Integração de imagens ópticas e SAR com processamento em nuvem no mapeamento de uso e cobertura da terra no Cerrado

Abstract

Land use and land cover mapping is important for environmental monitoring and land management. Synthetic aperture radar (SAR) from Sentinel-1 (S-1) and optical sensor MSI/Sentinel-2 (S-2) time series provide favorable conditions for land use and land cover mapping at orbital level due to their spectral, spatial and temporal resolutions. This work has been based on the hypothesis that the combination between S-1 and S-2 time series allows for greater accuracy in land use and land cover mapping in the Cerrado biome. The images were classified using the Random Forest algorithm by different hyperparameters in the Google Earth Engine cloud processing platform. The classifications obtained with the S-2 data alone (Kappa = 89.99) were better than those obtained with the S-1 data alone (Kappa = 75.78). The classification efficiency increased when combining the S-1 and S-2 data (Kappa = 93.07). Shortwave infrared, VH polarization from SAR data, CAI (Cellulose absorption index) and Hall Cover indices were the most important variables in the integrated classification of the S-2 and S-1 data, respectively.