Integrating machine learning and SAR-derived flood indices to assess the railway-induced waterlogging extent

Risk arising from waterlogging in low-relief floodplain areas is manifested primarily not only by extreme rainfall but also by large linear infrastructures such as elevated railway lines or road embankments, which disrupt natural drainage pathways. Conventional flood mapping approaches often fail to capture these anthropogenic controls. This study presents an integrated framework combining machine learning techniques and Sentinel-1 synthetic aperture radar (SAR) data to map flood extent and identify infrastructure-induced waterlogging along a railway corridor in Keonjhar district, Odisha, eastern India. Time series Sentinel-1 SAR data were analysed to extract inundation and surface moisture signatures using few flood indices. The infrastructure-induced topographic modification has been quantified using two Digital Elevation Models (DEM) representing two different time periods: the first one is the pre-infrastructure SRTM DEM, and the second one is the recent high-resolution DEM generated from drone-based orthophotos. Flow accumulation and watershed boundaries have been independently derived from both DEMs to evaluate changes in drainage pathways caused by the railway embankment. After watershed delineation from two DEMs, runoff coefficients were estimated, allowing a comparative assessment of pre- and post-infrastructure hydrological response. These terrain- and watershed-based variables, together with station-based rainfall data and SAR backscatter features, were used as input parameters in a Random Forest model to classify flooded, waterlogged, and non-inundated areas, with particular emphasis on zones adjacent to the railway alignment and cross-drainage structures. The results reveal that the persistent inundation patterns is largely as a consequence of natural flow obstruction by the railway embankment and inadequate cross-drainage connectivity. By highlighting these problems, the proposed methodology helps to identify infrastructure-driven flood augmentation and supports informed planning for designing any drainage-railway crossings, strategies related to flood mitigation, and climate-resilient transport infrastructure in vulnerable regions.