GEOBIA and multitemporal segmentation for land use and land cover mapping: a case study in Mato Grosso, Brazil

With expressive crop production and rampant deforestation rates, the state of Mato Grosso, Brazil, represents the struggle between conservation and economic development. This scenario reinforces the need for accurate land use and land cover (LULC) information to support sustainable development policies. Satellite image time series from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor are relevant to produce up-to-date LULC classifications at pixels level. However, there is a need for understanding how it performs at the geo-objects level. In this study, we used Geographic Object-Based Image Analysis (GEOBIA) to produce a LULC map for Mato Grosso from the 2016/2017 harvest period. We derive spatio-temporal geo-objects of a MODIS data cube via segmentation and use Support Vector Machine (SVM) to perform the classification - comparing it with a reference map. The overall accuracy (0.75) and Kappa (0.64) indicate pros and cons of combining GEOBIA and MODIS for LULC mapping.