In this study the Penn State-NCAR nonhydrostatic mesoscale model, MM5, is used to perform microphysical parameterization sensitivity tests on Typhoon Bilis (2000). For the 3-nested model runs, the microphysics schemes produced a track that was very close to observations. The 3-nested simulation with Goddard schemes deepened the tropical cyclone to 928 mb before impinging on Taiwan, while the Reisner '98 3-nest simulation produced a weaker storm, with a pressure bottoming out at 936 mb. Observed pressures from JTWC were down near 915 mb. The weaker storm in Reisner2 may be tied to its prediction of number concentrations of hydrometeors. The Goddard scheme uses the Marshall-Palmer distribution to predict number concentration. Rainfall was ~10% greater than observations (949 mm) in 3-nest simulation with Goddard scheme, while the Reisner '98 simulation produced rainfall about ~10% below observations.

The failure of the 3-nested model runs to produce a tropical cyclone with a lower pressure closer to observations is likely due to the small size of the 5 km domain. The domain nesting appears to have had an effect on the structure of the cyclone. The structure of the tropical cyclones simulated in the 3-nested runs had an almost complete absence of a northern eye-wall. Air from the 15 km domain was modified by the Grell cumulus scheme. This air entered the 5 km domain and got entrained into the cyclone. This air was too stable for the grid-explicit microphysics to produce precipitation.

Too avoid the issues described above, an expanded 5 km domain is used to resolve Typhoon Bilis throughout the entire simulation. Each of the 3-ice class scheme simulations were able to produce a complete eye-wall, and both schemes produced storms that had a minimum pressure of around 920 mb. Rainfall produced by the 2 schemes varied. The Goddard simulation had rainfall amounts close to observations. The Reisner '98 simulation gave lower values, 25% below actual rainfall. The lower rainfall amounts in the Reisner '98 scheme may be due to its prediction of number concentrations of hydrometeors.

Effects of PBL and latent heating will also be investigated.