Application of mixed-effects modelling approach in tree height prediction models

Abstract

In routine forest inventories, total tree height and diameter at breast height are very important growth parameters assessed to describe and estimate the stand structure and volume, respectively, of the forest. Height-diameter models are often used to predict the height for trees where only diameter is measured for all trees in a plot and a few trees measured for total height. This is because, tree diameter can be determined easily and accurately at little cost and time, but total tree height is more difficult to measure, time consuming and more costly. Africa has lagged behind in adopting modelling techniques that can assist in estimating tree height with higher precision and accuracy than that obtained using ordinary least squares and ordinary nonlinear least squares (which are the commonly used approaches). A study was carried out to demonstrate the utility of mixed-effects modelling approach in tree height prediction models. The ChapmanRichards model was selected as base height-diameter model and was fitted to model data using Ordinary Nonlinear Least Squares method. Using the same base model and fit data, a mixed-effects model was constructed using mixed-effects modeling approach. The two models were then compared in terms of predictive accuracy on independent data set (as well as model fit data for comparison). The mixed-effects model had a better predictive accuracy on both data sets, especially the independent data. Superiority of the mixed-effects model was more clearer when the two models were compared on a plot-by-plot basis. Forest modelers and managers in Africa should consider using mixedeffects modelling approach in development and use of height-diameter models in order to estimate tree heights with higher precision and accuracy.