Land use land cover modeling using optimized machine learning classifiers: a case study of Shiraz, Iran

Land degradation is one of the most critical environmental challenges affected by the change of the land use land cover in the last few decades. Land degradation has a negative impact on the livelihood and food security worldwide. The annual cost of land degradation caused by land use land cover change is estimated at 231 billion US dollars. It should be noticed that recent droughts caused by global climate change and, on the other hand, population growth have increased the rate of urbanization in Iran. In this study, to monitor the recent urbanization, several advanced and state-of-the-art classification algorithms, including small and large neural networks, support vector machine, and the genetic algorithm multi-layer perceptron (GAMLP) model, are developed in R and MATLAB programming languages for land use land cover modeling in the years 1990 and 2018 of the Shiraz City. For the year 1990, the SVM algorithm has the best performance in terms of overall accuracy and kappa indices with the values of 93.55 and 89.22, respectively. The SVM classifier has better performance for the year 2018 as well, with values of 98.37 and 95.76 for the overall accuracy and kappa indices. The developed GAMLP model has better performance over the other two small and deep neural network classifiers with the values of 92.88 and 89.09 for the year 1990 and values of 97.38 and 93.12 for overall accuracy and kappa indices for the year 2018. Based on the results, built up areas have increased in the study area where the vegetation regions decrease from 1990 to 2018.