Geoscience Reference
In-Depth Information
groundwater component is critically important for the assimilation of TWS retrievals
(Sect.
3.2
).
The Catchment model also includes a state-of-the-art, multi-layer, global snow model
(Stieglitz et al.
2001
). In each watershed, the evolution of the amount of water in the snow
pack (or snow water equivalent; SWE), the snow depth, and the snow heat content in
response to surface meteorological conditions and snow compaction is modeled using three
layers. The soil, vegetation, and snow model parameters used in the Catchment model are
from the NASA GEOS-5 global modeling system (Rienecker et al.
2008
).
The EnKF is a Monte-Carlo variant of the Kalman filter, which sequentially updates
model forecasts in response to observations based on the relative uncertainty of the model
and the observations. The key idea behind the EnKF is that the relevant parts of the model
error covariance structure can be captured by a small ensemble of model trajectories. Each
member of the ensemble experiences perturbed instances of the observed forcing fields
(representing errors in the forcing data) and/or randomly generated noise that is added to
the model parameters and prognostic variables (representing errors in model physics and
parameters). The model error covariance matrices that are required for the filter update can
then be diagnosed from the ensemble at the update time. The EnKF is flexible in its
treatment of errors in model dynamics and parameters. It is also very suitable for modestly
nonlinear problems and has become a popular choice for land data assimilation (Andreadis
and Lettenmaier
2006
; Durand and Margulis
2008
; Kumar et al.
2008a
,
b
; Pan and Wood
2006
; Reichle et al.
2002a
,
b
; Su et al.
2008
; Zhou et al.
2006
).
To realize the potential benefits from data assimilation, the assimilation system must be
supplied with appropriate input parameters for the description of model and observation
errors. For an ensemble-based system such as the GEOS-5 LDAS, for example, standard
deviations, spatial and temporal correlations, and cross-correlations must be specified for
the perturbations that are applied to each ensemble member. A detailed discussion of the
error parameters in the examples discussed here is beyond the scope of the paper. The
reader is referred to the references provided with each example as well as the overview
discussion of Reichle et al. (
2009
).
2.2 Assimilated Observations
The data assimilation examples discussed in this paper use various types of satellite
observations from a number of polar orbiting sensors/platforms, including passive and
active microwave observations (AMSR-E, SMOS, and ASCAT), visible and near-infrared
observations (MODIS), and gravimetric observations (GRACE).
AMSR-E, which operated with nominal performance between 2002 and 2011, is a
scanning, dual polarization radiometer that measured microwave emission from the Earth
at six frequencies (6.9, 10.7, 18.7, 23.9, 36.5, and 89.0 GHz), ranging in resolution from
*50 km at 6.9 GHz to *5 km at 89.0 GHz (Knowles et al.
2006
). Its successor, AMSR2,
was launched in May 2012 (
http://www.jaxa.jp/projects/sat/gcom_w/index_e.html
). The
training and validation of the empirical microwave radiative transfer model for snow-
covered land surfaces in Sect.
3.3.2
uses the 10.7, 18.7, and 36.5 GHz AMSR-E brightness
temperatures, while the snow assimilation example in Sect.
3.1
uses SWE retrievals that
are based on the difference between the 18.7 and the 36.5 GHz brightness temperatures
(Kelly
2009
). The soil moisture assimilation examples in Sect.
3.4
use surface (top 1 cm)
soil moisture retrievals that are derived from the 6.9 and 10.7 GHz brightness temperatures
(de Jeu et al.
2008
; Njoku et al.
2003
).
