Modeling the spatial changes of Zagros forests using satellite imagery and LCM model (Case study: Bastam, Selseleh)

During the last decades, the decrement of forest areas has come to attention in regional and global scales. This research aims to investigate the forest changes using satellite imagery, Artificial Neural Network and Markov chain in Bastam area, Lorestan province. Land use maps for the years 1364, 1379, and 1394 were prepared using the Maximum Likelihood Classification and Landsat TM and OLI Sensors. Land use changes modeling was done, using LCM model based on Artificial Neural Network and 7 effective variables include altitude, proximity of the residential areas, slope, direction, proximity of the roads, proximity of the river and land use map. The land use map for the year 1394 was predicted using the Markov chain modeling method. To evaluate the accuracy of the results, the error matrix was formed between the predicted map and the ground reality map for this year. The results of the first period showed that the highest increase in the agricultural area and the largest reduction in the forest area were 380 and 425 hectares respectively. In addition, the variables of altitude and proximity of residential areas with the Cramer correlation coefficient of 0.47 and 0.43 respectively, had the most effect on land use changes and forest degradation. Finally, the comparison of modeling and reality maps of 1394 showed the Kappa coefficient of 0.89, which indicates the proper performance of the LCM model in predicting land use changes.