A multiple land use change model based on artificial neural network, Markov chain, and multi objective land allocation

In this paper, a new combination of Artificial Neural Network (ANN), Markov Chain (MC), and Multi Objective Land Allocation (MOLA) was proposed and evaluated to simulate multiple land use changes using GIS-based techniques and multi temporal remote sensing data. The main objective of this paper is to predict land use changes for Tehran, the biggest and capital city of Iran. In this regard, by integration of ANN, MC, and MOLA, we found the pixels that have the highest tendency to change their states from one land use category to others. An ANN model was applied to create Transition Potential Maps (TPMs), and an MC model was used to calculate the quantity of the changes. Finally, a MOLA model was employed for spatial allocation of new changes. In order to analyze the effects of proximity, three types of neighborhood filters were combined with MOLA. The proposed method achieved 92.62%, 95.49%, and 92.74% of kappa index of agreement (KIA), overall accuracy (OA), and kappa of location (Klocation), respectively. This method was applied for Tehran to predict the situation in year 2020. The trend of the changes shows that the urban growth is moving toward southwest of the city, where the areas with poor infrastructure are situated.