Investigating the use of Lambertian reflection in the assimilation of surface-sensitive microwave sounding radiances over snow and sea-ice

This paper investigates two aspects aimed at an increased use of surface-sensitive microwave sounding  radiances for NWP, namely the assimilation of humidity-sounding radiances from ATMS over snow, as well  as the use of Lambertian reflection in the radiative transfer modelling over snow and sea-ice surfaces. The  Lambertian characteristics are modelled through a parametrization of the effective zenith  angle used to calculate the down-welling radiation in the radiative transfer model.

Using Lambertian rather than specular reflection significantly improves the forward modelling for 183  GHz humidity-sounding channels over snow and sea-ice in comparisons to observations, including a  reduction of significant zenith-angle dependent biases. For temperature-sounding channels in the 50 GHz  band, comparisons between observations and background equivalents also suggest improvements from  treating snow and sea-ice surfaces as Lambertian or semi-Lambertian, but the characteristics appear more variable, and there are indications of other sources of error such as biases in the skin-temperature. For both spectral regions the improvement from using non-specular reflection is clearest for snow-covered land surfaces.

Assimilation experiments show that the addition of 183 GHz humidity-sounding channels from ATMS over snow-covered land lead to a slight positive impact over the higher latitudes in winter. On top of this,  replacing the specular surface assumption over snow and sea-ice with a fully Lambertian for 183 GHz channels on ATMS and semi-Lambertian for 50 Ghz channels on ATMS and AMSU-A means that more observations pass quality control, giving a neutral to slightly positive impact at higher latitudes in winter. Combining the two changes leads to small forecast benefits that are statistically significant up to day 3 at higher latitudes.