Exploring agricultural classes from available land use and land cover maps

Our goal is to exploit land use and land cover maps obtained from five selected initiatives, analyzing the existence of discrepancies among them and determining whether these differences are attributed to natural changes or to potential errors and inconsistencies in representing mapped agricultural classes. To achieve this goal, we analyzed data from the MapBiomas Brazil Collection, the University of Maryland Global Land Analysis and Discovery laboratory LULC Maps and Soybean Maps, the Copernicus Global Land Cover Service, and the Environmental Systems Research Institute over a 5-year period starting in 2015 (2015-2020). We conducted local change and increment analyses for the agricultural class in specific areas of Mato Grosso and Paraná states in Brazil with distinct landscape characteristics. Our findings indicate a discrepancy among initiatives in their attempts to represent agricultural classes in homogeneous and heterogeneous areas, emphasizing the need for caution when selecting a specific mapping for representing croplands.