Geoscience Reference
In-Depth Information
Fig. 2 SWE and SCF fields for 6 days (MMDDYYYY) in the winter of 2009-2010 for a 75 km by 100 km
domain (1 km resolution) in northern Colorado. Blue (white) colors indicate low (high) SWE or SCF, black
shading indicates no snow, and orange shading indicates no data. The top two rows show SWE and SCF
satellite observations. The remaining rows show SWE (rows 3 and 4) and SCF (rows 5 and 6) for the
ensemble Open Loop (EnsOL) forecast (no assimilation) and the analyses obtained through data assimilation
(DA) of SWE or SCF. Adapted from De Lannoy et al. (
2012
)
timing of the onset of the snow season, albeit without a net improvement of SWE esti-
mates. In areas of deep snow, however, AMSR-E retrievals are typically biased low and
require bias correction (or scaling of the observations) prior to data assimilation. De
Lannoy et al. (
2012
) also showed that the interannual SWE variations could not be
improved through the assimilation of AMSR-E because the AMSR-E retrievals lack
realistic interannual variability in deep snow packs. These deficiencies in the AMSR-E
SWE retrievals motivated the development of the empirical microwave radiative transfer
model (Sect.
3.3.2
) toward a radiance-based snow analysis.
Of course, horizontal downscaling is not only important for snow assimilation. Low-
frequency passive microwave brightness temperature observations such those from AMSR-
E and SMOS (and the corresponding soil moisture retrievals) are at the coarse resolution of
*50 km. But for applications such as weather prediction, soil moisture estimates are
