Mesoscale Modelling for Tropical Cyclone Forecasting over the North Indian Ocean - Monitoring and Prediction of Tropical Cyclones in the Indian Ocean and Climate Change - page 273

Fig. 2: Initial vortex position error (km) for each case. Cases 1-5 for TC Nargis, cases
6-9 for TC Gonu, cases 10-13 for TC Sidr, cases 14-16 for TC KhaiMuk, cases 17-20
for TC Aila and cases 21-24 for TC Laila.
symmetric with stronger wind prevailing in the right sector in case of 3DVAR
(Fig. 3b) and in left sector in case of CNTL (Fig. 3a) at the surface. Comparing
with the QSCAT and SSM/I winds, the actual wind is higher in right sector
(http://www.SSM/I.com/qscat/qscat_browse.html) which supports the wind
distribution simulated by 3DVAR experiment. The outer core (>17 m s -1 ) is
slightly larger in 3DVAR (Fig. 3b) as compared to CNTL (Fig. 3a). Hence, to
summarize, the 3DVAR could explain better the inner and outer core and the
asymmetry of wind at the surface compared to CNTL experiment. Unlike
CNTL, 3DVAR experiment could capture two vorticity maxima viz. one close
to the right of the centre in the eye wall region (210 × 10 -5 s -1 ) and another
about 100 km away from the centre in the right side (150 × 10 -5 s -1 ). Comparing
with satellite images and TRMM observed rainfall (not shown); the Nargis has
two convective zones to the right of the centre which brought out realistically
in 3DVAR experiment (Fig. 3d). Corresponding to two vorticity maxima, there
are also two maxima in the vertical velocity in 3DVAR (Fig. 3f) unlike in
CNTL experiment (Fig. 3e). The eye region is warmer more than 370 K at 700
hPa in the case of 3DVAR (Fig. 3h), while it is 365 K only in CNTL simulations
(Fig. 3g).
The 3DVAR experiment improves the track prediction in all the cases as
demonstrated in Fig. 4. The mean 24-hr, 48-hr, 72-hr and 96-hr forecast track
errors (in km) are 176, 291, 327 and 425 for CNTL experiment and 132, 228,
176 and 224 for 3DVAR experiment respectively. Hence, there is an
improvement of 25%, 22%, 46% and 47% respectively with 3DVAR
assimilation experiment. Out of 24 cases, CNTL fails to predict the landfall in
nine cases and 3DVAR in four cases. The 3DVAR experiment could improve
the landfall prediction in 17 cases (out of 20 landfall cases). The mean landfall
error is reduced by 28% after the assimilation satellites derived wind data. The
improved structure, intensity and track with the assimilation of QSCAT and