3D crown information enhances growth predictions and where the gain comes from

Individual-tree growth models traditionally rely on basic tree and crown dimensions and competition indices, reducing complex three-dimensional (3D) crown architecture to scalar variables. Terrestrial laser scanning (TLS) enables detailed quantification of crown structure. TLS-derived crown metrics may better describe the environment of a tree and provide proxies for its competition for light and other resources. They may therefore be strong predictors of tree growth. We evaluated whether TLS-derived crown metrics improve individual-tree growth predictions, whether improvements are driven by crown size, crown shape, or both, and whether their importance differs among species with contrasting crown plasticity. We combined TLS-derived 3D crown metrics with tree-ring chronologies for 675 trees of four common temperate species ( Fagus sylvatica L., Picea abies (L.) H.Karst., Pinus sylvestris L., and Quercus spp.) across 26 plots from 18 long-term experiments in southern Germany. Mean basal-area increment (BAI) over the 10 years before the TLS scan was modelled using a nested hierarchy: M1 included basal area and Hegyi competition index; M2 (M1 + crown-size metrics); and M3 (M2 + crown-shape metrics). We fitted linear mixed-effects models with random intercepts for experiment and plot, fitted for all species pooled and for each species separately. For all species pooled, marginal R² increased from 0.59 (M1) to 0.63 with crown size and to 0.64 with crown shape; both steps were significant (p < 0.001). Species-specific marginal R² increased from 0.32 to 0.49 for Fagus , 0.60–0.63 for Picea , 0.43–0.57 for Pinus , and 0.70–0.76 for Quercus . Crown variables also reduced the variance between stands by 46%, and in Picea explained most of the differences between experiments. Crown size gave the full gain in Fagus and Quercus , whereas crown shape added information in Picea and Pinus . These findings demonstrate that integrating TLS-derived 3D crown structure can improve individual-tree growth models.