Parameterization of a Bayesian Normalized Difference Water Index for Surface Water Detection

The normalized difference water index (NDWI) has been extensively used for different purposes, such as delineating and mapping surface water bodies and monitoring floods. However, the assessment of this index (based on multispectral remote sensing data) is highly affected by the effects of atmospheric aerosol scattering and built-up land, especially when green and near infrared bands are used. In this study, a modified version of the NDWI was developed to improve precision and reliability in the detection of water reservoirs from satellite images. The proposed equation includes eight different parameters. A Bayesian procedure was implemented for the identification of the optimal set of these parameters. The calculation of the index was based on Sentinel-2 satellite images of spectral bands collected over the 2015–2019 period. The modified NDWI was tested for the identification of small reservoirs in a subbasin of the Belice catchment in Sicily (southern Italy). To assess the effectiveness of the index, a reference image, representing the actual reservoirs in the study area, was used. The results suggested that the use of the proposed methodology for the parameterization of the modified NDWI improves the identification of water reservoirs with surfaces smaller than 0.1 ha.