IDENTIFICATION OF WATER BODIES IN SENTINEL-1 RADAR TIME SERIES

• Autor(a) principal: de Castro Porto Costa, Evelyn
• Data de Publicação: 2023
• Outros Autores: Pereira dos Santos, Mikaella, Thomaz de Aquino Ribeiro, Eduardo, Garcia Rosa, Milton, Maria Moura de Almeida, Paula, Sanchez Vicens, Raúl
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Revista Geographia (Niterói. Online)
• Texto Completo: https://periodicos.uff.br/geographia/article/view/56659
• topic: Imagem de Radar; Sensoriamento Remoto; Aprendizado de Máquina; Google Earth Engine; Água; Radar Image; Remote sensing; Machine learning; Google Earth Engine; Water; Imagen Radar; Detección remota; Aprendizaje automático; Google Earth Engine; Água

This article aims to address the study carried out for the identification and cartography of water bodies in the state of Rio de Janeiro, using multitemporal remote sensing techniques in Radar images. It’s aims to contribute methodologically to the mapping of water bodies and wetlands, very specific targets that require detailed criteria for their correct classification, due to their dynamics. As a research methodology, machine learning algorithms were used on the Google Earth Engine platform, adopting Sentinel 1 – C-band images to identify these objects. A monthly time series of radar images was used, making it possible to test their potential in identifying these objects. As a result, the classification and quantification of the natural coverage of the state of Rio de Janeiro was obtained, considering the 12 months of the year 2018. The result allowed identifying the spatiality of the water bodies, in the different periods of the year, without atmospheric interference, which corresponds to a methodological difference in the attempt to map the annual flood dynamics. Mapping validation showed an excellent Kappa index (0,93), highlighting the potential of using radar images for mapping water bodies.