Flooded wetlands mapping from Sentinel-2 imagery with spectral water index : a case study of Kampinos National Park in central Poland

Flood monitoring of wetlands and floodplains is a new issue in remote sensing, as compared to the mapping of open water bodies. The method based on spectral water indices, calculated on the basis of green, red and shortwave infrared bands, is one of the most popular methods for the recognition of a water body in multispectral images. The recently introduced Sentinel-2 satellite can provide multispectral images with high spatial resolution. This new data set is potentially of great importance for flood mapping, due to its free access and the frequent revisit capabilities. In this study, three popular water indices (Modified Normalized Difference Water Index, Normalized Difference Pond Index and Normalized Difference Turbidity Index) were used. The efficiency of the proposed method was tested experimentally using the Sentinel-2 image for the Kampinos National Park in Poland. The experiment compared four extraction algorithms including three based on individual water indicators and one on a combination of them. The results showed that the 10-metre false colour composite produced significantly improved the recognition of flooding in wetland areas by comparison with single spectral water indices. In this way, flooded wetlands were mapped based on the Sentinel-2 data set for the years 2017-2018.