In this study the CAM3 model with a new Zhang-McFarlane convection scheme (CAM3Z) is used to simulate the tropical intraseasonal oscillation (TIO). The results are compared to the observations and the original CAM3. The model is run using observed, time-varying monthly SST as the lower boundary condition for about 50 years(Sept 1949-Oct 2001). The comparison shows the following:

1) The NCAR CAM3 fails to simulate the tropical intraseasonal oscillation realistically: The model cannot produce robust signal of intraseasonal time scale oscillation; the simulated TIO of CAM3 does not travel eastward; and the space distribution and annual cycle are different from the observation.

2) On the other hand, the CAM3 with revised Zhang_Mcfarlane convection scheme greatly improves the simulation ability. It produces robust large-scale oscillation on intraseasonal time scale. This TIO travels eastward with the speed similar to observation. Also it has spatial distribution and annual cycle similar to those observed.

3) There are still some deficiencies in the TIO simulation ability of the CAM3 with revised Zhang_Mcfarlane convection scheme. The most serious one is the too strong amplitude of the signals, particularly in the western Pacific. For example, the simulated annual cycle of the CAM3Z is too strong, especially in boreal summer. The simulated TIO in the western Pacific is stronger than observed while in the Indian Ocean it is weaker.

4) The analysis shows that the vertical distribution of the latent heating plays a very important role in generation of the tropical intraseasonal oscillation.