Land use change and soil salinization in the Sundarbans: a machine-learning based analysis of long-term transformation and future projections

Abstract

Quantitative data on coastal land use changes are essential for effective resource management and sustainable development. In this study, we examined land use and land cover (LULC) changes, along with erosion and accretion, in the climate-sensitive Sundarbans of India using satellite imagery from 1973 to 2022. Over this period, approximately 478 km 2 of land was lost to erosion, while 408 km 2 were gained through accretion. Although single cropping remains the predominant land use system, the areas dedicated to settlements and aquaculture grew 7 times and 2.4 times, respectively. The double-cropped area expanded by 75.8%, while the mangrove area remained unchanged. Among all land use categories, single-cropped land was found to be the most vulnerable to loss. The soil salinity area of the arable land, calculated using the canopy response salinity index (CRSI) from satellite imagery, rose from 2246.2 km 2 to 2669 km 2 in the 30 years since 1989. Using a machine learning algorithm (cellular automata-artificial neural network (CA-ANN)), which incorporated both anthropogenic factors and projected temperature and rainfall data as explanatory variables, we estimated that by 2049, the settlement area will increase by 31.6%, aquaculture will expand by 30%, and vegetation cover will decrease by 12.6%, compared to 2019 levels. The LULC change trend and coastline dynamics are expected to further exacerbate land degradation as the model predicts an increase in soil salinity by 5% over the same period. The results help farmers and policymakers to develop effective land management plans to enhance community readiness and resilience against vulnerabilities.

Bibliographic Information