Roughness and vegetation parameterizations at L-band for soil moisture retrievals over a vineyard field

The capability of L-band radiometry to monitor surface soil moisture (SM) at global scale has been analyzed in numerous studies, mostly in the framework of the ESA SMOS and NASA SMAP missions. To retrieve SM from L-band radiometric observations, two significant effects have to be accounted for, namely soil roughness and vegetation optical depth. In this study, soil roughness effects on retrieved SM values were evaluated using brightness temperatures acquired by the L-band ELBARA-II radiometer, over a vineyard field at the Valencia Anchor Station (VAS) site during the year 2013. Different combinations of the values of the model parameters used to account for soil roughness effects (HR, QR, NRH and NRV) in the L-MEB model were evaluated. The L-MEB model (L-band Microwave Emission of the Biosphere) is the forward radiative transfer model used in the SMOS soil moisture retrieval algorithm. In this model, HR parameterizes the intensity of roughness effects, QR accounts for polarization effects, and NRH and NRV parameterize the variations of the soil reflectivity as a function of the observation angle, θ, respectively for both H (Horizontal) and V (Vertical) polarizations. These evaluations were made by comparing in-situ measurements of SM (used here as a reference) against SM retrievals derived from tower-based ELBARA-II brightness temperatures mentioned above. The general retrieval approach consists of the inversion of L-MEB. Two specific configurations were tested: the classical 2-Parameter (2-P) retrieval configuration where SM and τNAD (vegetation optical depth at nadir) are retrieved, and a 3-Parameter (3-P) configuration, accounting for the additional effects of the vineyard vegetation structure.Using the 2-P configuration, it was found that setting NRp (p=H or V) equals to −1 provided the best SM estimations in terms of correlation and unbiased Root Mean Square Error (ubRMSE). The assumption NRV=NRH=−1 simplifies the L-MEB retrieval, since the two parameters τNAD and HR can then be grouped and retrieved as a single parameter (method here defined as the Simplified Retrieval Method (SRP)). The main advantage of the SRP method is that it is not necessary to calibrate HR before performing the SM retrievals. Using the 3-P configuration, the results improved, with respect to SM retrievals, in terms of correlation and ubRMSE, as the structural characteristics of the vineyards were better accounted for. However, this method still requires the calibration of HR, a disadvantage for operational applications. Finally, it was found that the use of in-situ roughness measurements to calibrate the roughness model parameters did not provide significant improvements in the SM retrievals as compared to the SRP method.