Spatio-temporal flood frequency dynamics in Kaziranga National Park using multi-year Sentinel-1 SAR and vegetation indices

Floods along the Brahmaputra are frequent, devastating, and dynamic, bringing severe hydrologic, ecological, and geomorphological alterations in Kaziranga National Park (KNP), a UNESCO World Heritage Site. This study presents a spatio-temporal Flood Frequency (FF) assessment and its ecological implications using multi-sensor satellite data. A total of 228 Sentinel-1 Synthetic Aperture Radar (SAR) and 8 Sentinel-2 optical images (2015–2022) were analysed to derive annual FF maps, identify hotspot regions of major FF change, and quantify vegetation response using Normalised Difference Vegetation Index (NDVI), Normalised Difference Water Index (NDWI), and Modified NDWI indices. Flood inundation maps were generated through SMILE random forest classification of pre and during-flood SAR composites in Google Earth Engine. Results show typically high FF along the Brahmaputra channel and permanent waterbodies and wetlands, whereas elevated southern highlands exhibited low FF. Extreme flood years (2017, 2020, 2022) recorded mean FF >40%. FF trend analysis revealed 81.72% of KNP remained stable, 10.04% experienced increasing FF linked to erosion and channel encroachment, and 8.24% showed decreasing FF due to deposition. Flood ridges (70% FF) digitized between 2015 and 2022 indicated average channel migration of ∼157 m upstream and ∼311 m inward, with localized shifts exceeding ∼1 km. DBSCAN clustering of FF slopes delineated 68 clusters, comprising 43 hotspots and 25 cold spots, marking zones of hydro-geomorphic instability. Vegetation analysis revealed NDVI decline in hotspots due to inundation and erosion, while cold spots showed greening from sediment deposition and grassland regeneration. The integrated SAR–optical approach underscores dynamic floodplain alterations and ecological responses, providing vital inputs for monitoring riverine ecosystems and strengthening conservation strategies in KNP. • Flood frequency mapping using Sentinel-1 SAR imagery in Google Earth Engine • High flood frequency associated with Brahmaputra channels, wetlands, seasonal lakes, and erosion zones • Integration of Sentinel-1 and Sentinel-2 data reveals flood and vegetation dynamics • DBSCAN clustering delineates hydro-geomorphic instability zones in the floodplain