Geoscience Reference
In-Depth Information
Fig. 12 Mean skill for root zone soil moisture from the open loop (ensemble mean, no assimilation), and
the data assimilation (DA) of ASCAT, AMSR-E, and both surface soil moisture retrievals, averaged by land
cover class, with 95 % confidence intervals. The number of sites in each land cover class is given in the axis
labels. Skill is defined as the daily anomaly R value versus SCAN/SNOTEL and Murrumbidgee in situ
observations. Adapted from Draper et al. (
2012
)
(R) with in situ soil moisture observations from the SCAN/SNOTEL network in the US (66
sites) and the Murrumbidgee Soil Moisture Monitoring Network in Australia (19 sites).
These 85 sites are surrounded by terrain with low topographic complexity based on data
provided with the ASCAT observations. Averaged over these sites, the ASCAT and
AMSR-E surface soil moisture retrievals have similar skill (Draper et al.
2012
).
Figure
12
shows the estimated R values and their 95 % confidence intervals for root
zone soil moisture from the assimilation of ASCAT, AMSR-E, and both. The results are
benchmarked against an open loop (no assimilation) model integration and have been
averaged by land cover type (based on MODIS land cover classifications). Averaged across
all 85 sites, assimilating ASCAT and/or AMSR-E surface soil moisture retrievals signif-
icantly improved the root zone soil moisture skill (at the 5 % level). The mean skill was
increased from 0.45 for the open loop, to 0.55 for the assimilation of ASCAT, 0.54 for the
assimilation of AMSR-E, and 0.56 for the assimilation of both.
Assimilating the ASCAT or AMSR-E retrievals also improved the mean R value over
each individual land cover type, in most cases significantly. At the frequencies observed by
AMSR-E and ASCAT, dense vegetation limits the accuracy of soil moisture observations,
and so the improvements obtained over the mixed cover sites, which have 10-60 % trees or
wooded vegetation, are very encouraging. For each land cover type, the skill obtained from
the assimilation of ASCAT or AMSR-E retrievals was very similar. The combined
assimilation of ASCAT and AMSR-E retrievals generally matched or slightly exceeded the
mean R values from the single-sensor assimilation experiments.
Draper et al. (
2012
) also examined the contribution of the model skill and the obser-
vation skill to the skill of the assimilation estimates. The color surface in Fig.
13
shows the
skill improvements (DR) in root zone soil moisture, where DR is defined as the skill (R)of
the assimilation estimates (from the single-sensor assimilation of ASCAT or AMSR-E
retrievals) minus that of the open loop model estimates. The skill improvements are shown
as a function of the open loop model skill and the retrieval skill. Specifically, the ordinate
measures the skill of the open loop root zone soil moisture estimates, and the abscissa
measures the skill of the assimilated (ASCAT or AMSR-E) surface soil moisture retrievals.
Where the skill of the assimilated retrievals is no more than 0.2 less than the open loop skill
(below the dashed line), the assimilation improves the root zone soil moisture skill. The
improvements increase (up to 0.4) as the observation skill increases relative to that of the
