Land Use Change Prediction using a Hybrid (CA-Markov) Model

Landsat data for 1992, 2000, and 2013 land use changes for Ekbatan Dam watershed was simulated through CA-Markov” model. Two classification methods were initially used, viz. the maximum likelihood (MAL) and support vector machine (SVM). Although both methods showed high overall accuracy and Kappa coefficient, visually MAL failed in separating land uses, particularly built up and dry lands.Therefore, the results of SVM were used for Markov Chain Model and “CA” filter to predict land use map for 2034. In order to assess the ability of “CA Markov” model, simulation for 2013was performed. Results showed that simulated map was in agreement with the existing map for2013 at 84% level. The land use map prediction showed that built up area of 0.8298 km2 in 2013 will increase to 1.02113 km2 in 2034. In contrast, irrigated agriculture will decrease from 17.33 km2 to 17.16 km2, and rain fed agriculture from 45.07 km2 to 44.49 km2. Results of this research proved the application of “CA Markov” model in simulating the land use changes.

Land Use Change Prediction using a Hybrid (CA-Markov) Model Keywords:

Cellular Automata model , MAL , SVM