Long-Term Assessment of Flood Inundation Using SWOT Satellite Observations

Flood inundation mapping is crucial for effective disaster management and well informed land use planning in flood prone regions. Optical sensors are often ineffective during flood events due to persistent cloud cover, limiting the usability of visible and near infrared imagery. While the SAR overcomes these limitations, there are no missions that monitor water from space in two dimensions. In contrast, the SWOT mission, through its Ka-band Radar Interferometer (KaRIn), provides cloud penetrating measurements of water extent and elevation, making it a powerful tool for long term flood inundation mapping despite its 21 day revisit cycle and limited real time availability. However, basin scale application of SWOT data is challenged by large data volumes, variable spatial resolution (10 to 60 m), overlapping swaths, and the absence of a standardized spatial indexing system. As a result, identical geographic locations may not appear consistently across multiple granules, complicating continuous time series extraction. This study uses the L2_HR_PIXC dataset, which is irregularly sampled and more complex than standard raster products, but better preserves critical hydrological information that is often lost during resampling. Although often neglected due to processing complexity, it contains crossover observations that hold  valuable information critical for accurate inundation analysis. To address these challenges, a dedicated national scale database was developed for India to efficiently organize and integrate spatiotemporal SWOT pixel data at 30 m resolution. Leveraging this database, a probabilistic clustering based approach was implemented to map seasonal and permanent flood inundation using both water extent and elevation information. The developed framework enables efficient flood inundation mapping over large spatial extents, including river basin scales, overcoming limitations of conventional hydraulic models that are data and resource intensive and often restricted to regional applications. The methodology was applied to major flood prone regions of India namely Bihar, Assam, Manipur, Kerala, and Andhra Pradesh, spanning diverse latitudes and exhibiting substantial variation in SWOT observation frequency. The resulting inundation maps were validated using ground based observations and Sentinel satellite imagery across multiple flood events. These maps effectively differentiate recurrent seasonal flood zones from rarely inundated areas that may be more vulnerable during extreme events, offering valuable insights to support targeted flood risk management and disaster preparedness.

Keywords: SWOT, Cluster, India, Inundation map, Flood, Database