Evaluating the performance of mainstream Swedish growth models in uneven-aged forestry systems

Continuous Cover Forestry (CCF) practices are increasingly recognized for their potential in climate change adaptation and biodiversity conservation. Selection cutting, a key method within CCF, presents unique challenges for forest growth modelling due to its complex structure and distinct growth dynamics. Current models, largely developed from data obtained from even-aged stands, may exhibit lower accuracy when applied to uneven-aged stands. This study assessed the short-term (i.e., up to 15 years) predictive accuracy of the Swedish Heureka Decision Support System for stands managed with selection cutting. It assessed growth models for tree recruitment, growth, and mortality using data from 27 CCF field experiments covering a broad latitudinal and environmental range across Sweden. A linear mixed-effects modelling approach was used to analyse differences between observations and model predictions. Findings revealed potential species-specific biases, with an average underestimation of volume growth by 2 m3 ha-1 yr-1 after ten years of simulation, driven predominantly by underestimations in Norway spruce growth. While mortality predictions were generally accurate, they exhibited slight underestimation after recent cutting and overestimation otherwise. Ingrowth density predictions demonstrated minor biases, with spruce being underestimated and birch overestimated, but displayed high residual variability. Sensitivity analysis revealed correlations of residuals with stand variables, including site index, proportion of spruce, and stand basal area. The study faced limitations due to data scarcity and the short observation periods. Although most observed biases were not statistically significant, the findings underscore potential discrepancies when applying current Swedish models to selection cutting stands.