Application of a Bayesian SWE retrieval algorithm using X- and Ku- band radar measurements

When the snow radar is applied for the snow water equivalent retrieval, an advanced algorithm is required to separate the influence of the underlying soil, and taking the penetration depth and the stratigraphy of the natural snowpit into consideration. In this study, the Bayesian-based Markov Chain Monte Carlo method is applied to estimate SWE based on backscattering coefficient measurements at X- and Ku-bands, VV polarization for taiga snowpits at Sodankyla (Lemmentyinen et al., 2013). This algorithm samples the SWE as well as the snow and soil properties that can reproduce the radar measurements from a set of globally-available prior distributions of these parameters. The prior is from monthly-average land surface model SWE product and traditional snow classes. The active Microwave Emission Model of Layered Snowpacks (MEMLSa) is used as the observation model. In the retrieval system, the multiple-layer snow density and exponential correlation length will be estimated. Therefore, we will check how many layers are required to have a good estimate of SWE. The sensitivity of soil parameters for SWE retrieval and the possibility of applying multi-angle observations to separate the influence of soil roughness and moisture will be studied. Preliminary results show that, this algorithm is capable of predicting SWE with a root mean squared error of 35 mm.