Geoscience Reference
In-Depth Information
Fig. 6.5 Impacts by
observation category on a per
observation basis (multiplied
by 1,000) for 3 days of 12 h
COAMPS forecasts Sep
22-25 2011. The horizontal
grid spacing was ( a )90km
and ( b ) 30 km. Values are in
units of J kg 1
the model domain covers is equal for both cases, the analysis area is larger for the
90 km grid. Therefore, more observations are being assimilated on the 90 km grid
which reduce the per observation impact values. This can be seen by comparing the
amount of shading south of 15 ı NinFig. 6.6 a, b.
Many more experiments need to be conducted before any definitive statements
can be made about the effect of horizontal grid spacing on observation impact
calculations. One other consideration is the manner in which the error is calculated.
The truth is an analysis field on the same grid as the model forecasts, so it is not
consistent between the two grid configurations (30 km forecasts are compared with
30 km analyses and 90 km forecasts are compared with 90 km analyses). Future
experiments would benefit from comparisons with a consistent truth.
6.3.4
Other Metrics
As noted earlier, the error for all other experiments in this chapter was calculated in
a dry energy norm for the lowest 20 model levels. This is a metric that is suitable
for a global observation monitoring system; however, mesoscale applications may
require different error considerations. This section demonstrates that a change in the
error metric can also change the relative importance of observations in reducing that
metric. Impacts for the same COAMPS 12 h forecast as in Sect. 6.3.1 were computed
Search WWH ::




Custom Search