Methods and Software to Understand and Build the MapBiomas Data

MapBiomas is a multi-institutional initiative involving NGOs, universities, think-tanks and tech companies dedicated to developing the most advanced, complete, and detailed time series of annual land cover and land use change maps in Brazil that support and advance sustainable management and conservation of natural resources. Map data accuracy derives from various decisions concerning the selection of legend categories and the steps used to process the remotely sensed images. MapBiomas has developed various temporal and spatial filters applied as a post-classification process to eliminate temporal differences derived from data error; however, these filters have various influences on the data and we need a better understanding of these influences. This project will develop new ideas and methods (software and publications) to explain transitions among land cover categories, whether the explanation concerns processes on the ground or errors in the data.