Advanced possibilities have emerged in the recent years for semi-automatic crop type mapping at national level due to the availability of Sentinel-1 and -2 satellite data. In this study, 14 crop type classes were mapped over Bulgaria using three bi-monthly composite image mosaics for 2019 generated in the Google Earth Engine (GEE) cloud computing platform. The overall accuracy, when both Sentinel-1 and -2 mosaics were used, was 78%, while the accuracy was slightly less when only Sentinel-2 data was used (75%). The accuracy was highest for "Cereals", "Maize", "Sunflower", "Winter rapeseed", and "Rice" - over 80% for both user's and producer's. However, the accuracy for classes such as "Vegetables", "Technical crops", "Forage crops", "Fallow", etc. was low. These classes represent categories suitable for the agricultural practice and statistics but are too general and difficult to distinguish using satellite data. It was found also that accuracy tend to be higher for larger parcels. Using composites with higher frequency and adapting the legend classes to include only crops which are similar in phenology and morphology are suggested as possible ways forward.