Sentinel-1 and Sentinel-2 Based Crop Classification Over Agricultural Regions of Navarre (Spain)

The objective of this article is to compare the classification accuracies obtained using Sentinel-1 and Sentinel-2 data and to evaluate the eventual benefits of their combination, as a means to support Common Agricultural Policy controls. With this aim, data from two contrasting agricultural regions of Navarre (Spain) for year 2017 were used. The available Sentinel-1 and Sentinel-2 scenes were processed and a sample of farmers' CAP declarations and field inspections were used for training and testing a Random Forests classifier. Results showed a slightly better performance of Sentinel-2 data, which improved ~5% when combining both data types. A classification based on a selection of features (NDVI, Sentinel-2 B11 and VV backscatter) performed almost the same with a much smaller computational cost.