Forecasting tropical cyclone precipitation in the Philippines using MM5

This study evaluates the skill of the PSU/NCAR mesoscale model (MM5) in predicting precipitation within the Philippine Area of Responsibility (PAR) during typhoon/tropical storm events.

Accumulated rainfall and rain rate forecasts of six tropical cyclones that occurred in 2003 are compared with satellite and station data. Twelve cases using different cumulus, planetary boundary layer and explicit moisture parameterization schemes are used for each tropical cyclone. The model is initialized using large-scale forecast data from the Japan Meteorological Agency’s Global Spectral Model and run on a domain that includes the PAR (100-180E longitude and 0-60N latitude) with a 90 km horizontal resolution. Forecasts are made for 72 hours at 6-hour intervals.

Results show that cases involving the Betts-Miller cumulus parameterization scheme provide the best estimates in terms of spatial coverage when compared with satellite observations. The Betts-Miller cases also have high correlation with station data for accumulated rainfall but values are underestimated.

The case using the combination of Anthes-Kuo cumulus, Medium-Range Forecast planetary boundary layer and Schultz explicit moisture parameterizations provided a high correlation for rain rate although it also underestimates the amount of rainfall.