Statistical features for land use and land cover classification in Google Earth Engine

The possibility of identifying and quantifying agricultural areas objectively and quickly is a relevant aspect in the Brazilian agricultural context, given the territorial extent of the country and the cultivated areas. The objective of this research was to create a methodology to apply an automatic land use and land cover classification on the São Francisco Verdadeiro River hydrographic basin, western region of Paraná state, with Landsat-8 images in the Google Earth Engine geospatial processing platform. This approach has compared commonly performed classifications with classification by statistical features (median and standard deviation). Results indicate as the best dataset was the bands B2 to B7, of each image and NDVI available in the studied period, which obtained Overall Accuracy (OA) of 99.0% and Kappa index (K) of 0.9867. The dataset of statistical features obtained OA of 97.3% and K of 0.9644, being also a reliable and representative way for LULC mapping. •NDVI median and SD show potential for crop cycles aggregation and reduction.•Statistical metrics present well-defined groups in the dispersion diagram.•Statistical mapping can differentiate Agriculture, Pasture, Forest and Water Bodies.