Google Earth Engine based assessment of vegetation, water, and land use dynamics in a diverse mountain–plain transition zone of northeast India

This study assesses multi-temporal land use/land cover (LULC) change and associated vegetation–water dynamics in the Ri-Bhoi district (Meghalaya, NE India), an ecologically sensitive mountain–plain transition zone under increasing development and land-use pressure. Google Earth Engine (GEE) and supervised random forest (RF) classification were applied to Landsat-5 (2010), Landsat-8 (2015), and Sentinel-2 (2020 and 2025) surface reflectance products (cloud threshold ≤ 10%). Normalized difference vegetation index and normalized difference water index (NDWI) were derived to evaluate vegetation condition and surface-water variability. The RF classifications achieved overall accuracies of 86–90% with Kappa coefficients of 0.81–0.87 across the four study years, and detailed confusion matrices and publicly accessible GEE scripts are provided in the Supplementary Material (Tables S1–S4; Section S3). Results indicate a persistent decline in forest cover from 1999.61 km2 (2010) to 1756.92 km2 (2025), corresponding to a net loss of 242.69 km2, while agricultural land expanded from 319.49 to 492.92 km2 (+ 173.43 km2). Built-up area increased overall (24.54–89.96 km2; + 65.42 km2) but showed inter-annual fluctuations, consistent with class-wise uncertainty discussed in the manuscript. Water bodies varied markedly (26.61 km2 in 2010, 44.41 km2 in 2015, 22.44 km2 in 2020 and 27.14 km2 in 2025), reflecting changing hydrological conditions captured by NDWI. Together, the LULC and index trends highlight growing anthropogenic impacts and demonstrate the value of reproducible, cloud-based workflows for regional landscape monitoring and management.