Efficiency of crop identification based on optical and SAR image time series

Abstract

This paper assessed the use of optical and SAR imagery for crop identification in an operational context with a particular emphasis on actual crop diversity and information delivery time. Fifteen ERS and Radarsat and 3 optical images were used to discriminate agricultural crop types based on dedicated per-parcel classification and photo interpretation schemes. For crop area control, the efficiency concept was introduced as a complementary indicator of classification performance. A set of 6571 parcels were classified into 39 crop types from various combinations of images. The efficiency computed from an independent set of 899 parcels peaked based on a combination of optical images and 3 to 5 SAR images. Moreover, the delivery time of the relevant information was improved when SAR data was included. The hierarchical classification strategy based on nested classifications also improved the operational crop control system for all image combinations. Finally, this research documented the respective contributions of optical and SAR time series for any control system of agricultural land.