Geoscience Reference
In-Depth Information
Assimilated Glider minus FR Temperature
Assimilated Glider minus FR Salinity
0
100
200
3
2
1
0
0
100
200
0.4
0.2
0
06
10
14
18
22
26
06
10
14
18
22
26
Assimilated Glider minus FG Temperature
Assimilated Glider minus FG Salinity
0
100
200
3
2
1
0
0
100
200
0.4
0.2
0
06
10
14
18
22
26
06
10
14
18
22
26
Assimilated Glider minus 4DVR Temperature
Assimilated Glider minus 4DVR Salinity
0
100
200
3
2
1
0
0
100
200
0.4
0.2
0
06
10
14
18
22
26
06
10
14
18
22
26
Days of August, 2003
Days of August, 2003
Fig. 15.7
Absolute model temperature (
left
) and salinity (
right
) discrepancies to assimilated
observations for the free run (
top
), first guess (
middle
) and analysis (
bottom
)
The difference between the observations and the model is computed for all
assimilated profiles of temperature and salinity and plotted in chronological order
in Fig.
15.7
. It can be seen that the temperature differences are confined in the upper
100 m of the water column, with magnitudes sometimes reaching 3 K for both the
free run and the first guess. Salinity differences extend deeper in the water column,
to about 200 m, although the largest differences are confined to the upper 100 m. A
slight improvement can be noticed from the free run to the forecast solutions in the
temperature field, but not as much in the salinity field. However, the assimilation is
able to significantly reduce the forecast discrepancies in both the temperature and
salinity fields, with the exception of a few profiles at the beginning of each cycle.
The assimilation is able to reduce discrepancies as high as 3 K and 0.4psu to less
than 0.5 K and 0.1psu in temperature and salinity respectively.
The forecast solution is expected to have smaller discrepancies to the observa-
tions than the free run, because it is initialized with the analysis at the end of a
previous cycle. So, having only a marginal improvement from the free run to the
first guess is an indication that the gains from the assimilation are short-lived in the
forecast run as a consequence of inadequate forcing fields driving the model away
from future observations
15.4.5
Independent Observations
For verification and evaluation purposes, discrepancies are computed between the
withheld glider observations and the three model solutions: the free run, the first
guess and the analysis. Results in Fig.
15.8
show that all three solutions have similar
error levels with respect to the un-assimilated as to the assimilated observations.
Search WWH ::
Custom Search