Combining data cubes and GEOBIA for land use and land cover mapping

The state of Mato Grosso, Brazil, materializes the dichotomy between conservation and agricultural production. This scenario reinforces the need for accurate land use and land cover (LULC) information to subsidize policies related to food security, climate change, deforestation, and agriculture dynamics. Satellite Image Time Series derived from the Moderate Resolution Image Spectroradiometer (MODIS) sensor are valuable to produce up-to-date LULC classifications, especially by analyzing reflectances at the pixel level. However, studies assessing the use of spatiotemporal heterogeneity and correlation features are scarce. In this study, we applied GEOBIA to produce a LULC map of Mato Grosso from the 2016/2017 harvest period. We used the Multiresolution Segmentation algorithm to derive spatiotemporal objects of the MODIS data cube and the Support Vector Machine to classify it. Our results underline the heterogeneity of Mato Grosso’s landscape. The overall accuracy of 0.95 indicates the potential of the approach.