vegetation change detection using multi-temporal remotly sensed data during recent three decades by artificial intelligence technique (Case study: protected area of Bashgol)

Quantitative and qualitative information of vegetation and its changes in duration of time as a basic foundation of determination of  habitat quality, priority of protected area and also determination of price of ecosystem services in order to optimum management of natural resources and sustainable development is a very important technical point. In other hand, researchers are interested in remote sensing as an efficient tool to access timely and accurate information about land coverage especially vegetation coverage. In current study post classification comparison method among different methods of change detection was used because of possibility of achieving optimum accuracy by using an efficient and accurate classification method. In order to determine vegetation classes by field-based resources and Landsat images and also, slope-based and distance-based vegetation indices derived from these images, two artificial neural networks; percentage of vegetation cover (overall accuracy 94.3% and mean square error 5.7% for test data) and dry weight of standing biomass (overall accuracy 86.6% and mean square error 11.4% for test data) were built and vegetation maps according to these qualified models were prepared. Results of this research show high capacity of the artificial intelligence technique in vegetation classification if, the average number of field-based samples and variety of images in terms of time were possible. By the vegetation classification maps made, change detection maps with pixel by pixel comparison that show change classes from class … to class … in three duration of time; 2000-2015, 1986-2000 and 1986-2015 were prepared. Accuracy of these maps is totally depended on classification accuracy and demonstrated qualitative enhancement in a wide area of case study after conservation.