Impact of Variational Data Assimilation for Simulating Tropical Cyclones over Bay of Bengal Using WRF-ARW - Monitoring and Prediction of Tropical Cyclones in the Indian Ocean and Climate Change - page 237

Fig. 4: Time evolution of average of errors in (a) track (in km), (b) central sea level
pressure (hPa) and (c) maximum sustained winds (m/s) for the ten cyclones.

48, 72, 96 h simulation period) in CSLP evolution of the storms with assimilation
PrepBufr observations using 3DVAR.
The time variation of average errors in maximum winds of all 10 cyclones
followed same trend as that of CSLP. The errors are of the order of -16, -31, -
25 and -18 m/s at 24h, 48h, 72h and 96h respectively indicating overestimation
of intensity of simulated cyclones in control runs (Fig. 4c). The corresponding
average errors for maximum winds with 3DVAR experiments are
-15, -25, -19, -20 m/s respectively which indicate a reduction of 0.5%, 19%,
22%, -10% in the wind errors at 24, 48, 72, 96 h of simulation with 3DVAR.
To analyse the model simulated rainfall intensity in both 24 h accumulated
rainfall for the fourth day forecast from control and 3DVAR experiments are
compared with corresponding TRMM rainfall estimates. Spatial statistics of
simulated 24 h rainfall for four days from both experiments for the 10 tropical
cyclone cases are made with the corresponding TRMM spatial rainfall data
(Table 2).

Table 2: The statistical performance of the WRF model in comparison with TRMM

Forecast Average BAIS RMSE Correlation

day rainfall (mm)

TRMM

CTL 3DVAR CTL 3DVAR CTL 3DVAR CTL 3DVAR

Day-1 9.88 11.38 11.01 -1.49 -1.35 20.90 20.44 0.62 0.64

Day-2 8.40 13.46 12.77 -5.05 -4.47 31.14 30.69 0.45 0.53

Day-3 6.90 14.67 13.25 -7.77 -6.45 33.73 32.35 0.40 0.56

Day-4 5.94 10.02 9.34 -4.07 -3.82 27.45 26.75 0.33 0.45