On the use of temporal-spectral descriptors for crop mapping, monitoring and crop practices characterization

Irrespective if remote sensing data are acquired by active or passive sensors, high or medium resolution, key information is the temporal signature. This is particularly true - but not limited to - agriculture, where the spatio-temporal dynamic is significant. Spectral (here meant in frequency and polarimetric terms) information, definitely, complements the temporal one. In this paper, temporal-spectral descriptors are derived from sigma nought time series acquired from various Synthetic Aperture Radar (SAR) systems over different agro-ecological zones in Senegal, The Gambia, Vietnam. It is shown that: - a limited set of temporal descriptors is sufficient to generate a reliable crop map; - the selection of the appropriate time period is crucial; - the temporal combination of wavelengths and polarizations may enhance the level of detail and product´s reliability; - the use of temporal descriptors derived from multi-annual, annual, and seasonal time series data provides, from an agronomic perspectives, complementary information; - temporal-spectral descriptors have an agronomic meaning, hence they should be used in knowledge based classifiers; - by sparse time series the adoption of temporal-spectral descriptors is more effective than a dedicated crop detection algorithm.