Modelling total volume of dominant pine trees in reforestations via multivariate analysis and artificial neural network models

Abstract

The paper describes an artificial neural network (ANN) modelling approach for estimating the inside-bark and outside-bark total volume of dominant pine trees (Pinus brutia) in reforestations. Difficulties in modelling total volume are the great number of tree measurements that have to be used and the complex behaviour of the tree-bole growth and its dimensions. A basic novelty in this modelling approach is the reduction and selection of the minimal set of tree measurements by using multivariate analysis, introducing this set to ANN-based models in a way that they should be able to handle such reduced measurements. Furthermore, ANN model training was accomplished using the cascade-correlation algorithm in which the Kalman's learning algorithm is embedded. The selected ANN models provided unbiased predictions and systematically better accuracy compared with conventional regression models. The results demonstrate the superiority of the ANN models, due to their ability to overcome the problems in forest data, such as non-linear relationships, non-Gaussian distributions, multicollinearity, outliers and noise in the data. The ANN-based approach introduced in this study is sufficiently general and should be considered seriously for estimating the inside-bark and the outside-bark total tree volume.