Evaluation of Pixel- Based and Object Oriented classification approaches for Determination of Land Use Changes in Van Lake Basin and it Comparison with Lake Urmia Basin

The reason for dissimilar behaviors of Lake Urmia (Iran) and Lake Van (Turkey) during the recent decades is one of the main challenging questions. A part of the answer can be addressed by evaluation of their land use changes as the main indicator for the role of human impacts. This issue constructs the objective of present paper. For this aim, the pixel-based and object-oriented classification approaches were evaluated for obtaining the land use maps of Lake Van basin during 1987 to 2007. The applied pixel-based methods include: Mahalanobis Distance (MD), Maximum Likelihood (ML) and Support Vector Machine (SVM); and the object-oriented method includes SVM-fuzzy. Their comparison showed better performance of the object-oriented method such that Kappa Coefficient and Overall Accuracy were 0.81 and 0.86, respectively. Then, the results were compared with a similar research work for the Lake Urmia basin which was attempted. The results revealed that that the land use of Van basin has not significantly change, while the increase of cropped lands in Van basin was only 10000 ha, it was 136000 ha in Urmia basin. The most significant change in Urmia Lake relates to orchard area by increase of 273% (3513 ha to 13120 ha), whereas it was insignificant in Van (less than 20%). Definitely, these changes can increase consumption of water and reduce inflows to Lake Urmia.