Analysis of Harmonized Landsat Sentinel data for land use and land cover mapping in areas of agricultural expansion in the Cerrado

Abstract

Different initiatives for mapping land use and land cover using orbital remote sensing have been carried out with the aim of supporting agri-environmental planning in the Brazilian Savanna (Cerrado). However, the complex regional agricultural dynamics require more detail in the discrimination of agricultural uses, which requires time series of satellite images (SITS). This study aimed to evaluate how the integrated use of Landsat 8 Operational Land Imager (OLI) and Sentinel-2 Multispectral Instrument data can contribute to digital land use and land cover mapping in medium spatial resolution. The municipality of Balsas, a region of agricultural expansion in the Cerrado of Maranhão State and relevant economic and agro-environmental interest, was chosen as a study area to evaluate the performance of Random Forest classification models generated with data from Landsat 8, Sentinel-2, and the Harmonized Landsat Sentinel 2 (HLS). The MSI, HLS, and OLI models presented similar performance, with Global precision of 0.857, 0.844, and 0.832, and Kappa index of 0.819, 0.803, and 0.787, respectively. The T-Test indicated that these results are not statistically different. However, despite the models’ performance, the final map created with the OLI data showed a high loss of information due to the absence of pixels in the time series. The new HLS data proved suitable for land use and coverage mapping, even on a large scale, at a resolution of 30 m, with precision and reduced processing time compared to the MSI data. These results can support decision-making in the planning and execution of new agricultural mapping of the Cerrado Biome.