Neural Networks for the Prediction of Species-Specific Plot Volumes Using Airborne Laser Scanning and Aerial Photographs

Abstract

Parametric and nonparametric modeling methods have been widely used for the estimation of forest attributes from airborne laser-scanning data and aerial photographs. However, the methods adopted suffered from complex remote-sensed data structures involving high dimensions, nonlinear relationships, different statistical distributions, and outliers. In this context, artificial neural networks (ANNs) are of interest as they have many clear benefits over conventional modeling methods and could then enhance the accuracy of current forest-inventory methods. This paper examines the ability of common ANN modeling techniques for the prediction of species-specific forest attributes, as exemplified here with the prediction stem volumes (cubic meters per hectare) at the field plot and forest stand levels. Three modeling methods were evaluated, namely, the multilayer perceptron (MLP), support vector regression (SVR), and self-organizing map, and intercompared with the corresponding nonparametric k most similar neighbor method using cross-validated statistical performance indexes. To decrease the number of model-input variables, a multiobjective input-selection method based on genetic algorithm is adopted. The numerical results obtained in the study suggest that ANNs are appropriate and accurate methods for the assessment of species-specific forest attributes, which can be used as alternatives to multivariate linear regression and nonparametric nearest neighbor models. Among the ANN models, SVR and MLP provide the best choices for prediction purposes as they yielded high prediction accuracies for species-specific tree volumes throughout.

Support Vector Regression for the Estimation of Forest Stand Parameters Using Airborne Laser Scanning

“…The nonparametric k-most similar neighbor method has been successfully tested for species-specific stand attributes estimation from laser data [9]. In a comparison with k-most similar neighbor and artificial neural network models such as self-organizing map and multilayer perceptron, support vector regression (SVR) turned out to be one of the best suited methods for prediction purposes [10].…”

Analysis of Land Cover Classification in Arid Environment: A Comparison Performance of Four Classifiers

“…It has been reported by several researchers (Lippmann, 1987;Cybenko, 1989) that a single hidden layer should usually be sufficient for most problems, especially for classification tasks. The major efforts were focused on controlling the complexity of the model in order to avoid a too complex model structure which may lead into an over fitted ANN model (Niska et al, 2010).…”

Prediction of Species-Specific Volume Using Different Inventory Approaches by Fusing Airborne Laser Scanning and Hyperspectral Data

“…They demonstrated that using only ALS variables resulted in less accurate estimates compared to using combined ALS and images variables. Niska et al [18] compared the kMSN method with three artificial neural network modeling methods: the multilayer perceptron (MLP), support vector regression (SVR), and self-organizing map (SOM) at the plot and stand level. The results revealed that the SVR and MLP models reached the greatest prediction accuracy, the kMSN a lower accuracy, and the SOM the smallest prediction accuracy.…”