Ecosystem services value assessment and forecasting using integrated machine learning algorithm and CA-Markov model: an empirical investigation of an Asian megacity

Abstract

The ecosystem services in an area are quite dependent on its ambient land use and land cover (LULC) attributes. Here we assess the spatio-temporal distribution of Ecosystem Services Value (ESV) for the years 1990, 2000, 2010, 2020, based on the then existing LULC aspects of the Kolkata Urban Agglomeration in eastern India. Further, these are simulated for 2030, 2040 and 2050 to determine the future potential ESV. The respective LULC layers were extracted from Landsat images using the support vector machine method and future projections were done using Markov Chain-Cellular Automata models. Results reveal that all LULC aspects are likely to significantly decrease except built-up tracts. The available ESV shall concomitantly decline by 29.7%, especially due to wetland loss. The ESV patterns also showed strong spatial correlation/clustering, with higher ESV patches locating along rivers/wetlands. These results can better inform management of high-value EVS components for sustaining/improving the urban environmental quality.