Dry snow cover is conventionally though to have only a minimal influence on microwave emission at L-band (~1-3 GHz), due to the inherent high penetration depth of microwave energy in dry snow at this frequency range. However, recent studies have indicated a non-negligible influence of dry snow cover on L-band emission as detected by means of microwave remote sensing, due to the dual effects of refraction and impedance matching at the soil-snow interface (Schwank et al., 2014; Schwank et al., 2015; Lemmetyinen et al., 2016; Naderpour et al., 2016; Roy et al., 2016), corroborated by experimental data.

Besides influencing retrievals of soil properties in the presence of dry snow, these effects also introduce the potential of retrieving characteristics of snow cover (namely, snow density) in dry snow conditions using microwave remote sensing at L-band.

The theme was covered within two recent ESA studies, the “Combined use of multifrequency radiometry (L- to Ka-Band) and SAR imagery for enhanced monitoring of terrestrial cryosphere processes” and the “SMOS Frost2Study”. In these studies, a simplified emission model for addressing the influence of dry snow cover on L-band emission was developed (Schwank et al., 2015), and extended to cover also more complex satellites scenes containing a mixture of forested and open areas as well as water bodies (e.g. lakes). A two parameter retrieval method proposed by Schwank et al. (2015) was applied to retrieve simultaneously dry snow density and ground permittivity from L-band observations, applying the forward model in a numerical inversion scheme. Retrievals using SMOS L3 brightness temperatures were performed. The retrieved soil permittivities show distinct spatial features indicating the stage of soil freezing, low values indicating frozen soil. While showing overall a realistic mean level, the retrieved snow densities exhibit high spatial scatter and artefacts inherent of SMOS data, pointing to further development needs of the retrieval algorithm. Proper validation is a challenge as large scale reference data is difficult to obtain. Further development is focusing on using selected test sites and e.g. physical snow models to provide reference data and support algorithm development.