INFORMATION CONTENT OF MULTI-SPECTRAL SAR DATA

The prospects of multi-frequency (MF) SAR data have been investigated for six thematic applications and CNR ISSIA has been involved for the agricultural applications (i.e. crop mapping, crp biomass retrieval and soil moisture); following the indications from a literature survey and the selection of dataset suited for implementing multi-spectral retrieval and classification, then the study focused to demonstrate the benefits of integrating multi-temporal and MF SAR data for the three inter-linked agricultural applications. For crop mapping, the contribution of SAR multi-temporal and MF data has been investigated. Two classification methods have been adopted and results compared. The first method is the supervised ML classification, whereas the second is an unsupervised approach referred to as MFT. Results indicate that: the crop classification accuracy obtained with Sentinel-1-only data is usually higher than 85%; the additional improvement brought by X- and L-band data is secondary and depends on the number and type of crop classes and on the time in the growing season they are acquired. The dominant role of C-band-only data is mostly due to the Sentinel-1 systematic acquisition scenario. Under these circumstances, the inclusion of X- and/or L-band SAR time series is recommended when they are acquired with consistent revisit time, incidence angle and spatial resolution. In addition, multi-spectral SAR data can be particularly valuable when they are acquired during the early stages of growing season. For the crop biomass application, semi-empirical relationships relating MF SAR observations to crop-specific biomass have been used for retrieving time series of crop biomass. The main advantage of using MF SAR data relies on the possibility to simultaneously retrieve crop biomass information over a diversity of crops, ranging from wheat to sugar beet and corn or sunflower. Results indicate errors ranging from 0.6 kg/m2 for cereals to 2 kg/m2 for broad leaf crops (i.e. maize and sugar beet). It is worth mentioning that the exploited relationships have not been fully validated due to the limited data sets available and that, at least, the SAR-biomass relationships for the main crops, such as wheat, maize, sunflower, rapeseed and soybean, should be further consolidated. For the soil moisture (SSM) application, the SMOSAR SSM retrieval algorithm (Mattia et al., 2011) has been applied to time series of C-, L- and X-band SAR data in order to assess the extension to which MF SAR data can be combined to constitute a composite SSM time series with improved temporal resolution. Results indicate that MF SAR data can lead to an improvement of the temporal resolution of high resolution SAR SSM products. Either the combined use of independent single-frequency retrieval or the integrated use of multi-frequency SAR data for SSM retrieval have shown promising results that, however, need to be further assessed. One issue has been the lack of sufficiently long and consistent time series of MF data needed to fully validate the proposed approaches.