Geoscience Reference
In-Depth Information
Section
4
describes the experiments, which attempt to assess the effect of realistic snow
layering on radiative transfer. Coincident satellite PM measurements are compared with
HUT simulations at CLPX, confirming that HUT simulations are close to observed values.
The effect of including or neglecting detailed snow layering is assessed by comparing DT
B
simulations when snowpits are resampled to different layering profiles. Profiles include
between 1 and 5 layers, along with an N-layer case where layers are prescribed based on
the snowpit measurement resolution. The N-layer case is taken as truth, and the difference
in simulated DT
B
s for fewer-layer models relative to this truth allows statistical estimation
of the bias and variance introduced through simplification of stratigraphy to fewer layers,
which are reported as a function of snow depth and number of layers.
The results are related in Sect.
5
and discussed in Sect.
6
where it is indicated that
neglect of statigraphy may mean that Globsnow has unaccounted variance in its assimi-
lation step. The results have relevance to a user who may use these to calculate variance
introduced due to simplified stratigraphy, or alternatively may choose an optimal layering
structure based on the criteria of computational expense and acceptable levels of variance.
However, it is cautioned that these results are only derived for snow typical of that present
during CLPX.
2 Current Snow Mass Estimation
2.1 General Circulation Models and Reanalyses
Without the global coverage of space-based remote sensing, alternative methods of snow
mass estimation have relied on a combination of models and observations. Coupled GCMs
are a modelling approach and have been used to estimate SWE climatologies for current
conditions, and the spatial and temporal components of these climatologies have been
explored by Clifford (
2010
) and Roesch (
2006
) among others.
However, due to the chaotic nature of the system, fully coupled models are only capable
of estimating climatology and, in order to produce a time series corresponding to the real-
world realisation of weather, regular assimilation of observational data is required.
As such a number of reanalysis products have been produced, coupling LSMs which
simulate the snow cover with an atmospheric model. These reanalyses regularly assimilate
observations of both the atmosphere and the land surface, although no fully coupled land-
atmosphere reanalysis yet assimilates microwave radiances for the purpose of snow mass
estimation. Instead, in situ synoptic station measurements of snow depth and estimates of
SCA based on satellite data are used.
The full details of these reanalyses and their assimilation schemes are beyond the scope
of this paper; the reader is directed to the references in Table
1
, which details selected
reanalyses and other gridded products which offer snow mass or snow depth.
A number of assessments of reanalysis performance in terms of snow variables have
been undertaken. Khan and Holko (
2009
) noted that reanalyses performed well in much of
the Aral Sea Basin, although there were underestimates of snow depth and SWE in
mountainous areas. Betts et al. (
2009
) determined that both the European Centre for
Medium Range Weather Forecasts (ECMWF) 40 year and Interim Reanalyses (ERA-40,
ERA-Interim) suffer from early snow melt out. Meanwhile, Clifford (
2010
) reported the
spatial and temporal characteristics of different approaches to snow mass estimation in
more detail, and that the potential for future improvements remains clear. Improved
modelling
is
one
opportunity,
with
Salzmann
and
Mearns
(
2012
)
comparing
SWE
