Geoscience Reference
In-Depth Information
Table 4 continued
Observation
State
Parameter/model
EnKF/EnKS
Other
Surface soil or skin
temperature,
evapotranspiration,
retrievals or in situ
Pipunic et al. (
2008
),
Ghent et al. (
2010
),
Reichle et al. (
2010
),
Xu et al. (
2011
)
Castelli et al. (
1999
),
Lakshmi (
2000
),
Boni et al. (
2001
),
Schuurmans et al.
(
2003
), Bosilovich et al.
(
2007
), Renzullo et al.
(
2008
), Sini et al.
(
2008
), Meng et al.
(
2009
), Barrett and
Renzullo (
2009
),
Mackaro et al. (
2011
)
Caparrini et al. (
2004
),
Kalma et al. (
2008
),
Gutmann and Small
(
2010
)
Water stage,
retrievals
Andreadis et al. (
2007
),
Durand et al. (
2008
),
Biancamaria et al.
(
2010
)
Matgen et al. (
2010
),
Giustarini et al. (
2011
)
Montanari et al. (
2009
)
Terrestrial water
storage, retrievals
Zaitchik et al. (
2008
), Su
et al. (
2010
), Li et al.
(
2012
), Forman et al.
(
2012b
)
-
G¨ nter (
2008
), Lo et al.
(
2010
)
Discharge, gauge
Weerts and El Serafy
(
2006
), Vrugt et al.
(
2006
), Pauwels and De
Lannoy (
2006
,
2009
)
Aubert et al. (
2003
),
Moradkhani et al.
(
2005a
), Seo et al.
(
2009
), Lee et al.
(
2011
), Vrugt et al.
(
2012
)
Madsen (
2003
),
Moradkhani et al.
(
2005b
), Montanari and
Toth (
2007
), Vrugt
et al. (
2008
), Quets
et al. (
2010
)
Leaf area index,
remotely sensed
Pauwels et al. (
2006
),
Nearing et al. (
2012
)
Jarlan et al. (
2008
),
Albergel et al. (
2010
),
R¨diger et al. (
2010
)
Lewis et al. (
2012
)
Screen-level
observations
-
Balsamo et al. (
2004
),
Seuffert et al. (
2004
),
Drusch and Viterbo
(
2007
), Mahfouf et al.
(
2009
), Draper et al.
(
2011
), Mahfouf and
Bliznak (
2011
)
-
Synthetic observation studies are classified by the observation type that is mirrored. For land surface
(-coupled) state updating, the studies are divided into sets using either the EnKF or EnKS (Ensemble
Kalman Smoother) and those using any other assimilation technique. For parameter and model structure
updating, examples relate to either forward models or land surface(-coupled) models
error covariance matrix, is typically assumed to be diagonal, although this is not always
justified. R includes errors of the measurements themselves, E, and errors of representa-
tiveness, F; R = E ? F. B is the background error covariance matrix in variational
methods (the analogue in the KF and ensemble methods is P
f
); its diagonal elements
determine the relative weight of the forecasts, and its off-diagonal elements determine how
information is spread spatially. Estimating B or P
f
is a key part of the data assimilation
method (Bannister
2008a
,
b
). Estimating model error Q is a research topic.
In the EnKF, the background (or forecast) errors are represented by the spread of the
ensemble. This simplifies the computation of P
f
, implicitly accounts for the model error
Q and avoids the calculation of Eq. (
2b
). For land data assimilation, the relative fraction of
