Anais do XX Simpósio Brasileiro de Sensoriamento Remoto

Brazil faces the challenge of balancing environmental conservation and economic development. An intensive use of irrigated agricultural areas may offer a powerful alternative to address this issue making methods for detailed mapping of irrigated croplands vital. In this study, we employed land surface phenological information derived from dense satellite image time series to identify irrigated croplands in Juazeiro, Bahia, a region which makes use of such management practice in the Caatinga biome. We used a 16-day composite Sentinel-2/Multispectral Instrument (S2/MSI) data cube developed by the Brazil Data Cube project and a Random Forest classification scheme to perform the analysis for the year 2020. The map was compared with MapBiomas and achieved 74,86% of agreement, showing the potential of phenology-derived information from dense S2/MSI time series to map irrigated croplands across the Caatinga biome.