Geoscience Reference
In-Depth Information
09 km horizontal resolution to evaluate the impact of radiance observations
relative to GTS data assimilation. The impact of the WRF model simulation
with the 3DVAR based initial conditions is examined in the section, where the
predicted meteorological fields during the cyclone period are compared with
the available observations.
We calculated the average root mean square errors (RMSE) of analyses
increments from GTS and RAD experiments for zonal, meridional winds, and
temperature at each model vertical sigma level. It is clearly noticed that the
mean RMSE from GTS (1.79 m/s, 1.09 m/s and 1.72 °C) is relatively high as
compared to the mean RMSE of analyses increments obtained from RAD (0.81
m/s, 0.61 m/s and 1.09 °C) for the above mentioned parameters respectively.
This information helps in diagnosing how best WRF-3DVAR has used different
components of the observing system in the best fit analysis. We also calculated
the RMSE of analyses increments at different model levels (figure not provided)
obtained from GTS and RAD experiments. In the model, 13 vertical levels are
below 850 mb while there are 23 levels below 500 hPa. Within the boundary
layer, RMSE is smaller for RAD data-assimilated analysis compared to GTS
analysis for all three variables. It is also clear that the range or spread of RMSE
is less in RAD analysis which clearly indicates its ability in improving the
background field for the different model vertical levels. The reduced departure
in the RAD-based initial conditions clearly depicts the best fit of observed
radiance data. It can also infer that the assimilation of both radiance and GTS
data produced relatively better analysis as compared to that of assimilation of
GTS data alone.
The model initial time and simulated (48 hrs, 96 hrs and 108 hrs) forecast
of mean sea level pressure (MSLP) obtained from three numerical experiments
are illustrated in Fig. 2(a-l). The MSLP patterns during the cyclone period
over the experimental domain are more or less well simulated in all the
simulations. However the intensity, movement and position of the cyclone with
different forecast time are significantly well simulated in the RAD experiments
as compared to other experiments. The vector displacement errors (VDEs)
from three numerical experiments are calculated with different forecast time
and provided in Table 2. Figure 2(a-c) shows the MSLP obtained from the
three analyses such as global FNL analysis, after assimilation of only GTS
(conventional and non-conventional observations) and after assimilation of
radiance observation along with GTS respectively. The initial position error
(64 km) of the cyclone is reduced significantly after assimilating the radiance
observations along with GTS (RAD expt.) as compared to the GTS (72 km)
and CNTL (140 km). The observed MSLP of the cyclone is 998 hPa at initial
time (00 UTC 26 December 2011), while the RAD analysis (996 hPa) revealed
a relatively more active system as compared to both the analyses FNL (997
hPa) and GTS (997 hPa). It is observed that the system crossed the Tamil Nadu
coast around 01 UTC of 30 th December 2011. None of the model simulations
show the landfall of the cyclone during the period. It is clearly seen from the
Search WWH ::




Custom Search