Spatio-temporal Modeling of Landscape Changes using Markov Chain Compilation Model and Automated Cells (Case Study: Arid and Semi-Arid Area Dehloran)

Detection and prediction of changes in landscape, is necessary for the maintenance of an ecosystem, especially in developing countries with rapid changes and without planning. The object of this research, is monitoring landscape changes in past and it’s simulation for future using Markov chain Consolidated and automated cells (CA-Markov) in arid and semi-arid region of Meymeh Dehloran, Ilam. Landsat satellite images of (TM) 1985, Landsat (TM) 2000 and Landsat (ETM) 2016 were used. Change detection maps were prepared in seven classes of agriculture, Forest, fair range, poor range, rocky protrusions, residential land and salt land using supervised classification ARTMAP FUZZY neural network. Accuracy of the classification landscape maps for 1985, 2000 and 2016, are 93, 95 and 93 percent, respectively. Changes in landscape were predicted for 2030, using Markov chain model and automated cells. Predicted matrix results based on 2001 and 2016 maps showed that in span of 2016-2030, it is likely that 13% of agricultural land, 54% of Forest, 48% of the fair range, 82% of poor range, 55% of rocky protrusions, 52% of the residential land, 93% of salt lands and marsh land converted to other land uses. To validating the model, simulated landscape map of 2016, were compared with satellite image classification of the same year. Kappa coefficient was 87%, which shows the high capabilities of CA-Markov model to simulate landscape changes in arid and semi-arid region of Meymeh Dehloran.