Geoscience Reference
In-Depth Information
gradients through the mixing ratio - dewpoint depression - pseudo relative humidity
sequence. The NAVDAS adjoint code is checked by comparing calculations of
ıe f
e f
in observation and model space and with values of
.
6.3
Observation Impacts for COAMPS/NAVDAS
As discussed in Sect. 6.1.4 , the adjoint observation impact framework has been
primarily implemented for global modeling systems. This section will highlight
some of the unique aspects of the framework for a limited area model such as
COAMPS.
The flow of error information is shown in Fig. 6.1 for a 12 h COAMPS forecast
with 60 km grid spacing valid at 1200 UTC 05 May 2010. The forecast error was
computed on the lowest 20 model levels (out of 30) from the surface to the upper
troposphere in a dry energy norm. All of the experiments in this chapter (except in
Sect. 6.3.4 ) will use a dry energy norm in the lowest 20 model levels. The sum of
the components of
e f
in the vertical at each model grid point in the horizontal is
presented in Fig. 6.1 a. The blue shaded areas indicate where the error in the forecast
from the analysis field is less than the background forecast valid at the same time.
The COAMPS adjoint model integrates this information backward in time so that
the components of
ıe f
in analysis space can be shown (Fig. 6.1 b). Similar plots in
observation space are shown for radiosondes and aircraft data (Fig. 6.1 c, d) with the
aid of the NAVDAS adjoint.
A small majority of observations contribute to the overall reduction in forecast
error, which is why there are many observation locations with red shading in
Fig. 6.1 c, d. The percentage of beneficial observations in COAMPS/NAVDAS is
similar to values for other global systems ( Gelaro et al. 2010 ). The area of the
forecast error calculation is smaller than the model domain, but the COAMPS
adjoint model spreads error information to grid points outside the forecast error area.
The smaller error area was chosen because of the influence of the lateral boundaries
which will be discussed in the next section.
6.3.1
Lateral Boundaries
In the original framework for an adjoint observation impact system, Langland and
Baker ( 2004 ) noted that the reduction in forecast error after the analysis procedure
was due entirely to the assimilation of new observational information. This is
true for a global NWP model, but for a limited area model like COAMPS, the
lateral boundary conditions are also updated with the new analysis. Fortunately the
framework allows for a quantitative estimation of the lateral boundary effects.
To illustrate, three impact experiments were conducted that only differed in the
area over which the forecast error reduction was calculated. The dark boxes in
Search WWH ::




Custom Search