For many applied problems in agricultural monitoring and food security it is important to provide reliable crop classification maps. Using only optical data consistent identification amongst winter crops and separation of summer crops like sugar beet, maize, sunflower and soybeans may be difficult. Taking into account that the Sentinel-1 radar satellite provides free imagery since October 2014, we propose a new
approach to pixel and parcel-based classification based on multitemporal combination of optical satellite imagery and multitemporal
dual-polarization SAR data. Pixel-based classification maps are derived from an ensemble of neural networks, in particular multilayer perceptrons (MLPs). Parcel-based classification maps are obtained using different fusion methods of raster and vector information. The proposed approach is applied at regional scale for crop classifications using multi-temporal Landsat-8 images for the JECAM site in the Kyiv oblast of Ukraine in 2013-2015 and in the Odessa oblast in 2014-2015. In 2015, Sentinel-1A imagery for the period March-August is integrated.