Geoscience Reference
In-Depth Information
GPS
TPWater
Radiance
TerrPress
Sfcwind
SpecHum
v−wind
u−wind
Temp
−0.05
0
0.05
0.1
−1
0
1
2
3
4
x 10
−6
fcst s
i
−sensitivity ( J Kg
−1
)
o
fcst s
i
−sensitivity/obs ( J Kg
−1
)
o
Fig. 9.8
The sensitivity of the 24-h forecast error to the observation error covariance weight
coefficient
s
i
. Displayed are average values per assimilation/forecast cycle (
left side
graphic) and
average values per observation (
right side
graphic)
By analogy with the OBSI estimates, the adjoint approach provides all-at-once
R
-sensitivity information for each data type, instrument, and observation location in
the time-space domain. For each observed parameter, the forecast error sensitivity
to the observation error covariance weight coefficient
s
i
(
9.44
) is shown in Fig.
9.8
.
-sensitivities have units of J/Kg
since all the weight coefficients are non-dimensional scalar parameters. Positive
values identify those data types whose reduced
The comparison is facilitated by the fact that all
s
i
values will be of potential
benefit to the forecasts and it is noticed that total precipitable water observations
exhibit the largest sensitivity/observation. Radiances and specific humidity are
identified in Fig.
9.8
as data types whose negative sensitivity values point toward
o
o
-
inflation. The presence of both positive and negative
-sensitivity values indicate
that an optimal weighting between the information provided by the background and
observations may not be achieved by adjusting a single scalar covariance coefficient
(e.g.,
s
i
s
b
-inflation) and that a systematic analysis of each data type and instrument is
necessary to optimize the DAS performance. To further illustrate this aspect, results
of forecast sensitivity to the
-weight coefficient for the Special Sensor Microwave
Imager (SSMI) total precipitable water and for the radiosonde specific humidities
are contrasted in Fig.
9.9
for each assimilation/forecast episode.
o
Search WWH ::
Custom Search