Retrieval of soil relative permittivity using passive microwave brightness temperatures

Accurate determination of the relative permittivity of frozen soil is required in order to account for emission of thermal microwave radiation from soil when producing estimates of snow depth from satellite passive microwave measurements. On a global scale, observations of the spatial variability of the relative permittivity of the underlying frozen soil are very limited.

A new retrieval method aimed at estimating the relative permittivity of soil was developed by forcing a radiative transfer model, DMRT-ML, with a snow microstructure evolution framework from the JIM snow evolution model and an ensemble of soil dielectric configurations. The ensemble of simulated brightness temperatures (TBs) were then compared to TB observations collected from an area of intensive observation in Sodankylä, Finland.

A radiometer configuration consisting of a 10 GHz H-pol receiver retrieved soil permittivity to within 10% of the observed value for all incidence angles between 30 - 60 degrees. The use of a 10 GHz V-pol configuration produced less accurate and less precise retrievals of soil relative permittivity. Accurate retrieval of soil relative permittivity is only weakly dependent on assumed snow density. Relative permittivity values obtained using this procedure were then input into DMRT- ML to estimate snow depth from TBs differenced at 19 GHz and 37 GHz. Snow depth errors as a proportion of total snow depth were largest (up to 45%) during December for all soil dielectric parameterisations considered; reducing to <13% from January to April.

This demonstrates the importance of accurately retrieving permittivity at the onset of the snow season, immediately after ground freezing and before the first significant snowfall. When this is not taken into account, errors in estimation of snow depth from passive microwave sensors persist through peak annual snow depth the following spring.