GSI/WRF Regional Data Assimilation System and Its Application in theWeather Forecasts over Southwest Asia - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 614

48-h forecasts (not shown). The larger magnitude of MSEs is randomly distributed
over the central Saudi Arabia, southeast Iraq, northwest Iran and west Himalaya
mountain region (Fig. 25.8 b). The corresponding SD error (Fig. 25.8 c) reveals that
nonsystematic errors compose a substantial portion of the total error. The fraction
of squared biases to the MSE (Fig. 25.8 d) is far less than 50 % except for west
Himalaya mountain region (E), which showed clearly that a larger contribution to
the MSEs is from a nonsystematic total model forecast error. Compared to 24-h
forecasts, the 48-h forecasts' bias is higher over most of study areas (not shown).

25.5.1.5 Winds at 200 hPa

Similar to the upper level temperature forecasts, the wind forecasts (the first 24-h
forecasts shown only) at 200 hPa (Fig. 25.9 ) indicate that the MSEs are dominated
by nonsystematic errors in either the zonal or meridional wind component or
both. For the zonal wind forecasts, the large MSE over Himalaya mountain region
is consistent with nonsystematic error, as well as, the Arabian Sea also has a
strong nonsystematic error signature. For the meridional wind component, the larger
forecast errors occur over a different place relative to the zonal wind forecasts. The
larger MSEs for the zonal wind forecasts in the Himalaya mountain region disappear
in the meridional wind field.
In summary, the 2-m temperature forecast error is typically caused by systematic
error and is most associated with the elevated terrain; by contrast, precipitation,
10-m wind speed, and upper level forecast errors are dominated by the nonsystem-
atic errors, which do not appear correlated with terrain.

25.5.2 Diurnal Variation

Based on model forecasts, the Southwest Asian domain-wide mean of the 2-m
temperature exhibits a minimum near 0000 UTC followed by a sharp increase to a
maximum near 1200 UTC (not shown). The difference of variables (temperature and
wind speed) at maximum (1200 UTC) and minimum (0000 UTC) time is defined as
the diurnal cycle variation in this study.
The Southwest Asia region's mean diurnal cycle of 2-m temperature during the
30-day study period (Fig. 25.10 a) shows that the amplitude of temperature diurnal
cycle for model forecasts is considerably lower than the value in the WMO GTS
observations. Note a slight deepening in the diurnal temperature cycle on May 2, 7
and 17 in GTS observations that is not reproduced in the model forecasts. These two
points indicate the near surface diurnal temperature cycle in model forecasts has a
serious problem.
The 10-m model forecast wind speeds exhibit a different behavior from that of
temperature. Similar to the NCEP GFS analysis data (GFS ANL), the amplitude
of the wind speed diurnal cycle in the model (Fig. 25.10 b) shows a strong diurnal