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 271

Table 1: Brief description of ARW and NMM

Model WRF-ARW WRF-NMM

Dynamics Nonhydrostatic Nonhydrostatic

Horizontal grid system Arakawa C-grid Arakawa E-grid

Map projection Mercator Rotated latitude-longitude

Horizontal resolution 9 km 9 km

Vertical coordinate Terrain following sigma Terrain following hybrid sigma

vertical coordinates pressure vertical coordinates

Vertical levels 51 51

Cumulus convection Kain-Fritsch Simplified Arakawa Schubert

PBL parameterization Yonsei University (YSU) Yonsei University (YSU)

Land surface physics Monin-Obukhov NMM

Microphysics WRF single-moment Ferrier (new eta)

3-class (WSM3)

Radiation RRTM LW/ Goddard SW GFDL LW/SW

3. Results and Discussions

3.1 ARW Model Performance in Real Time
Prediction of TCs over NIO

The performance of ARW model for the prediction of TCs over the NIO is
assessed. This evaluations based on 97 cases (initialized at 00 and 12 UTC of
each day during TC period) of 16 landfalling TCs over the NIO during 2007-
10. The initial and boundary conditions are obtained from National Centers for
Environmental Prediction (NCEP) global forecasting system (GFS) analyses
and forecast fields respectively. The model based TC centres are obtained based
on minimum sea level pressure.
The mean direct position errors (km) for the TCs over the NIO (based on
97 cases) using ARW model are given in Fig. 1. Along with the model error
calculation, the skill of the model with respect to persistence method is also
calculated. The ARW model forecasts become increasingly skillful with increase
in lead time compared to persistence forecasts. Further, the model forecasts
are more consistent for all forecast lengths as standard deviation of model
errors is smaller compared to those of persistence track errors (not shown).
The mean initial vortex position error in the GFS analyses is about 79 km
which is considerably large and results in negative skill (of about -8%) at 12-
hour model forecast. The mean track forecast errors vary from 130 km to 350
km from 12 to 72 hour forecast. The mean landfall position errors are 73, 108
and 139 km in advance of 24, 48 and 72-hour lead for the NIO TCs and the
mean time errors (standard deviation) are -7 (15), -3 (13) and -5 (6) hours for
72, 48 and 24-hour forecast (not shown).