Random forest models based on Sentinel-2 multispectral indices for flood mapping

Optical satellite sensors represent a reference for Earth imaging applications, including land monitoring and flood management, directly allowing the visual interpretation of acquired scenes or the exploitation of surfaces’ spectral signatures. An extensive literature exists that proves the ability of multispectral satellite sensors in mapping flooded areas and water bodies (Albertini et al., 2022). Several multispectral indices have been developed for water segmentation in different contexts of varying degrees of landscape complexity. Simultaneously, the advancements in Machine Learning (ML) methods led to a proliferation of supervised and unsupervised algorithms applied to classification problems in the field of flood hazard and risk mapping. In the present study, four random forest (RF) models were used in combination with three spectral indices, namely the Modified Normalized Difference Water Index (MNDWI), the Normalized Difference Moisture Index (NDMI) and the Red and Short Wave Infra-Red (RSWIR) index, to map the extent of the flood event occurred along the Sesia River (Vercelli, Italy) in October 2020. A Sentinel-2 scene was acquired soon after the flooding event and spectral bands at 20m resolution were used in the analyses. The performances of the RF methods implemented with the use of the mentioned spectral indices were evaluated and compared using as reference map the delineation product delivered by the Rapid Mapping service of the Copernicus Emergency Management Service (CEMS). Results revealed some very interesting findings regarding the performances of the examined methods, which can become a well-established operational technique. Last but not least, the validation framework itself underlined the added value of Sentinel-2 and the Copernicus platform as a robust, rapid and cost-effective solution in flood mapping.

Keywords: floods mapping, spectral indices, machine learning, Sentinel-2, Italy

References:

Albertini, C.; Gioia, A.; Iacobellis, V.; Manfreda, S. Detection of Surface Water and Floods with Multispectral Satellites. Remote Sens., 14, 6005, 2022. (doi: https://doi.org/10.3390/rs14236005).