GSI/WRF Regional Data Assimilation System and Its Application in theWeather Forecasts over Southwest Asia - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 628

25.7.2 Discussion

In this study, the weather forecasts using the WRF-ARW system were evaluated over
the mountain areas of Southwest Asia. Due to the complexity of the high terrain and
lack of knowledge in the estimation of physical processes in this area, forecasters
should have greater awareness of these limitations of the model forecasts in this
region.
First of all, the parameterization of physical processes plays a significant role
in the forecasting of surface temperature. For the 2-m temperature forecasts, the
systematic error component is larger than the random errors, and it is related to
the elevation of terrain. It should be noted that the areas of high bias shown in
Fig. 25.3 a correspond with the areas of rapid elevation change. These are the areas
where a difference in terrain height between the datasets would have the largest
effect. They are also the areas where the difference between the observational
station elevation and mean grid point elevation has the largest value. The lapse rate
effects due to these terrain height differences is probably another reason for the
2 m temperature bias. In contrast with the temperature fields, random errors play
a much bigger role in the forecasting of the upper level precipitation and 10-m
wind fields. The random errors constrain forecasters from presenting high quality
forecast guidance and are caused by a combination of uncertainty in the initial
conditions and unreasonable model scales. The detailed statistical results presented
in Sect. 25.4 are specific to the surface and the upper levels at nine locations. The
basic error characteristics for one forecasting variable change by the selected region,
and may not be representative of errors of other forecast variables. For example, in
the preliminary investigation of temperature errors, the results demonstrated that the
maximum 2-m temperature biases occurred over the high mountain areas while the
temperature biases at 500 hPa were found over most of Southwest Asia and it was
not related to the terrain configuration.
Note that the results presented here are for only one month of experimental
model runs; the accuracy of the forecast performance needs to be further verified
and investigated with more real-time forecasts. As expressed by Manning and
Davis ( 1997 ), “These statistics would provide additional information to model users
and alert model developers to those research areas that need more attention.” The
additional and complementary need for verification strategies in the WRF-ARW
model is elucidated in reference papers ( Skamarock et al. 2005 ).
Second, random errors are very complicated. It is only partially attributed to
the uncertainty in initial conditions. An accurate representation of initial conditions
would help users to compare the latest forecast guidance with current observations
and make appropriate adjustments in real time. The assimilation of satellite radiance
observations into a numerical weather prediction (NWP) model provides initial
conditions more closely representative of the true state of the atmosphere. The
results shown here demonstrate the positive impact of satellite data on weather
prediction in most of the Southwest Asia areas, but the impacts are not as obvious in
the high terrain areas, such as the Himalaya Mountain and Iranian Mountain regions.