Land Use Change Assessment of Helmand Plane of Afghanistan Using Maximum Likelihood, Decision Trees and Support Vector Machines Classification Algorithms

One of the major challenges in the Iran water resources management is reduction of Helmand River’s inflows that has happened several times in recent years. Increasing in the basin’s cultivated area (mainly after Kajaki dam) and the recent droughts are declared as the reasons; however none of them is analyzed in a quantative manner. This paper aimed to address this issue and at this stage evaluated land cover changes in the last two decades using remote sensed data. Due to importance of the selected algorithm for classification, three methods have been applied for the evaluation, including maximum likelihood classifier (MLC), decision trees (DTs) and support vector machines (SVMs). It is obvious that direct sampling was not possible from the study area, therefore the FAO land use maps, watershed atlas of Afghanistan (AIMS), and Google Earth images were applied in this regard. Considering the capabilities of Landsat (ET and ETM), three images for 1990, 2001 and 2011 during growing season (all in May) were prepared and processed for this study. The results were evaluated using Kappa coefficient and overall accuracy, which showed almost similar performances of the algorithms. It was concluded that the total cultivated area in the region increased from 103’000 ha in 1990 to 122’000 ha in 2001. Notably that this amount increased to 167’000 ha in 2011. These changes show definite impact of land use change on the river inflows. In spite of similar results of the applied methods, we found DT method more suitable for such an analysis due to its computation cost and efficiency as well as less relying on training samples.