Abstract
Detailed and frequent mapping of the snow water equivalent (SWE) is required by several applications including hydrology, water management and climate changes. The capabilities offered by remote sensing techniques in estimating SWE at high spatial resolution and with frequent revisiting are therefore extremely attractive.Besides the well assessed capability of microwave radiometers, Synthetic Aperture Radar (SAR) sensors have also revealed some potential for operational SWE monitoring. Although the X-band (i.e. the highest frequency currently available on satellite SAR systems) is not the most suitable frequency for the retrieval of snow parameters, especially for shallow snow covers, past research on dry snow demonstrated that SWE values greater than 100-150 mm can be retrieved at X-band too, by using appropriate algorithms and models.Moreover, the new generation of X-band SAR satellites, offered significant improvements in terms of revisiting time and spatial resolution, which are crucial features for operational monitoring applications.This is particularly true for the Italian Space Agency (ASI)’s COSMO-SkyMed (CSK) mission that, thanks to its 4-satellite constellation, allows frequent revisiting of the target areas.This work aims at exploiting the information derived from CSK images acquired in StripMap HIMAGE mode at 3 m ground resolution, together with in-situ measurements for SWE retrieval in the entire province of South Tyrol (~7,400 km²),in north-eastern Italy.The SWE data considered as ground truth have been derived from manual snow profiles performed during the field campaigns by the operators and provided by the Hydrographic Office of the Autonomous Province of Bolzano. From the available dataset, 45 SWE values have been selected in correspondence to the CSK acquisitions in dry snow conditions. Data refer to the winter months of January and February of three years, from 2013 to 2015. The CSK X-band backscattering coefficients were compared with the in-situ SWE data to evaluate the sensitivity of microwave measurements to the target parameter. For interpreting the scattering mechanisms and assessing the obtained relationships, the experimental data have been also compared with model simulations based on the Dense Media Radiative Transfer theory in the Quasi-Crystalline Approximation (QCA) of Mie scattering of densely packed Sticky spheres (DMRT –QMS).Based on the results of the sensitivity analysis, the SWE retrieval has been attempted by using two machine learning techniques, namely Support Vector Regression (SVR) and Artificial Neural Networks.Training of both algorithms was based on experimental data and DMRT model simulations, by considering backscattering and incidence angle as inputs of the algorithm and SWE as output.After training, the algorithms were applied to the available CSK images to generate SWE maps of the entire area covered by the SAR acquisitions. The obtained results were encouraging, although more analysis and validation are needed to exploit the potential and assess the limits of CSK application to snow parameters retrieval.This work is carried out by EURAC, CNR/IFAC and ASI in the framework of the 2019-2021 project ‘Development of algorithms for estimation and monitoring of hydrological parameters from satellite and drone’, funded by ASI under grant agreement n.2018-37-HH.0