Comparative analysis of Sentinel and MODIS data in the problem of crop classification

Sentinel-2 and MODIS data sources are widely used for crop classification. However, they differ in time and spatial resolution and the question of a better data source for classification is still open. The article aims to compare both data sources in terms of crop classification quality. We based our research on the Random Forest classifier of the NDVI time series. The data sources were compared using two datasets with large and small fields. As a result, we found that the Sentinel data give almost the same probability of correct classification as MODIS data. However, for fields less than 50 hectares, the classification using Sentinel data is slightly better than the classification using MODIS data. The results indicate that the high frequency of MODIS imaging allows a similar classification probability at a lower spatial resolution. However, for small fields, the Sentinel-based classifier remains the preferred choice.