Geoscience Reference
In-Depth Information
Weather Service (Deutscher Wetterdienst) in 2000 (Hess
2001
). Along the same line,
ECMWF developed a point-scale simplified EKF soil moisture analysis (Seuffert et al.
2004
). Preliminary investigations at local scale showed that the OI and the EKF soil
moisture analyses give similar results when they both use screen-level parameters (Seuffert
et al.
2004
). The ECMWF simplified EKF soil moisture analysis was extended to be used
at global scale (Drusch et al.
2009
) and implemented in operations in 2010 (de Rosnay
et al.
2012
).
In the following section, differences between EKF and OI soil moisture analyses are
presented in terms of soil moisture increments and low-level atmospheric parameters
forecasts.
3.2 Comparison Between the OI and EKF Soil Moisture Analyses
de Rosnay et al. (
2012
) quantified monthly mean global soil moisture increments for both
the OI and the EKF schemes for an entire annual cycle. They showed that the OI scheme
systematically adds water to the soil. The global monthly mean value of the OI analysis
increments was shown to be 5.5 mm, which represents a substantial and unrealistic con-
tribution to the global water cycle. In contrast, the EKF global mean soil moisture analysis
increments are much smaller, representing more reasonable global monthly mean incre-
ments of 0.5 mm. The reduction in increments between the EKF and the OI is mainly due
to a systematic reduction in increments below the first layer. The OI increments computed
for the first layer are amplified for deeper layers in proportion to the layer thickness,
explaining the overestimation of the OI increments. In contrast, the EKF dynamical esti-
mates, based on perturbed simulations, allow optimising soil moisture increments at
different depths to match screen-level observations according to the strength of the local
and current soil-vegetation-atmosphere coupling. The EKF accounts for additional con-
trols due to meteorological forcing and soil moisture conditions. Thereby, it prevents
undesirable and excessive soil moisture corrections (de Rosnay et al.
2012
). Figure
3
illustrates monthly mean increments accumulated in the first metre of soil, for August
2009. In agreement with de Rosnay et al. (
2012
), it shows that larger increments are
accumulated into the soil with the OI than with the EKF analysis scheme. It is, however,
interesting to note that large increments remain with the EKF in the US Great Plains and in
South America, showing these regions are affected by systematic soil moisture bias in the
LSM. Further investigation will be carried out in the future to address this feature.
The impact of the soil moisture analysis scheme on analysed soil moisture was also
studied using ground data from SMOSMANIA (Soil Moisture Observing System-Meteo-
rological Automatic Network Integrated Application) (Calvet et al.
2007
). de Rosnay et al.
(
2012
) showed that ECMWF soil moisture is generally in good agreement with ground
observations, with mean correlations higher than 0.78. Using the EKF instead of the OI
scheme improves significantly the soil moisture analysis, with mean correlation between
ECMWF and ground truth soil moisture higher than 0.84 when the EKF soil moisture
analysis is used (de Rosnay et al.
2012
).
Figure
4
shows the monthly mean impact of the EKF soil moisture analysis on the 36-h
forecast of 2-m temperature at 0000 UTC for July-August-September 2009. It shows the
difference in temperature error (in K) between the OI and EKF experiments. Positive
values indicate that the EKF generally improves the 2-m temperature forecasts compared
to the OI soil moisture analysis. It is consistent with the results shown by de Rosnay et al.
(
2012
) to indicate that in most areas the 2-m temperature errors for OI are larger than the
