Assessing a reliable modeling approach of features of trees through neural network models for sustainable forests

Financial exploitation of forests comprises an important part of man activity. There are efforts being made to conserve the sustainable exploitation while simultaneously avoiding degradation of the environment. One tool used in these efforts is the modeling of tree features, such as total tree height, sawn-timber tree height, merchantable tree height, and total or sawn-timber tree volume, which yields an estimate of the forest in finance recoverable goods. Sustainable forest management design must be supported by the adjustment of computational techniques. The purpose of this paper is to assess a reliable modeling approach for estimating individual tree heights for the maturity of trees for logging through determining the applicability of different types of neural network models and identifying a neural network procedure for accurate estimation of these variables. These models serve as an alternative to the traditional regression approach. All types of model estimations are evaluated and compared in this paper. Back Propagation Artificial Neural Network (BPANN), Cascade Correlation Artificial Neural Network (CCANN), and Generalized Regression Neural Network (GRNN) models are developed to estimate individual tree heights for the logging of mature trees, such as sawn-timber height and merchantable height. The results reported in this research suggest that the selected BPANN and CCANN models are reliable and demonstrate their adequacy and potential for estimating sawn-timber and merchantable tree height. The results also illustrate that the CCANN models are superior to the BPANN and GRNN models and lead to higher estimation accuracy. Moreover, the NN models were found to be superior to the tested nonlinear regression models.