The role of cumulus schemes in the reproducibility of tropical cyclones by the NCAR Community Climate Model (CCM3)

It is a challenging problem to realistically simulate various aspects of tropical cyclones in a long-term integration with an atmospheric general circulation model. Although the NCAR Community Climate Model (CCM3) produces marked improvement in the mean properties of climate, such as energy balance and precipitation, it has been reported that its performance was degraded with respect to the simulation of tropical transiences, such as tropical cyclones (TCs) and tropical intra-seasonal oscillations, for which its predecessor CCM2 demonstrated a good capability. The purpose of this study is to shed light on this problem by investigating the role of two cumulus schemes in the CCM3 regarding the frequency of TC appearance with the hope of achieving better model performance on both the mean climate states and tropical transiences.

As the moist convection processes in the CCM3, Zhang and McFarlane (ZM, 1995) scheme for deep convection and Hack (1994) scheme for shallow convection are sequentially applied to adjust the vertical distributions of temperature and moisture and to produce convective precipitation. In the ZM scheme, moist convection occurs wherever positive convective available potential energy is present. In this study, as an additional condition for triggering the ZM scheme, we implement a relative humidity threshold (RHc) so that deep convection does not occur unless the relative humidity in the model's sub-cloud layer exceeds a threshold value. We conducted one-year integration each with the T42 version of CCM3 for four cases of threshold values: RHc=0, 80, 85, and 90%, and examined the frequency of simulated TCs, defined as lows over the oceans between 40N and 40S with specified criteria for sea level pressure gradient and warm-core structure. Climatological sea surface temperature data are prescribed in the model.

Although virtually no TCs appear in the simulation with the original CCM3 (RHc=0%), the frequency of TC appearance dramatically increases as the threshold value increases. In the case of RHc=85%, the model produces almost the same TC frequency as in the CCM2, which is in agreement with the observed frequency of TCs. What has happened in these experimental runs is that the use of relative humidity threshold changes the partition between the "deep" ZM convection and the "shallow" Hack convection, while maintaining the global mean convective precipitation rate almost unchanged. For example, the contribution of the ZM scheme to the global mean convective precipitation rate is 74% and 37% for the case of RHc=0% and 85%, respectively. Apparently, the increased contribution of the Hack scheme associated with the decreased contribution of the ZM scheme is beneficial for the reproducibility of TC formation, which implies that such modifications enhance the interaction of cumulus-scale and large-scale circulations that plays an essential role in the formation process of TCs.