Search for a command to run...
As climate change progressively alters temperature and precipitation regimes, understanding how these gradual shifts influence tree growth is essential for improving growth models and supporting adaptive forest management. Tree growth responses to climate are often nonlinear, temporally lagged, and mediated by stand structure and site conditions, making them difficult to capture with traditional approaches. This study investigates the effects of gradual climate change, competition, initial tree dimensions, age, and soil properties on basal area increment (BAI) using Generalized Additive Mixed Models (GAMMs). GAMMs allow flexible, nonlinear relationships while accounting for hierarchical data structures typical of long-term forest experiments. The analysis focuses on Norway spruce (Picea abies) in pure stands to avoid confounding effects of species mixtures. We used long-term data from experimental research plots of the Northwest German Forest Research Institute, which provide consistent measurements across multiple decades and capture slow-changing climatic trends relevant for growth modelling.The final GAMM for Norway spruce explained more than 60% of the observed variation in BAI, demonstrating a strong capacity to reproduce growth dynamics across a wide range of stand and site conditions and management regimes. To minimize multicollinearity and reduce the risk of overfitting, highly correlated predictors were grouped into broader site-factor categories (e.g. climate, soil, dendrometry and stand structure), from which a single representative variable was selected for model inclusion. Model development followed a structured, two-step selection procedure in R. An initial screening using mgcv::bam() was applied to efficiently evaluate candidate predictors based on Akaike Information Criterion, followed by refinement using mgcv::gam() to allow closer inspection of smooth terms and to ensure biologically meaningful response shapes. Random effects were used to account for plot-level variability and repeated measurements over time.Model performance was evaluated using an independent dataset, confirming the robustness and transferability of the fitted relationships. Sensitivity analyses further demonstrated that the model responds smoothly and consistently to gradual changes in climate variables, supporting its suitability for simulating long-term growth trends. Overall, the results highlight the value of long-term experimental research plots for disentangling climate signals from stand development processes and underline the advantages of GAMMs for integrating gradual climate effects into growth models. The developed BAI functions provide a solid foundation for incorporation into the TreeGrOSS forest growth simulator, enabling improved assessment of future growth trajectories of Norway spruce under changing climatic conditions. Keywords: Tree growth; Statistical modelling; Climate change; Basal area increment; Generalized additive mixed models; Long-term experimental research plots; Norway spruce (Picea abies)