EVALUATING VARIOUS METHODS OF VEGETATIVE COVER CHANGE TREND ANALYSIS USING SATELLITE REMOTE SENSING PRODUCTIONS (CASE STUDY: SISTAN PLAIN IN EASTERN IRAN)

As vegetation of any region can change over time due to various natural and human factors, the study of changes in the vegetation trend, especially in arid and semi-arid regions, has always been of great importance for the management of water and soil resources as well as vegetation. In this study, the NDVI products of Terra Satellite MODIS sensor (MOD13A3), with a spatial resolution of 1x1 km for a 15-year statistical period (2000-2014), were used to study the changes in the vegetation trend on a pixel-based scale during April, May and June in Sistan plain in eastern Iran. Four statistical methods, namely, simple moving average, simple exponential smoothing, double exponential ordering, and classical linear regression were used to detect long-term changes in the vegetation trend of this plain. The error rates of the models were then calculated using the three indicators of mean absolute deviation (MAD), mean square deviation (MSD) and mean absolute percentage deviation (MAPD). Analysis of these indicators showed that classical linear regression was the best model for detecting changes in the vegetation trend thanks to its lower error than others. Based on the selected statistical method, the most increasing and decreasing changes in the NDVI values were observed in the northeast, and the east and center of the plain, respectively. Finally, it was found that the use of trend analysis along with the classic linear regression method in a pixel-based scale could be a suitable method for revealing long-term vegetation changes in an arid and hyper arid climate.