Comparison of linear mixed effects model and generalized model of the tree height-diameter relationship

Abstract

Models of height curves generated using a linear mixed effects model and generalized model were compared. Both tested models were also compared with local models of height curves, which were fitted using a nonlinear regression. In the mixed model two versions of calibration were tested. The first calibration approach was based on measurement of heights only in trees of the mean diameter interval, while the second calibration approach was based on measurement of tree heights in three diameter intervals. Generalized model is the mathematical formulation of a system of uniform height curves, which is commonly used in the Czech Republic. The study took place at Training Forest Enterprise called Masaryk Forest at Křtiny and was carried out for Norway spruce (Picea abies [L.] Karst.). It was found that the mixed model behaves correctly only in the case of calibration based on selection of trees in three diameter intervals. Selection of a total of nine trees was confirmed as the most suitable to calibrate the model. In most of the calculated quality criteria, the mixed model achieved better results than the generalized model, even with a smaller number of measured heights. The bias of both models from the local model was very similar (0.54 m for the mixed model and 0.44 m for the generalized model). The mixed model can therefore fully replace the commonly used generalized model even with a smaller number of measured heights.

New Insights into Tree Height Distribution Based on Mixed Effects Univariate Diffusion Processes

“…Mathematical technique of a system of uniform diameter and height regional functions is the approach known as the generalized model. The mixed effects regression models are able to achieve the same results than the generalized model [10, 33]. In this study new developed mixed effects parameters height-diameter relationships demonstrated similar statistical indexes as in the nonlinear generalized height-diameter regression models presented by Petrauskas et al [34].…”

A new paradigm in modelling the evolution of a stand via the distribution of tree sizes

“…The mathematical technique for a system of uniform diameter and height regional functions is the approach known as the generalised model. The mixed effects regression models are able to achieve the same results as the generalised model 34 .…”

Reconstructing past forest status using inventory and tree ring data to support uneven-aged forest management

“…Their data showed that the best model performance was obtained for fir with the coefficient of determination 0.94 -0.95 and RMSE 1.94 -1.97 m. Better results than those our are conditioned by long term data series and more or less even-aged stand structure. Recent studies suggest using general mixed effect height-diameter models rather than general ordinary least square models because of their higher prediction accuracy (Adamec 2015;Mehtätalo et al 2015;Sharma et al 2016). This could also improve modelling of heightdiameter relation in markedly uneven-aged forests.…”