on soil textural properties (sand, silt, clay), along with additional parameters such as upper and lower boundaries, thickness, and gravel content per genetic hori- zon. F rom these data, the available water capacity is determined by means of pedotransfer functions (Schmidt-Walter et al., 2020). Soil textural properties were derived from the 25 m resolution product for Swiss soils provided by Baltensweiler et al. (2021) . Three pedotransfer functions were tested: ROSETTA (Zhang & Schaap, 2017), Wessolek (2009) and Puhlmann & von Wilpert (2012). In the ROSETTA model, van Genuchten (VG) model curves are calculated, and the available water capacity (AWC) is retrieved based on matric potentials. The Wessolek method is based on field capacity, wilting point,and available water content using a pre-defined van Genuchten function. The third method, similar to Wessole k, calculates field capacity, wilting point, and available water content using a pre- defined van Genuchten function, yet it requires bulk density and soil organic carbon data, which are rarely available and uncertain to determine, especially in forest soils. Ultimately, the Wessole k pedotransfer function was chosen as it was developed for German sites and has been tested for forest soils. It furthermore allowed us better to compare the performance of For- Clim and LWFBrook90, as it had also been adopted in Meusburger et al. (2022). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 74 The Wessolek function estimates soil water content (θ) based on pressure head (ψ) using the van Genuchten model. The function takes parameters such as alpha, npar, mpar, ths, thr, and calculates wet- ness and water content accordingly. Then, field capacity (FC) and wilting point (WP) are calculated, and the available water content is retrieved for the given pressure heads (63 hPa for FC and 15 ’848 hPa for WP) for each layer n as: 𝐴𝑊𝐶 𝑛 = 𝐹𝐶 𝑛 − 𝑊𝑃 𝑛 Eq. SM 3 The volumetric available water content is then converted to gravimetric units (cm) by considering soil layer thickness and gravel content of each layer as: 𝑘𝐵𝑆 𝑚𝑒𝑎𝑛 = 𝐴𝑊𝐶̅̅̅̅̅̅̅ = ∑( 𝐴𝑊𝐶 𝑛 × (1 − 𝑔𝑟𝑎𝑣𝑛 100 ) × 𝑡ℎ𝑛 ) 𝑁𝐿 1 Eq. SM 4 Where NL is the number of soil layer s. The resulting soil hydraulic properties for each 25 x 25 m cell were averaged to obtain the mean values from which the mean bucket size was determined, as shown in Table D2; for each site, the minimum value of kBS was used to constrain kBSmin as: 𝑘𝐵𝑆𝑚𝑖𝑛 = 𝑚𝑖𝑛𝑖=1 𝑁𝐿 𝐴𝑊𝐶 𝑛 Eq. SM 5 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 75 Table D2: Soil properties derived from the Wessolec pedotransfer function. For each site (Sid), the number of soil layers ( NL) and their depth (upper and lower limits, cm) are given. For each soil horizon, texture (text, derived from clay,sand,silt (%)), and gravel content (grav, in %) are retrieved from the soil map of Baltensweiler et al. (2021) and averaged for each 25 m cell, and then used to calculate soil hydraulic properties, namely: ths (saturation water content fraction), thr (residual water content frac- tion), alpha (alpha parameter of the van Genuchten water retention function, npar (n parameter of the van Genuchten water retention function), mpar (m parameter of the van Genuchten water retention function), ksat (saturated hyraulic conductivity parameter of Mualem hydraulic conductivity function (mm/d)) and tort (tortuosity parameter of the Mualem hydraulic conduc- tivity function). Sid NL upper lower text grav ths thr alpha npar mpar ksat tort GZ 1 0 -0.05 Lu 0.01 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 GZ 2 -0.05 -0.15 Tu3 0.09 0.46 0 5.5 1.08 0.08 1,237.65 0 GZ 3 -0.15 -0.3 Tu3 0.09 0.46 0 5.5 1.08 0.08 1,237.65 0 GZ 4 -0.3 -0.6 Lt2 0.14 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 GZ 5 -0.6 -1 Lu 0.12 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 VA 1 0 -0.05 Tu3 0.06 0.46 0 5.5 1.08 0.08 1,237.65 0 VA 2 -0.05 -0.15 Tu3 0.1 0.46 0 5.5 1.08 0.08 1,237.65 0 VA 3 -0.15 -0.3 Tu3 0.13 0.46 0 5.5 1.08 0.08 1,237.65 0 VA 4 -0.3 -0.6 Tu3 0.22 0.46 0 5.5 1.08 0.08 1,237.65 0 VA 5 -0.6 -1 Lt2 0.27 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 TB 1 0 -0.05 Lt2 0.03 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 TB 2 -0.05 -0.15 Lt3 0.05 0.45 0.16 4.95 1.17 0.15 443.4 -4.1 TB 3 -0.15 -0.3 Lt3 0.09 0.45 0.16 4.95 1.17 0.15 443.4 -4.1 TB 4 -0.3 -0.6 Lt3 0.13 0.45 0.16 4.95 1.17 0.15 443.4 -4.1 TB 5 -0.6 -1 Lt2 0.17 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 UH 1 0 -0.05 Tu3 0.08 0.46 0 5.5 1.08 0.08 1,237.65 0 UH 2 -0.05 -0.15 Lu 0.15 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 UH 3 -0.15 -0.3 Lt2 0.33 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 UH 4 -0.3 -0.6 Lt2 0.42 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 UH 5 -0.6 -1 Lt2 0.4 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 BB 1 0 -0.05 Lt2 0.07 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 BB 2 -0.05 -0.15 Lt2 0.17 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 BB 3 -0.15 -0.3 Lt2 0.28 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 BB 4 -0.3 -0.6 Lt2 0.38 0.44 0.15 7.01 1.25 0.2 625.31 -3.18 BB 5 -0.6 -1 Ls3 0.4 0.41 0.07 6.83 1.21 0.17 982 -3.23 RB 1 0 -0.05 Lu 0.01 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 RB 2 -0.05 -0.15 Lu 0.06 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 RB 3 -0.15 -0.3 Lu 0.07 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 RB 4 -0.3 -0.6 Lu 0.13 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 RB 5 -0.6 -1 Lu 0.17 0.43 0.05 4.32 1.17 0.14 826.8 -3.23 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 76 D3. Modeling of Predisposing and Inciting factors (after Manion, 1981) D3.1. Proxy for the memory of carbon pools Trees are capable of resisting and recovering from a drought event (Körner, 2019). To mimic this behavior in the absence of a full-tree carbon balance in ForClim, we followed the protocol developed by Zamora- Pereira et al. (2021) to constrain the diameter increment in subsequent years by means of a proxy. The constraints are based on yearly ratios of measured tree -ring width data for Abies alba, Picea abies and Fagus sylvatica (Bottero et al. 2019 and Cailleret et al. 2017). In th is protocol, the smallest and largest ratios of inter-annual changes of tree-ring width are identified , respectively (Table D3.1, bold values). For the sake of consistency and parsimony, we selected the same two approximate values for all species, namely 0.2 and 5.0. In the current implementation, the yearly diameter increment of each cohort is corrected based on the values on the previous year’s growth. When the ratio of the current year is lower or higher than the em- pirical values, growth will be constrained accordingly. No lag effects longer than one year are considered. Table D3.1: Ratios of yearly tree-ring widths according to Maxime Cailleret and Zamora-Pereira et al. (2021). Species Quantile Ratio* Ratio** Ratio*** Silver fir 0.001 0.1420 0.2456 0.2086 0.999 4.3799 3.5517 3.7822 Norway spruce 0.001 0.3266 0.1632 0.1737 0.999 2.8598 5.7124 5.3877 Beech 0.001 0.1124 0.1426 0.1401 0.999 16.1329 6.0804 6.9725 * Bottero et al. 2019, ** Cailleret et al. 2017, *** both databases (used for constraints) D3.2. Seasonal soil moisture and REW critical in spring and autumn The empirical threshold of “critical Relative Extractable Water content” (REW) suggested by Granier et al., 1999 has been used as reference for more than 20 years with a value ~0.4. However, several studies have reported that REW varies strongly and seems to depend not only on soil properties but also on species-specific responses to drought via stomatal conductance, canopy conductance or, more generally, by transpiration strategies (Lindroth et al., 2018; Niu et al., 2023; Ruehr et al., 2012; Vilhar, 2016) . In .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 77 Table D3.2.2.1 we summarize the range of variability of REW c and report the temporal extent, species, biomes, and countries in which these values were assessed. These values were crucial to identify a range of plausible variability in the parameter space for the spring and fall period. Based on the literature review, we decided to set the range of variability for spring to [1.0 to 0.9] as in Chuste et al. (2019; cf. their Figure 2-A for control plots C in DoY 100-150), while for the fall period we identified the range [0.7 to 0.5] as in Vilhar et al. (2016). To assess the importance of choosing an exact value as well as to test the plausi- bility of these parameter ranges, we assessed the impact of the variation of REW across the beech sites in spring and fall, as shown in Figure D3.2. Table D3.2: Critical REW (REWc) values according to a literature review. The temporal extent, main investigated species, biome and country are reported. REWc Temporal extent Species Biome Country Reference 0.7-0.4 growing season (May-October) Fagus syl- vatica Temperate Slovenia Vilhar U. Comparison of drought stress indices in beech forests: a modelling study. (2016) iForest 9: 635-642. doi: 10.3832/ifor1630-008 0.4-0.5 growing season (2014-2015) Fagus syl- vatica Temperate Lorraine re- gion (France) Chuste, P. A., Massonnet, C., Gérant, D., Zeller, B., Levillain, J., Hossann, C., An- geli, N., Wortemann, R., Bréda, N., & Maillard, P. (2019). Short-term nitrogen dy- namics are impacted by defoliation and drought in Fagus sylvatica L. Branches. Tree Physiology, 39(5), 792–804. https://doi.org/10.1093/treephys/tpz002 < 0.1 15 days during growing season Quercus spp, Temperate NW China Niu, Xiaodong, and Shirong Liu. Environmental and stomatal control on evapotran- spiration in a natural oak forest." Ecohydrology 15.4 (2022): e2423 0.4-0.5 Late summer (August-Septem- ber) Ponderosa pine Temperate Central Ore- gon (USA) Ruehr, N. K., Martin, J. G., & Law, B. E. (2012). Effects of water availability on carbon and water exchange in a young ponderosa pine forest: Above - and below- ground responses. Agricultural and Forest Meteorology, 164, 136–148. https://doi.org/https://doi.org/10.1016/j.agrformet.2012.05.015 < 0.4 annual- growing period Quercus al- iena Warm Temperate Central China Niu, X., Chen, Z., Pang, Y., Liu, X., & Liu, S. (2023). Soil moisture shapes the en- vironmental control mechanism on canopy conductance in a natural oak forest. Sci- ence of the Total Environment, 857. https://doi.org/10.1016/j.sci- totenv.2022.159363 < 0.2 or < 0.05-0.5 (global sensitiv- ity analysis) Pinus syl- vestris Boreal Hyytiälä (Fin- land) Launiainen, S., Guan, M., Salmivaara, A., and Kieloaho, A.-J.: Modeling boreal forest evapotranspiration and water balance at stand and catchment scales: a spatial approach, Hydrol. Earth Syst. Sci., 23, 3457–3480, https://doi.org/10.5194/hess-23- 3457-2019, 2019 0.04 and 0.37 (Pi- nus), 0.18 and 0.39 (P. abies) Late summer (June to Septem- ber) Pinus syl- vestris, Picea abies Boreal Sweden Lagergren, F., & Lindroth, A. (2002). Transpiration response to soil moisture in pine and spruce trees in Sweden. In Agricultural and Forest Meteorology (Vol. 112). < 0.4 Annual Fagus syl- vatica, Picea abies,Pinus sylvestris Temper- ate/ Boreal Hesse (F), Hainich (D), Sorø (DK), Tharandt (D), Hyytiälä (FI), Loobos (NL) Bréda, N., Huc, R., Granier, A., & Dreyer, E. (2006). Temperate forest trees and stands under severe drought: a review of ecophysiological responses, adaptation processes and long-term consequences. Annals of Forest Science, 63(6), 625 -644. 0.33-0.45 Growing period Pinus tae- da, Lirio- dendron tu- lipifera, Liquidam- bar styraci- flua, Carya spp., Quer- cus spp. Temperate North Caro- lina (USA) Oishi, A. C., Palmroth, S., Butnor, J. R., Johnsen, K. H., & Oren, R. (2013). Spatial and temporal variability of soil CO2 efflux in three proximate temperate forest eco- systems. Agricultural and Forest Meteorology, 171–172, 256–269. https://doi.org/10.1016/j.agrformet.2012.12.007 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 78 Figure D3.2: Relative extractable water content variation across the six beech-dominated sites in the fall (SON) and spring (MAM) months. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 79 D3.3. Sensitivity analysis for new model parameters Persistent uncertainty regarding parameter selection is a recurring challenge in ecological modeling (Hu- ber et al., 2020) . We confronted this when introducing four new parameters (i.e., kREWfall, kREWspring, kDrTh, kEg) that required rigorous constraints. From the literature review, we identified the parameter range for kREW as presented in section D.3.2, hence the selected ranges for fall [0.7-0.5] and spring [1.0-0.9] were chosen as upper and lower boundary conditions. For the parameter kDrTh, we identified a threshold range spanning from 0.1 (lower boundary) to 0.3 (upper boundary) corresponding to a very low to very high drought sensitivity class associated with the drought index (gDr, Eq.9 in the main text), which usually varies between 0 (no drought) to 0.2 (me- dium-strong drought) during the growing season (Figure D3.3.1). For kEg, we assumed a variation be- tween 1 (i.e., any deviation from the optimum E/D would count) or a slight threshold such as 0.9. This condition implies that the ratio of transpiration to demand should be sufficiently high during the growing season. The scarcity of empirical data posed a challenge in this regard. To confine this uncertainty and to choose a plausible combination of parameter values within determined ranges, we adopted a two -fold approach aimed at exploring the parameter space effectively (cf. Table D3.3.1). D3.3.1 Short-term patterns We focused on identifying a target pattern which operates in the short term, specifically the basal area losses triggered by the droughts of 2018-2020 at the six beech-dominated sites. The analyses of the short- term pattern allowed us to confirm the pre-conceived notion about the optimal parameter ranges. Yet, to ensure the model's generalizability across diverse environmental conditions and future case studies, we extended our analysis to include sites of the European gradient, particularly those characterized by high aridity. Therefore, we selected a long-term pattern focusing on basal area development over time and the frequency of mortality events reducing it. This strategy, inspired by the pattern-oriented modeling frame- work (Grimm et al., 2005) , facilitated a comprehensive exploration of parameter values that accounted for both short -term disturbances and long -term ecological dynamics, enhancing the overall robustness and applicability of our model. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 80 We developed a scoring system in the range [0,1] to eliminate the scenarios which were failing to reproduce the pattern in the short term. A parameter combination leading to basal area losses in the exact years of observation (2018-2020) would score 1. Furthermore, the scores varied based on how widespread the mortality events were across multiple sites: i) if mortality events occurred in more than two out of six sites in the exact year of observation the score was 0.75; ii) if mortality events occurred in two out of six sites the score drop ped to 0.5; iii) if mortality events occurred in one out of six sites the score further dropped to 0.25. Only the scenarios with a score ≥ 0.75 were selected for testing on long-term simulations across the European gradient of sites (cf. main text). The scenarios that reproduced a scoring greater than 0 are shown in Table D3.3.2. Table D3.3.1: Pattern-oriented modelling framework to assess parameter uncertainty (after Huber, 2019) Pattern Point in time Level Short-term Basal area losses 2018-20 (drought-induced mortality event) Total stand Long-term Frequency in basal area losses Dynamic equilibrium (years -1500) Total stand .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 81 Table D3.3.2: Scenario definition as combination of the parameters REW in autumn and spring and the two drought tolerance thresholds kDrTh (Eq. 8) and g (Eq. 9) in the main text. Highlighted in red are the parameter combinations with a score of 0.75. In the red shaded rows, we show the best parameter set, which was chosen in this study. Scenario kREWfl kREWsp kDrTh kEg 1 0.5 1 0.1 1 2 0.6 1 0.1 1 3 0.7 1 0.1 1 4 0.5 0.9 0.1 1 5 0.6 0.9 0.1 1 6 0.7 0.9 0.1 1 7 0.5 1 0.2 1 8 0.6 1 0.2 1 9 0.7 1 0.2 1 10 0.5 0.9 0.2 1 11 0.6 0.9 0.2 1 12 0.7 0.9 0.2 1 13 0.5 1 0.3 1 14 0.6 1 0.3 1 15 0.7 1 0.3 1 16 0.5 0.9 0.3 1 17 0.6 0.9 0.3 1 18 0.7 0.9 0.3 1 19 0.5 1 0.1 0.9 20 0.6 1 0.1 0.9 21 0.7 1 0.1 0.9 22 0.5 0.9 0.1 0.9 23 0.6 0.9 0.1 0.9 24 0.7 0.9 0.1 0.9 25 0.5 1 0.2 0.9 26 0.6 1 0.2 0.9 27 0.7 1 0.2 0.9 28 0.5 0.9 0.2 0.9 29 0.6 0.9 0.2 0.9 30 0.7 0.9 0.2 0.9 31 0.5 1 0.3 0.9 32 0.6 1 0.3 0.9 33 0.7 1 0.3 0.9 34 0.5 0.9 0.3 0.9 35 0.6 0.9 0.3 0.9 36 0.7 0.9 0.3 0.9 We selected a total of 36 scenarios as result of the combination of the 4 parameter values (Table D3.3.2), which were evaluated for the six beech sites. All scenarios led to different degrees of basal area losses (cf. Figure D3.3.2), most commonly including the years 2003-2004, 2008-2009 and the years 2016-2017, except for scenario 28 which was identified as best matching the observations (Figure D3.3.3) and those scenarios with a score greater than 0.5 which were then selected for further testing (red values in Table D3.3.2, cf. also Figure D3.3.2). .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 82 Figure D3.3.1: Drought sensitivity threshold classes from very low (0.1) in red to very high (0.3) in blue, in relation to the drought tolerance classes ( kDrTol) of the tree species parameterized in ForClim. Each class corresponds to different threshold values (kDrTh) per species group. As examples, 8 representative species of the 32 species parameterized in ForClim are labelled and shown in the plot. Figure D3.3.2: Basal area development in scenarios 1 and 30 resulting from the combination of the four parameters kREWfall, kREWspring, kEg and kDrTh (cf. Table D3.3.2). The black arrows indicate the mortality years considered as artifacts. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 83 Figure D3.3.3: Combination of the four parameters kREWfall, kREWspring, kEg (shown as g in the plot) and kDrTh according to the parameter range selected in Table D3.3.2. Each colored bar represents a unique parameter combination (scenario) with its score. The scenarios which scored a value> than 0.25 are shown in the plot. D3.3.2 Long-term patterns We assessed the long-term patterns by evaluating the mortality frequency on selected default sites, namely Davos, Adelboden and Potsdam , resulting in 33 simulation scenarios. Simulations started in the year 0 and ended in the year 1500 , using model variant 22 (Huber et al. 2020), while the simulation setup was the same as in the main experiment . To systematically assess the differences within each scenario (S) - site (N) combination, we developed a cost function (𝐶𝑠) that calculates the sum of squared differences between the simulated basal area (Eq. SM6) and its lagged value (time points, (𝑇), cf. Eq. SM7 ). To facilitate the choice of the optimum parameter set, the scenario with the lowest cost (𝑠∗, Eq. SM8.), indi- cating the most favorable outcome in terms of the evaluated variable, was selected. Basnt ∼ 𝒩(μs, σ2) Eq. SM 6 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 84 𝐶𝑠 = ∑ ∑(𝐵𝑎𝑠𝑛𝑡 − 𝐵𝑎𝑠𝑛(𝑡−1)) 2 ] 𝑇 𝑡=2 𝑁 𝑛=1 Eq. SM 7 𝑠∗ = arg min 𝑠∈{1,…,𝑆} 𝐶𝑠 Eq. SM 8 This calculation helped us quantify the deviation or change in the variable under consideration, aiding in the evaluation of scenario performance by assigning higher cost to scenarios that result in larger changes in basal area over time (Figure D3.3.4). While no differences were observed at the site Adelboden, sce- nario 6 yielded the lowest cost at Davos (cold part of the gradient ), while scenarios 6, 12, 24 and 30 resulted in low costs at Potsdam (dry part of the gradient). Therefore, we concluded that the parameter combination of scenario 6 can be used as a second-best alternative to the chosen parameter selection for scenario 28. Figure D3.3.4: Cost function (see Eq. 14) associated with each scenario for three sites along the European climatic gradient. D4: kDD re-estimation In the new formulation of the growth equation (cf. Eq. 6 of the main text) , we enhanced the growth response to climatic extremes by applying Liebig’s law of the minimum between temperature and soil water dynamics, rather than multiplying the two factors and keeping them within the root function. Thus, the growth reduction associated with temperature (DDGF) and moisture (SMGF) became substantially larger in ForClim v4.1. This has considerable impacts on forest dynamics, particularly in cold and dry .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 85 conditions. We therefore applied a formal equivalence between the two DDGF formulations, and defined the system: 𝐷𝐷𝐺𝐹𝑣4.0.1 = √(𝑚𝑎𝑥(0, 1 − 𝑒𝑎 ∗ (𝑘𝐷𝐷 𝑣4.0.1− 𝑢𝐷𝐷))) Eq. SM 9 𝐷𝐷𝐺𝐹𝑣4.1 = max(0,1 − 𝑒𝑎 ∗ (𝑘𝐷𝐷 𝑣4.1− 𝑢𝐷𝐷)) Eq. SM 10 where kDD stands for the species-specific minimal annual degree-day sum. uDD is the monthly sum of degree-days (°C·days), and a is the slope parameter. Values for kDD v4.1 were approximated iteratively: the Newton-Raphson iteration algorithm starts with an initial guess by setting kDD v4.1 to the value of kDD v4.0.1 and the function iteratively updates the guess based on the differences between DDGF v4.0.1 and DDGFv4.1, converging towards a value of kDD v4.1 that satisfies a specified tolerance value set to 0.1. The kDD and uDD values used for testing were retrieved from simulations along the European site gra- dient. We fit ted a linear model to predict the approximate values of kDD v4.1 based on kDD v4.0.1. The highest performance was obtained with the following model (Table D4.1): 𝑘𝐷𝐷𝑣4.1 = −18.61 + 0.72 · 𝑘𝐷𝐷𝑣4.0.1 Eq. SM 11 Table D4.1: Summary statistics of linear model for kDDv4.1 estimation Residuals Min 1Q Median 3Q Max -281.02 -107.28 -19.36 116.68 256.79 Estimate (Intercept) -18.60847 62.55123 -0.297 0.768 kDD 0.71979 0.07938 9.067 1.54e-11 Multiple R2 Adjusted R2 F-statistic Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 0.6566 0.6486 82.22 The newly estimated parameter values can be found in Table D4.2. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 86 Table D4.2: Current (v4.0.1) and newly estimated (v4.1) kDD parameter values for the European tree species in ForClim. kName kDD v4.0.1 kDD v4.1 Alnus viridis 350 233 Larix decidua 350 233 Pinus cembra 350 233 Picea abies 350 233 Betula pendula 425 287 Populus tremula 425 287 Pinus montana 500 341 Sorbus aucuparia 500 341 Pinus sylvestris 500 341 Alnus incana 500 341 Pseudotsuga menziesii 633 437 Abies alba 650 449 Acer pseudoplatanus 650 449 Salix alba 650 449 Populus nigra 750 521 Alnus glutinosa 750 521 Sorbus aria 750 521 Tilia cordata 750 521 Fagus sylvatica 850 593 Corylus avellana 850 593 Fraxinus excelsior 850 593 Acer platanoides 850 593 Ulmus glabra 850 593 Tilia platyphyllos 850 593 Quercus petraea 950 665 Carpinus betulus 950 665 Acer campestre 950 665 Taxus baccata 1050 737 Quercus robur 1050 737 Castanea sativa 1050 737 Quercus pubescens 1150 809 D5: kRedMax and kHmin D5.1. Dynamic calculation of maximum tree height and climatological averaging of drought and de- gree-days When utilizing weather time series in the ForClim simulations, it is logical to dynamically calculate the maximum height that trees (gHMax) can reach, rather than to determine it prior to the simulation based on the assumption of a constant “current” climate, as done in previous versions of ForClim (cf. Rasche et al. 2012). This change in the approach was necessary because weather time series are inherently non - .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 87 stationary. By recalculating gHMax annually based on a moving window approach , we can reflect the influence of changing climatic conditions on tree growth (“site index”), ensuring that the model is re- sponsive to temporal fluctuations in temperature, precipitation, and other critical environmental factors. This dynamic adjustment is crucial for capturing realistic growth patterns and ecological responses of trees in a varying climate, including drought-induced mortality. To this end, averages of the drought index and the degree-day sum are calculated over a 30 -year period using a moving average, thereby treating these averages as a climatology underlying the calculat- ing of “site index”. D5.2. Minimum height at the species distribution limit Rasche et al. (2012) used the 𝑘𝑅𝑒𝑑𝑀𝑎𝑥 parameter to quantify the maximum reduction in tree height based on suboptimal temperature and drought conditions, using data from the Swiss National Forest Inventory (NFI) and Growth-and-Yield plots. This parameter was designed to reflect the maximum height reduction due to environmental stressors, particularly drought and degree-days, but its estimation is not suitable for simulations outside productive forests, thus potentially misrepresenting ecological dynamics at the distri- butional limits of the species (Körner et al. 1998). To address this limitation, we assumed that the param- eter 𝑘𝐻min should reflect the minimum height that trees (which survive for a long time) can achieve at their upper and lower distributional limits, in the most extreme case at treeline. Yet, the definition of a tree height threshold to define forests varies from 5-3 m (Jeník & Lokvenc 1962, Körner 2012) in No rthern Central Europe to 2 m (Holtmeier et al. 2009) in Central Southern Europe (Cudlín et al. 2017). Moreover, recent studies (Gelabert et al. 2024) indicate that up to around 1 500 m a.s.l., trees reach a fairly constant maximum height in the European Alps, decreasing at a rate of -1.27 m per 100 m of elevation. Despite the limitations of the study, we approximated the minimum height of trees to be 15 m by taking the mean between the estimated maximum tree height at the highest site of the European gradient (2300 m a.s.l., site of GrandeDixence), which amounts to 27 m and the maximum tree height reported by Körner for the Swiss Alps of 3 m . We therefore set 𝑘𝐻min to 15 m and recalculated 𝑘𝑅𝑒𝑑𝑀𝑎𝑥 according to Eq. 19. Furthermore, the model updates the maximum tree height (𝑔𝐻𝑀𝑎𝑥) annu- ally to incorporate the most recent trends and environmental conditions (Eq. SM 19-24). This adjustment .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 88 ensures that calculated 𝑔𝐻𝑀𝑎𝑥 reflects the latest data trends and environmental conditions, thus enhanc- ing the reliability of our simulations. 𝑘𝑅𝑒𝑑𝑀𝑎𝑥 = 1 − 𝑘𝐻min 𝑘𝐻Max Eq. SM 19 𝑔𝑅𝑒𝑑(𝑚𝐷𝑟Se) = min(max (𝑚𝐷𝑟Se 𝑘𝐷𝑟Tol ⋅ 𝑘𝑅𝑒𝑑Max, 0), 𝑘𝑅𝑒𝑑Max) Eq. SM 20 𝑔𝑅𝑒𝑑(𝑚𝐷𝐷Se) = min(max ((1 − 𝑚𝐷𝐷𝑆𝑒 − 𝑘𝐷𝐷Min 𝑘𝐷𝐷Opt − 𝑘𝐷𝐷Min ) ⋅ 𝑘𝑅𝑒𝑑Max, 0) , 𝑘𝑅𝑒𝑑Max) Eq. SM 21 𝑔𝐻𝑀𝑎𝑥(Dr) = (1 − 𝑔𝑅𝑒𝑑(𝑚𝐷𝑟𝑆𝑒))⋅ 𝑘𝐻max Eq. SM 22 𝑔𝐻𝑀𝑎𝑥(DD) = (1 − 𝑔𝑅𝑒𝑑(𝑚𝐷𝐷𝑆𝑒))⋅ 𝑘𝐻max Eq. SM 23 𝑔𝐻𝑀𝑎𝑥 = min(𝑔𝐻𝑀𝑎𝑥(Dr), 𝑔𝐻𝑀𝑎𝑥(DD)) Eq. SM 24 D6: Site parameters for the six beech sites as well as the LWF sites Lägeren and Visp Table D6: Summary table of site characteristics at the six drought-prone beech sites and at the ICP-Level II plots. Site kPatchSize (m2) kBSmean (cm) kBSmin (cm) kAvN (kg ha-1 y-1) SN1 (p ha-1) Blattenberg 550 8.6 8.2 100 180 Grosszinggibrunn 550 15.4 14.4 180 245 Rossberg 550 15.5 13.3 180 187 Tüeliboden 550 14 12.6 180 170 Usserholz 550 11.2 10.2 180 165 Vogtacher 550 13.7 12.6 180 201 Visp 750 7.0 9.0 100 232 (i) Lägeren 550 8.5 7.3 180 503(ii) 1 Simulated Stem Number. Tree number with a DBH > 8 cm (i) Observed stem number per hectare 226 p ha -1 in 1996 (Dobberdin et al 2005) (ii) Observed stem number per hectare was 503 in 2011 (SwissFluxnet, 2024) .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 89 D7: Site parameters along the climatic gradient Table D7.1: Summary table of average temperature (T, °C) and precipitation sum (P, cm) for the 12 European sites. Weather data were obtained from Bugmann & Solomon (2000). Site T P Grande-Dixence 1.3 1017 Bever (S) 1.5 841 Davos 3 1007 Adelboden 5.5 1351 Huttwil 8.1 1287 Bern 8.4 1006 Schaffhausen 8.6 882 Basel 9.2 784 Schwerin 8.2 625 Potsdam 8.7 589 Cottbus 8.77 573 Sion 9.7 597 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 90 Table D7.2: kBS values across the European gradient (a: Swiss sites, b: German sites). Buffer from kBSmin,M from Meusbuerger et al, kBSmin,41 as the value chosen for ForClim v4.1 and kBS for ForClim v4.0.1 as in Bugmann and Solomon (2000). a) Swiss sites site kBSmin,M kBSmean,M kBSmin,41 kBSmean,41 kBS Δ[kBSmean,M - kBSmean,41] Δ[kBS - kBSmean,M] Δ[kBS - kBSmean,41] Grande Dixence 9.5 10.0 7.0 8.0 10.0 2.0 0.0 2.0 BeverS 8.3 9.4 8.1 9.0 10.0 0.4 0.6 1.0 Davos 8.2 9.3 5.0 7.0 10.0 2.3 0.7 3.0 Adelboden 9.1 10.6 9.0 10.0 15.0 0.6 4.4 5.0 Huttwil 13.5 15.2 12.0 13.0 20.0 2.2 4.8 7.0 Bern 13.5 14.9 9.0 14.0 20.0 0.9 5.1 6.0 Schaffhausen 9.5 11.6 9.0 13.0 15.0 -0.5 3.4 2.0 Basel 12.4 13.9 11.0 14.0 15.0 -0.1 1.1 1.0 Sion 10.0 11.4 9.0 10.0 15.0 1.4 3.6 5.0 b) German sites site kBSmin,BGB kBSmean,BGB kBSmin,41 kBSmean,41 kBS Δ[kBSmean,BGB - kBSmean,41] Δ[kBS - kBSmean,BGB] Δ[kBS - kBSmean,41] Schwerin 5.0 18.5 4.0 19.0 24.0 -0.5 5.5 5.0 Potsdam 5.0 23.2 5.0 25.0 24.0 -1.8 0.8 -1.0 Cottbus 4.6 9.6 4.0 10.0 24.0 -0.4 14.4 14.0 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 91 Table D7.3: ForClim v4.1 is based on ForClim v4.0.1 (Huber et al., 2020), which offered a total of 504 model variants. The table shows the setup of ForClim v4.1 and salient differences to ForClim v4.0.1 beyond those explained in the main manuscript. Acronyms: A stands for the species-specific light-dependent allocation into height vs diameter growth based on the scaled shade tolerance; E represents the establishment probability routine according to which the establishment probability (kEstP) is set to 4%. The mortality M is distinguished in background mortality (𝑀0) and stress-induced mortality. The latter is further differentiated in growth dependent (𝑀SG) and the predisposing-inciting stress dependent mortality (𝑀S-PI). ForClim version Model variant Weather approach Soil bucket Height reduction Establishment routine Allocation of growth

mortality Stress-induced mortality 4.0.1. 22 Generated One kRedMax 𝐸6* 𝐴2 𝑀0 𝑀S-G 4.1. 22 Time series Distributed Minimum height 𝐸6* 𝐴2 𝑀0 𝑀S-PI .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 92 Figure D7.1: Available Water Capacity (mm) maps by Meusburger et al. (2022). Symbols indicate the nine Swiss sites of the European gradient. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 93 Figure D7.2: Available Water Capacity (mm) maps produced by BGB. Symbols indicate the three German sites of the European gradient. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 94 Figure D7.3: Basal area differences (m2 ha-1) between ForClim v4.0.1 variant 22 minus ForClim v4.1 variant 22 across the 12 default sites. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 95

in the Supplementary Materials Bréda, N., Huc, R., Granier, A., & Dreyer, E. (2006). Temperate forest trees and stands under severe drought: a review of ecophysiological responses, adaptation processes and long-term consequences. An- nals of Forest Science, 63(6), 625-644. Bottero, A., Cailleret, M., van der Maaten, E., Cochard, H., & del Castillo, E. (2021). Growth resistance and resilience of mixed silver fir and Norway spruce forests in central Europe: Contrasting responses to mild and severe droughts. Annals of Forest Science, 78, 1-28. https://doi.org/10.1007/s13595-021- 01052-3 Cailleret, M., Nourtier, M., Amm, A., Durand-Gillmann, M., & Davi, H. (2017). Drought-induced de- cline and mortality of silver fir differ among three sites in Southern France. Annals of Forest Science, 74, 65. https://doi.org/10.1007/s13595-017-0665-0 Chuste, P. A., Massonnet, C., Gérant, D., Zeller, B., Levillain, J., Hossann, C., Angeli, N., Wortemann, R., Bréda, N., & Maillard, P. (2019). Short-term nitrogen dynamics are impacted by defoliation and drought in Fagus sylvatica L. Branches. Tree Physiology, 39(5), 792–804. https://doi.org/10.1093/treephys/tpz002 Enhancing gap model accuracy by modeling dynamic height growth and dynamic maximum tree height. (2012). Ecological Modelling, 240, 9-18. https://doi.org/10.1016/j.ecolmodel.2012.04.005 Lagergren, F., & Lindroth, A. (2002). Transpiration response to soil moisture in pine and spruce trees in Sweden. Agricultural and Forest Meteorology, 112(2), 67-85. https://doi.org/10.1016/S0168- 1923(02)00060-6 Launiainen, S., Guan, M., Salmivaara, A., & Kieloaho, A.-J. (2019). Modeling boreal forest evapotran- spiration and water balance at stand and catchment scales: a spatial approach. Hydrology and Earth Sys- tem Sciences, 23, 3457–3480. https://doi.org/10.5194/hess-23-3457-2019 Niu, X., Chen, Z., Pang, Y., Liu, X., & Liu, S. (2023). Soil moisture shapes the environmental control mechanism on canopy conductance in a natural oak forest. Science of the Total Environment, 857. https://doi.org/10.1016/j.scitotenv.2022.159363 Niu, Xiaodong, and Shirong Liu. (2022). Environmental and stomatal control on evapotranspiration in a natural oak forest. Ecohydrology, 15(4), e2423. https://doi.org/10.1002/eco.2423 Oishi, A. C., Palmroth, S., Butnor, J. R., Johnsen, K. H., & Oren, R. (2013). Spatial and temporal varia- bility of soil CO2 efflux in three proximate temperate forest ecosystems. Agricultural and Forest Mete- orology, 171–172, 256–269. https://doi.org/10.1016/j.agrformet.2012.12.007 Pastorello, G., Trotta, C., Canfora, E. et al. (2020). The FLUXNET2015 dataset and the ONEFlux pro- cessing pipeline for eddy covariance data. Scientific Data, 7, 225. https://doi.org/10.1038/s41597-020- 0534-3 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint 96 Puhlmann, H., & von Wilpert, K. (2012). Pedotransfer functions for water retention and unsaturated hy- draulic conductivity of forest soils. Journal of Plant Nutrition and Soil Science, 175(2), 221–235. https://doi.org/10.1002/jpln.201100139 Ruehr, N. K., Martin, J. G., & Law, B. E. (2012). Effects of water availability on carbon and water ex- change in a young ponderosa pine forest: Above- and belowground responses. Agricultural and Forest Meteorology, 164, 136–148. https://doi.org/10.1016/j.agrformet.2012.05.015 Vilhar, U. (2016). Comparison of drought stress indices in beech forests: a modelling study. iForest, 9, 635-642. https://doi.org/10.3832/ifor1630-008 Zamora-Pereira, J. C., Yousefpour, R., Cailleret, M., Bugmann, H., & Hanewinkel, M. (2021). Magni- tude and timing of density reduction are key for the resilience to severe drought in conifer-broadleaf mixed forests in Central Europe. Annals of Forest Science, 78, 1-28. https://doi.org/10.1007/s13595- 021-01052-3 Zhang, Y., & Schaap, M. G. (2017). Weighted recalibration of the Rosetta pedotransfer model with im- proved estimates of hydraulic parameter distributions and summary statistics (Rosetta3). Journal of Hy- drology, 547, 39–53. https://doi.org/10.1016/j.jhydrol.2017.01.004 SwissFluxnet site Lägeren. accessed in February 2024. URL: https://www.swissfluxnet.ethz.ch/in- dex.php/sites/site-info-ch-lae/. .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted December 16, 2025. ; https://doi.org/10.64898/2025.12.15.694299doi: bioRxiv preprint