Simulation of the North American monsoon by the NCAR CCM3 and its sensitivity to convection parameterization

Two 9-yr runs of the NCAR Community Climate Model version 3 (CCM3) are compared in their simulations of the North American summer monsoon. In a control simulation, the Zhang-McFarlane deep convection scheme is used. For an experimental simulation, the following modifications to the scheme are implemented. The closure is based on the large-scale forcing of virtual temperature, and a relative humidity threshold on convective parcels lifted from the boundary layer is applied. The sensitivity to these modifications for simulating the North American monsoon is investigated. Model validation relies on hourly precipitation rates from surface gauges over the United States, hourly precipitation rates derived from the combination of microwave and radar measurements from NASA's Tropical Rainfall Measuring Mission (TRMM) satellite over Mexico, and CAPE values as calculated from temperature, specific humidity, and pressure fields from the NCEP-NCAR reanalysis. Results show that the experimental run improves the timing of the monsoon onset and peak in the regions of core monsoon influence considered here, though it increases a negative bias in the peak monsoon intensity in one region of northern Mexico. Sensitivity of the diurnal cycle of precipitation to modifications in the convective scheme is highly geographically dependent. Using a combination of gauge-based rainfall rates and reanalysis-based CAPE, it is found that improvements in the simulated diurnal cycle are confined to a convective regime in which the diurnal evolution of precipitation is observed to lag that of CAPE. For another regime, in which CAPE is observed to be approximately in phase with precipitation, model phase biases increase nearly everywhere. Some of the increased phase biases in the latter regime are primarily because of application of the relative humidity threshold.