Existing forward snow emission models (SEMs) are limited by knowledge of both the temporal and spatial variability of snow microphysical parameters, with grain size being the most difficult to measure or estimate. This is due to the sparseness of in situ data and the lack of simple operational parameterizations for the evolution of snowpack properties. This paper compares snow brightness temperatures predicted by three SEMs using, as inputs, predicted snowpack characteristics from the Variable Infiltration Capacity (VIC) model. The latter is augmented by a new parameterization for the evolution of snow grain morphology and density. The grain size dynamics are described using a crystal growth equation. The three SEMs used in the study are the Land Surface Microwave Emission Model (LSMEM), the Dense Media Radiative Transfer (DMRT) model, and the Microwave Emission Model of Layered Snowpacks (MEMLS). Estimated brightness temperature is validated against the satellite [Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E)] data at two sites from the Cold Land Processes Experiment (CLPX), conducted in Colorado in the winter of 2003. In addition, a merged multimodel estimate, based on Bayesian model averaging, is developed and compared to the measured brightness temperatures. The advantages of the Bayesian approach include the increase in the mean prediction accuracy as well as providing a nonparametric estimate of the error distributions for the brightness temperature estimates.
All Science Journal Classification (ASJC) codes
- Atmospheric Science