Masters Thesis Abstract: Tristan Shepherd
A numerical study of Tropical Cyclone Sidr (2007): sensitivity analysis using Advanced Weather Research and Forecasting (WRF-ARW)
Tropical cyclones are a special class of large-scale rotating wind system occurring over tropical and subtropical oceans. They are fundamentally a mechanism to transport heat and moisture poleward when the Hadley Cell is unable to complete this exchange rapidly enough. Their development is based upon a set of key criteria, being: an underlying ocean with surface temperatures exceeding 26°C, small wind speed and direction changes with height, a pre-existing region of lower tropospheric horizontal wind convergence, cumulonimbus cloud development resulting from turbulent mixing and saturation of surrounding air, and a location poleward of about 4°-5° of latitude.
Tropical Cyclone Sidr, part of the 2007 Northern Indian Ocean hurricane season officially claimed the lives of 4,234 people in Bangladesh (Centre for Research on the Epidemiology of Disasters (CRED), International Disaster Database, 2008). According to the Bangladesh Ministry of Disaster Management, an estimated 2.7 million people were affected, 773,000 houses damaged, and total damages in the order of $450 million forecast.
In this study, the dynamics of Sidr are explicitly simulated using a mesoscale meteorological model, being version 2.2 of the National Centre for Atmospheric Research (NCAR) non-hydrostatic, two-way interactive, movable, multiple nested grid (WRF-ARW) model.
The model is initialised with the National Centres for Environmental Prediction (NCEP)-NCAR reanalysis dataset. Three initialisation times were tested, a 96-hr, a 72-hr, and a 48-hr, to examine the model’s ability to simulate important features of Sidr, such as track, and cyclone intensity and structure. Three cumulus parameterisation schemes were used each time, the Kain-Fritsch, Betts-Miller-Janjic, and Grell-Devenyi, to resolve convective processes over the model domain. Simulations were conducted using a single domain of 15 km grid resolution with 150 points in the north-south and east-west directions (area coverage 2500 x 2500 km2). Effectively resolving the multi-scale processes remains a critical issue of hurricane research.
Simulations were verified against observations from the MODIS and TRMM satellite missions, and QUICKSCAT scatterometry winds, with the JTWC best track dataset also used for comparison. Preliminary results indicate the 72-hr Kain-Fritsch simulation performed best with regard to forecast cyclone track, although both the 72-hr and 48-hr simulations poorly resolved cyclone structure and intensity. The 96-hr Kain-Fritsch simulation captures reasonably well the evolution and inner core structures of the storm, but is somewhat erroneous with regard to storm track. The Kain-Fritsch simulations were consistently the best performing across all initialisation times but tended to produce spurious precipitation with increasing radius from the eye wall. The Betts-Miller scheme tended to best resolve convective processes compared to those observed.
The 96-hr Kain-Fritsch simulation reproduced minimum cyclone surface pressure (948hPa) and maximum surface wind (Vmax) of 144 km/h, which was the closest to the observed best track data. When compared to observed pressure of 918hPa and Vmax (259 km/h), the skill of the model is somewhat poor. It is hoped that by improving model setup, the accuracy of WRF will improve for this event.
