Geoscience Reference
In-Depth Information
1 Introduction
Snow is extremely important hydrologically, with more than one-sixth of the global
population situated in areas where snow precipitation is greater than half of annual runoff
(Barnett et al.
2005
). Snow affects both timing and quantity of runoff as well as the surface
energy balance and atmospheric chemistry.
Currently, remote sensing products exist for snow-covered area (SCA), albedo, grain
size, surface contaminants, melt and snow water equivalent (SWE). Measurements of snow
surface properties such as SCA are regularly used (e.g., Brown and Mote
2009
; Dye
2002
;
Frei et al.
2003
) and generally have more well-characterised uncertainties (Hall and Riggs
2007
; Rittger et al.
2013
) than estimates of bulk properties such as SWE.
Measurement of surface properties has allowed the identification of snow season
duration, surface melt (Koskinen et al.
1997
) and a determination of snow's contribution to
radiative feedback in response to warming (Flanner et al.
2011
). In terms of hydrological
relevance, Painter et al. (
2012
) developed a Moderate Resolution Imaging Spectroradi-
ometer (MODIS) algorithm for determining radiative forcing from impurities in near-
surface snow. When realistic values of these radiative forcings were included in the
analysis of snow in south-western Colorado, it was estimated that the impurities reduced
snow cover duration by 21-51 days, increased peak outflow, changed the runoff profile and
reduced total seasonal runoff (Skiles et al.
2012
).
Despite the successes of surface measurements, there remain large uncertainties in
global estimates of SWE, with regional disagreements between products derived from
remote sensing, general circulation models (GCMs) and reanalyses.
This study reviews continental-scale SWE products and describes the key techniques
and their relative strengths, including the assimilation of remotely sensed passive micro-
wave (PM) observations. A recent product that assimilates PM, Globsnow (Takala et al.
2011
), is described in detail as it has been suggested as a suitable product for validation of
land surface models (LSMs (Hancock et al.
2013
)).
The assimilation of PM observations requires an observation operator, which converts
the state vector of snow properties into a vector of observable microwave brightness
temperatures. In the case of Globsnow, the snow is described by density, grain size and
snow depth of a single layer. The observation operator is the Helsinki University of
Technology (HUT) radiative transfer model (Pulliainen et al.
1999
), which produces a
brightness temperature difference between two PM channels, DT
B
, for comparison with
satellite retrievals.
Although Globsnow assumes a single homogeneous layer, snowpacks typically consist
of multiple layers that often feature complex stratigraphy which affects the radiative
transfer. The current Globsnow approach neglects this both in the radiative transfer sim-
ulation and in calculating the weighting function that determines the size of the PM-driven
update to the forecast.
Globsnow's performance might be improved by the relaxation of the one-layer
assumption, and here the effect of this relaxation on simulated DT
B
s is assessed based on
realistic snow profiles obtained from the snowpits of the National Aeronautics and Space
Administration Cold Land Processes Experiment (NASA CLPX).
Section
2
reviews the historical methods of snow mass estimation, including separate
estimates from snow models, ground stations and PM. Section
3
introduces the principles
behind the assimilation of passive microwaves and details Globsnow, identifying its
simplified snow stratigraphy as a possible source of error and suggesting that layering
might be included in a future scheme.
