Impact of Radiance Data Assimilation on Simulation of Tropical Cyclone Thane Using WRF-3DVAR Modelling System - Monitoring and Prediction of Tropical Cyclones in the Indian Ocean and Climate Change - page 386

Fig. 4: Time series (12 hrly interval) of (a) MSLP (hPa) and (b) maximum wind
sustained (m/s) at 10 m.
is clearly observed from the figures that the assimilation experiments simulated
the intensity of the storm very well as compared with the CNTL simulation in
the whole forecast period. However, the intensity of the storm is more realistic
in the RAD simulation and the MSLP pattern is also close to the observed
pattern for the whole forecast period. Similarly, the assimilation experiments,
mainly the RAD simulation, well simulated the surface wind as compared
with the CNTL simulation (Fig. 4b). The magnitude of surface wind simulated
by CNTL experiment is not consistent with the observed wind pattern
throughout the forecast period. The RMSE of MSLP (3.5 hPa) and surface
wind (3.3 m/s) are found less in the RAD as compared to the CNTL (6.5 hPa,
5.6 m/s) and GTS (4.3 hPa, 4.2 m/s) respectively. It is noticed that the RMSE
values are also less in the GTS corresponding to the CNTL simulation.
6. Conclusion
The mesoscale model WRF (ARW) and 3DVAR data assimilation systems
with a single domain at 09 km horizontal resolution are used to simulate the
very severe cyclone Thane that occurred over BoB. For this purpose, three
numerical experiments are carried out to assess the impact of the high-resolution
analysis prepared using the radiance observations through 3DVAR analysis
system. The improvements in the high-resolution analysis (3DVAR) from RAD
experiments are uniformly better as compared to the global low-resolution
FNL as well as GTS analyses.
Comparing the simulated results obtained from RAD experiments with
CNTL and GTS experiments, it is clearly seen that the intensity, movements
and position of the cyclone with different forecast hours are substantially
improved in the RAD simulation.
The mean VDEs are significantly decreased due to the assimilation of
radiance observations along with GTS as compared to the CNTL and GTS
experiments. It is clearly demonstrated that the gain skill of the radiance data
assimilation over CNTL is significantly high around 62% compared to that of
the GTS experiment (32%). Similarly, the gain skill of RAD simulation over
GTS is also high around 45% in whole forecast period.