Land Use/Cover Change Modelling in a Mediterranean Rural Landscape Using Multi-Layer Perceptron and Markov Chain (mlp-Mc)

Abstract

Mediterranean land use and land cover (LULC) have a very dynamic structure as a result of continuous transformation process due to anthropogenic effects and environmental gradients. LULC dynamics are important indicator of environmental condition in temporal and spatial scales. The aim of this paper was to simulate the future LULC of a Mediterranean type watershed located at the Eastern Mediterranean Region of Turkey by incorporating multi-layer perceptron (MLP), artificial neural network (ANN) and Markov chain (MC) approaches. Landsat TM/OLI images in 1990, 2003 and 2014 over the study area were classified using hybrid classification approach. The Kappa statistics of the hybrid classification that combines K-means, decision tree and object based classification method for these three images were 0.81, 0.85 and 0.87 respectively. The LULC map of 2014 was simulated using LULC maps of 1990 and 2003 for calibration and validation. The simulation results were compared with the actual 2014 LULC map to assess the accuracy of the simulation, and the rate of overlap was found as 89%. LULC map of 2025 was estimated using LULC maps of 2003 and 2014. These results indicated that, the area of bareground will reduce 13.31% whereas the rate of forest and agricultural area will increase 8.70% and 6.51% respectively.