Climate-change-driven growth decline of European beech forests

Global environmental change is impacting ecosystems worldwide. Forests, key terrestrial ecosystems, are increasingly showing cascading effects from anthropogenic climate change, including significant consequences for water and carbon cycles and societal services. Large-scale analyses encompassing the full distribution range of key species are crucial to address these evolving impacts. While forest cover prediction research has a long history of exploring climate change's link to forest presence/abundance, less is known about ecologically-based predictions of species growth. Considering that tree stems comprise about 70% of a tree's biomass, secondary growth serves as a reasonable proxy for total carbon sequestration and an effective indicator of tree health and performance. Dendroecological analyses, traditionally localized, have offered valuable regional insights into growth responses to local habitat and climate change. However, spatio-temporal studies of actual and predicted growth at scales incorporating species' geographic and climatic distributions are rare. Existing tree-ring databanks, while valuable, often exhibit biases or limitations for certain taxa, biomes, and trailing-edge populations, hindering ecologically-focused applications. To overcome these challenges, a dense, species-specific network covering the entire ecological spectrum of *Fagus sylvatica* L. (beech) was created, including over 780,000 ring width measurements from 5800 trees across 324 European sites. Beech is ecologically and socio-economically significant in Europe, having played a dominant role in postglacial reforestation. However, its resilience to predicted climate change and ecological plasticity across its range remain poorly understood. This research analyzed past beech growth and projected growth variability under various climate change scenarios (CMIP6 Shared Socioeconomic Pathways) to understand 21st-century performance at continental scales. A generalized linear mixed-effects model (GLMM) was developed to model tree growth and facilitate spatio-temporal comparisons, identifying regions of growth increase or decrease in recent decades. The model, after validation, predicted radial growth until the late 21st century, considering various climate change scenarios.

Existing research demonstrates the significant impact of climate change on forest ecosystems (IPCC, 2014; Cailleret et al., 2017; Forzieri et al., 2021; Bonan, 2008). Studies focusing on the relationship between climate change and forest cover (Buras & Menzel, 2019; van der Maaten et al., 2017) are more prevalent than those examining the impacts on species-specific growth performance. While dendroecological analyses provide valuable insights at regional scales (Büntgen, 2019), large-scale spatio-temporal assessments incorporating species' full geographic and climatic distributions are limited (Klesse et al., 2020). Challenges in using tree-ring data include biases and limitations in international dendrochronological databanks (Zhao et al., 2019; Babst et al., 2018; Klesse et al., 2018). The importance of beech (*Fagus sylvatica*) in European ecosystems (Yousefpour et al., 2018) and its postglacial distribution (Giesecke et al., 2007) are well-established. However, understanding its resilience and plasticity under climate change is still needed, particularly given the potential risks to its range (Fang & Lechowicz, 2006).

A dense, species-specific tree-ring network was established, encompassing 324 sites, 5800 trees, and ~780,000 ring width measurements across Europe, covering the entire geographic and climatic range of *Fagus sylvatica*. Increment cores were processed using standard procedures (Speer, 2010), and ring widths were measured. Instead of applying detrending methods, the data was converted to annual basal area increment (BAI) using the `dplR` package to account for size dependency. Climate data (monthly precipitation and maximum/minimum temperatures) were obtained from CHELSA (Karger et al., 2017) for 1901-2016, with prevailing moisture conditions defined by the De Martonne aridity index (De Martonne, 1926; Martinez del Castillo et al., 2019). A generalized linear mixed-effects model (GLMM) was developed using the `lme4` package in R (Bolker et al., 2009; Calcagno & Mazancourt, 2010), incorporating variables for moisture conditions, elevation, latitude, and seasonally varying climate conditions (previous-year summer to current-year autumn). The model included 21 variables and 66 interactions, with previous-year BAI as a random effect to account for size-dependency. Model performance was evaluated using AIC scores, likelihood ratio tests and compared with reduced models and a null model (Table 1). The validated model predicted BAI for 1950-2016 (Supplementary Fig. 1). Growth rate differences between 1955-1985 and 1986-2016 were calculated for an average beech tree. Future growth under SSP1-2.6 and SSP5-8.5 CMIP6 climate scenarios (Meinshausen et al., 2020) was projected using multi-model ensemble means from CMIP6 for temperature and precipitation (Karger & Zimmermann, 2018), added to CHELSAcruts climate variables for three future periods (2020-2050, 2040-2070, 2060-2090). Applicability domains (AD) were calculated to assess model reliability under future conditions (Supplementary Fig. 2).

The full GLMM model accurately predicted 86% of growth variability. Geographical variables interacting with aridity index were significant predictors, showing differing latitude effects between xeric and mesic climates. Precipitation correlated positively with growth, while temperature effects varied seasonally. The most significant factor explaining growth variability across the distribution range was seasonal temperatures. Past growth analysis (1955-2016) revealed substantial spatio-temporal differences (Fig. 3). Growth rates were higher at low altitudes in NW and central Europe but lower at the southern range limits and higher altitudes (Fig. 2). Comparing 1955-1985 and 1986-2016 revealed a general growth decline across most areas, with greater decline at lower latitudes (Fig. 3). Future growth projections (Fig. 4) under SSP1-2.6 showed growth reductions up to 30% in southern Europe by 2020-2050, compared to 1986-2016, and increases at higher latitudes. SSP5-8.5 projected more dramatic decreases, exceeding 50% in southern Europe and positive trends in northern areas. These patterns intensified in later periods (2040-2070, 2060-2090).

The study reveals a pervasive growth decline in European beech forests since the 1980s, particularly in southern areas. The GLMM model successfully captured the complex interactions between geographic location, climate, and growth, highlighting the importance of considering variable interactions across environmental gradients. The findings reconcile discrepancies in previous studies, which often reported varying trends based on methodological differences. Results indicate a significant role of temperature increases in limiting beech growth, exceeding the effect of precipitation reduction. Increased temperatures exacerbate drought stress through higher atmospheric pressure deficits, constrained stomata closure, and increased tree water demand, potentially leading to hydraulic failure, defoliation, and reduced metabolic reserves. While beech is a competitive species, warming-induced growth declines pose significant challenges. The model projections suggest that future growth reductions will be substantial, particularly in southern Europe, potentially leading to increased forest mortality. This challenges previous predictions of increasing terrestrial carbon stocks, as the carbon sink capacity of beech forests will likely decrease.

This study, using an extensive tree-ring network, reveals a widespread growth decline in European beech forests since 1955, with significant future declines projected under climate change scenarios. The findings highlight the vulnerability of beech forests to warming temperatures and drought, emphasizing the need for forest adaptation strategies. Future research could focus on the combined effects of multiple stressors (extreme weather events, soil composition, competition), refining growth projections and developing more robust forest management practices.

The study's projections rely on climate models, which inherently have uncertainties, particularly regarding precipitation. Other factors potentially affecting growth, such as extreme weather events, soil conditions, and species competition, were not explicitly included in the model, potentially leading to some underestimation or overestimation of the effects. The model's predictive power is reduced outside the defined applicability domain.