Enhancing the Normalized Difference Water Index for Improved Urban Flood Detection

Accurate urban flood detection is crucial for effective disaster management and urban planning. Traditional indices like the Normalized Difference Water Index and Modified Normalized Difference Water Index often produce inaccurate results due to spectral confusion in urban areas and sensitivity to shadows. Moreover, MNDWI's reliance on the Shortwave Infrared band limits its use with certain sensors and UAVs. To overcome these limitations, this study introduces the Enhanced Normalized Difference Water Index, which incorporates a ratio of green band reflectance to NDWI to minimize urban noise. We tested ENDWI using satellite datasets from Landsat, Sentinel-2, WorldView-4, and Pleiades-Neo during and after flood events in two urban areas. The results demonstrate that ENDWI significantly improves the delineation of water bodies, wetlands, and non-water areas by reducing noise from urban structures. Unlike NDWI and MNDWI, ENDWI values center near zero, potentially enabling automated thresholding for water extraction. While shadows from tall buildings remain a challenge, these findings highlight ENDWI's potential for automated urban flood detection, offering valuable insights for flood management and urban planning. Further research is recommended to refine its application and address the remaining challenges.