Coupling SAR X-band and optical data for NDVI retrieval: model calibration and validation on two test areas

Sustainability of modern agro-hydrology requires the knowledge of spatial and temporal variability of vegetation biomass to optimize management of land and water resources. Diversely from optical imaging, temporal resolution of active sensors, such as SAR, is not limited by sky cloudiness; thus, they may be combined with optical imageries to provide a more continuous monitoring of land surfaces. Several new SAR missions (e.g., ALOS-PALSAR, COSMO-SkyMed 1 and 2, TerraSAR-X, TerraSAR-X2, Sentinel 1) acquiring at X-, C- and L-bands and dual polarization capability, are characterized by a short revisit time (from 12 h to ~10 days) and high spatial resolution (<20 m). These satellites could provide suitable data for operative crop monitoring. The new COSMO-SkyMed (owned by Italian Space Agency-ASI) represents a valuable source of SAR data in X-band, opening new opportunities to develop agro-hydrological applications. Although, X-band backscattering is not the most suitable to model agricultural and hydrological processes; it could help assessing the vegetation development if combined with optical vegetation indices (VIs). Recently, two models to infer a VI from σ° have been setup. One of these achieved accurate results in VI retrieval at a certain time using σ°(VISAR), once known the NDVI derived from optical images (VIopt) at a reference time. This paper aims to further validate this model through two independent datasets. The model was firstly implemented on COSMO-SkyMed and optical DEIMOS-1 data acquired over the Sele plain (Campania, Italy). It has been further validated using different optical images. To this aim, a dataset of 2 COSMOSkyMed images and 2 Landsat 7 SLC-off images were acquired in the southwestern part of Sicily (Italy) between 8 and 25 August 2011. Determination coefficients of the validation set were similar to those of the calibration set. Results confirm that VISAR obtained using the combined model is a suitable surrogate of VIopt if estimated at parcel scale.