Abstract
<title>Abstract</title> <p>Wetlands are increasingly recognised as critical water infrastructure providing wastewater treatment and reuse, water storage, flood buffering, and water-quality regulation yet global and national land-use/land-cover (LULC) products are not designed to support wetland-based water resources management. This gap is acute in peri-urban, wastewater-fed systems, where water functions and livelihoods hinge on fine distinctions that generic maps overlook. Focusing on the East Kolkata Wetlands (EKW), a Ramsar site whose community-managed, wastewater-fed aquaculture forms one of the world's largest natural wastewater-treatment and reuse systems, this study develops a hierarchical LULC framework tailored to water management. Using multi-temporal Sentinel-2 MSI imagery (February–May 2025), a multi-season water-frequency (hydroperiod) analysis, and a Random Forest classifier (80:20 training/testing) in Google Earth Engine, we implemented a seven-stage classification distinguishing aquaculture ponds, aquatic vegetation, agriculture, other vegetation, fallow land, fallow ponds, built-up areas, and active/closed landfill. The map achieved an overall accuracy of 89.29% (95% CI 85.8–92.7%; κ = 0.88; macro-F1 88.5%), with allocation disagreement (8.8%) exceeding quantity disagreement (2.0%). Water and water-dependent classes occupy 44.3% of the mapped wetland (≈ 5,593 ha), with managed aquaculture ponds the active water-storage and treatment surface covering 32.2% (≈ 4,071 ha), against 9.2% built-up encroachment concentrated on the western urban fringe. By resolving hydrologically meaningful classes managed versus idle water bodies, aquatic macrophytes, and hydroperiod-defined fallow ponds the framework converts Earth-observation data into an operational basis for wastewater-reuse management, water-quality surveillance, storage and flood-cushion protection, and encroachment control, directly supporting SDG 6 and the sustainable governance of urban water systems.</p>