An investigation of past and future LULC using multilayer perceptron-Markov chain techniques: a case study of a Himalayan smart city (2005-2040)

This study focuses on the Dehradun tehsil of India for urban growth modeling. It is one of the cities under 100 smart city plans by the Government of India. Land Use Land Cover (LULC) mapping is carried out for different years, i.e., 2005, 2009, 2013, 2017 and 2021. The LULC analysis shows that the built-up was 12.04% in 2005 and has continuously increased, reaching 26.53% in 2021. As per the change detection outputs, the maximum built-up gain of 5099 hectare occurred between 2005 and 2013, mainly at the loss of agriculture and forest. A Multilayer Perceptron (MLP) neural network was used to perform transition potential modeling. Four driving variables (slope, DEM, evidence likelihood and distance from road) and five LULC transitions (water to built-up, forest to built-up, agriculture to built-up, barren to built-up and forest to agriculture) were considered to generate transition potential. The transition potentials generated through MLP helped to predict LULC using the Markov Chain. The predicted LULC of 2021 was validated against the actual LULC for the same year. Finally, LULC for the years 2030 and 2040 was predicted. They forecasted a rise in commercial, industrial and residential area at the expense of forest cover and agricultural lands. The findings of this study could be valuable for government agencies, urban planners and respective departments.