Combining topography and reflectance indices for better surface water detection

Abstract

Since the Normalized Difference Water Index (NDWI) was proposed, water indices have served as useful tools for surface water detection. However, existing water indices are highly influenced by atmospheric and other environmental conditions and suffer from limited performance, especially in urban areas. At the core of the limitation is the sole dependency on the spectral distribution of reflectance signals. To overcome this, we propose to utilize topographic data as additional information for better water detection. Accordingly, the new index, namely Combined Water Index (CWI), is developed as the product of the topographic index and the reflectance-based index. These two indices excellently compensate each other: the former is free from noise issues but invariant over time while the latter can capture temporal dynamics of waterbody extents. The CWI is applied to four study areas of different development levels (natural, medium-sized cities, and megalopolis) in the Han River basin, South Korea. The water detection results of the CWI is promising, particularly in the heavily developed urban setting, demonstrated through visual images as well as various statistical measures.