Multivariate mixed-effects models for stand characteristics of hybrid aspen plantations in southern Finland and southern Sweden

Abstract

Hybrid aspen, a hybrid between the European aspen and North American trembling aspen (Populus tremula L. × P. tremuloides Michx.), is a promising species because of its fast-growth and its suitability for multi-purpose use. However, models for predicting the age-dependent development of stand characteristics are still missing. The main objectives of this study were therefore to develop the models for predicting stand characteristics of hybrid aspen plantations and to validate the model applicability. The target response variables were stand basal area (BA), basal area-weighted mean diameter (DG) and basal area-weighted mean height (HG). Data were obtained from clonal hybrid aspen trials in southern Finland and southern Sweden. Multivariate mixed-effects modelling was used to estimate the parameters of seemingly unrelated regression for BA, DG, and HG. Model fit provided the following predictor variables: stand age (AGE), the number of trees per hectare (TPH), site index (SI), growing degree-days (GDD5), soil and site type, and thinning treatment. The chosen predictors differed slightly by response variable, but all parameters were highly significant (P < 0.0001), and model goodness-of-fit statistics presented high accuracy: RMSE of 2.59 m2 ha−1 for BA, 1.21 cm for DG, 1.05 m for HG in arithmetic scale. The applied simulations illustrated clear differences in the predicted development of stand characteristics when input variables SI, TPH or GDD5 changed. The developed models were assessed to be easily applicable and useful for predicting the stand and tree characteristics of clonal hybrid aspen plantations, especially for the stands with AGE ≤ 30 years and TPH ≤ 2000 trees ha−1.