Exploring the impact of land use/land cover changes on the dynamics of Deepor wetland (a Ramsar site) in Assam, India using geospatial techniques and machine learning models

Wetlands are significant landscapes that help in maintaining the ecological services, providing habitat for flora and fauna, controlling floods and regulating climate. Wetlands have been under constant threats due to urbanization and land use changes. The present study has examined the impact of land use/land cover (LULC) changes on spatio-temporal dynamics of Deepor Beel wetland, a Ramsar site in India. Multi-temporal Landsat satellite images for 32 years (1990–2022) were utilized to analyze the wetland dynamics. Wetland was delineated using modified normalized difference water index (MNDWI). Neural networks (Nnet), random forests (RF) and support vector machine (SVM) models were employed for preparing land use and land cover (LULC) maps. Water consistency was assessed using water presence frequency (WPF). The Landscape Fragmentation Tool (LFT) was utilized for computing various fragmentation indices. The findings revealed a consistent decline in wetland area from 15.16% in 1990 to 8.96% in 2022 primarily due to its transformation into built-up and agricultural lands. RF model was found more suitable than other models for LULC classification and change detection. Water presence frequency (WPF) analysis has shown marked variations in wetland area during pre- and post-monsoon seasons. Fragmentation analysis indicated that the number of patches has increased in periphery of wetland. Thus, this study calls for effective land use planning, reduction in wetland dependency of communities, creation of awareness among communities towards wetland restoration and conservation. The findings of the study may help the policymakers and conservationists for long-term sustainability and effective wetland management.