Comparing different classifications of satellite imagery in forest mapping (Case study: Zagros forests in Iran)

Forest mapping is essential to manage natural resources and environment, land use plans and also to determine land potential and it is defined as one of the main resource for adjusting development programs. Forest mapping is a difficult process that in all projects has financial and time constraints. Satellite data are one of the quickest and low cost methods for forest mapping. In recent years, researchers have produced different types of forest maps using these data. There are several methods for classification of satellite images. The present study was aimed to determine the most appropriate image classification of ETM+ in Bivareh forests, a part of Zagros forests. Results showed that support vector machine (SVM) with Kappa coefficient 0.7069 and overall accuracy 88.65% is more accurate than other methods. The methods followed the order of accuracy form SVM, the maximum likelihood mahalanobis distance, the minimum distance, spectral information divergence, binary codes, parallelepiped to spectral angle mapping, respectively. The results of this study can be used as an effective tool for extracting forest maps with high accuracy.

Image classification, Forest mapping, Remote sensing, Kappa coefficient, Ilam province