The paper aims to concept and to implement an operational and optimized multi-temporal Sentinelimages based tools for management support of citrus area. A time series of optical and Synthetic Aperture Radar (SAR) Sentinel images covering the period of January - November 2017 wereacquired in order to identify 7 thematic categories (Young Citrus, Adult Citrus, Old Citrus, other plantations, cereals, vegetables and spring bare soil). Thirteen automatic supervised classification schemes with two tested classifiers (the classical Maximum Likelihood (ML) versus the advanced Support Vector Machine (SVM)), two used approaches (pixel based versus object based), and different SAR polarizations were investigated. To ensure an unbiased assessment, all schemes are tuned based on the same training and validation independent sets ground collected in April 2017. All classifications are evaluated based on F-score accuracy assessment metric. Overall classification map accuracy based on Sentinel-2 time series was 72.4%, Sentinel-1A – 63.6%, at the same time, data fusion of Sentinel-1A and Sentinel-2 provides overall accuracy of 80% that was higher by +8.3% and +17.1% for Sentinel-1 and Sentinel-2 based classifications, respectively. The last one was retained for crop and Citrus maps generation.
