Geoscience Reference
In-Depth Information
available from the SYNOP network for NWP soil moisture analysis systems. As shown by
Mahfouf (
1991
), it provides indirect, but relevant information to analyse soil moisture. So,
most of the current operational soil moisture analysis systems rely on analysed screen-level
variables (2-m temperature and relative humidity).
In 1994, a nudging approach was implemented at ECMWF to analyse soil moisture,
using the lowest atmospheric level specific humidity analysis increments. It was the first
NWP centre to implement a soil moisture analysis scheme, mainly to prevent the soil
moisture from drifting to unrealistic dry conditions in summer time. However, the nudging
scheme was not accounting for processes that modulate the relation between soil moisture
and specific humidity. So, soil moisture increments were affected by systematic biases with
successive negative and positives increments at both the diurnal and seasonal scales
(Mahfouf et al.
2000
).
In 1999, the nudging soil moisture analysis was replaced by a 1D Optimal Interpolation
analysis, as originally proposed by Mahfouf (
1991
). The OI soil moisture analysis
implementation at ECMWF and evaluation are detailed in Douville et al. (
2000
) and
Mahfouf et al. (
2000
). The OI soil moisture analysis relies on the relation between soil
moisture and screen-level temperature and relative humidity. When soil moisture is
underestimated, air temperature is expected to be overestimated and air humidity under-
estimated. In contrast, when soil moisture is overestimated, screen-level air temperature is
too low and air humidity too large. Based on this relation between soil moisture and
screen-level parameters, the soil moisture correction is computed as a function of the
screen-level parameters correction. So, a dedicated screen-level parameters analysis was
implemented, based on an OI approach, and its increments are then used as input of the soil
moisture analysis (Mahfouf et al.
2000
). The same method has been used to analyse soil
temperature and snow temperature using the screen-level air temperature increments. The
OI soil moisture and temperature analysis has been widely used for NWP applications in
several NWP centres. It was used at ECMWF for operational NWP from July 1999 to
November 2010. It has been used in ECMWF re-analyses ERA-40 (Uppala et al.
2005
) and
is still in use for ERA-Interim (Dee et al.
2011
). An OI soil moisture analysis is currently
used at M
´
t
´
o France (Giard and Bazile
2000
), Environment Canada (B
´
lair et al.
2003
)
and in the High Resolution Limited Area Model (HIRLAM) (Rodr´guez et al.
2003
).
As shown by Drusch et al. (
2009
), the OI soil moisture analysis improves screen-level
parameters forecasts, without, however, any positive impact on soil moisture. Furthermore,
an important weakness of the OI approach is its lack of flexibility to easily account for new
types of data including new generations of satellite data (Mahfouf et al.
2009
). Also, it uses
calibrated coefficients that would require to be updated for each change in the LSM. The
OI, by using fixed coefficients, does not account for local processes such as cloud cover or
soil moisture conditions that influence the coupling strength between soil moisture and
screen-level parameters and therefore the magnitude of the increments (de Rosnay et al
2012
).
Several NWP centres started to investigate the use of satellite data to analyse soil
moisture, using a range of approaches based on simplified EKF (Draper et al.
2011
) or the
equivalent simplified 2D-Var (Balsamo et al.
2007
), as well as EKF and an Ensemble
Kalman Filter (Reichle et al.
2002
,
2008
). At M´t´o France, an offline EKF approach was
evaluated, and the impact of ASCAT (Advanced SCATterometer) soil moisture data
assimilation on the low-level atmospheric parameters was addressed (Mahfouf
2010
). A
limited impact of the EKF soil moisture analysis on relative humidity and air temperature
was found. A simplified Extended Kalman Filter (EKF) soil moisture analysis, using
screen-level parameters information, was implemented operationally at
the German
