Global climate simulations of precipitation are typically evaluated on the basis of monthly or seasonal means over various geographic regions. However, the method by which precipitation is generated and the frequency with which it falls have implications for how diabatic heating feeds back onto the atmosphere’s large-scale thermal structure.

In this paper, we examine the distribution of precipitation as a function of rainfall rate in the Amazon Basin as represented in GPCP daily mean rainfall retrievals and in three versions of the Community Atmospheric Model (CAM). The CAM is run through three 16-month simulations with the following configurations: 1) the standard cumulus parameterization, which serves as a control, b) a run in which the standard parameterization is replaced by a series of 2D cloud resolving models (CRMs) running in each grid box with surface friction computed on the GCM grid, and 3) a second CRM simulation in which surface friction is computed on the CRM grid.

JJA and DJF basin-wide mean rainfall is slightly underestimated in the control run, whereas the CRM runs show greater underestimate. However, when the distributions of rainfall rates are examined, too much Amazon Basin rainfall results from very light precipitation rates in the control than is seen in the GPCP. CRM precipitation distributions generally show better agreement with GPCP distributions, except in the 10-30 mm/day range, which accounts for a large fraction of total rainfall. Furthermore, the spatial distribution of precipitation is more accurately represented in the CRM runs.

The above observations provide a starting point for further investigations in how simulated convection interacts with the larger scale environment. In some ways, the CRM runs are improvements over the control. The challenge is to determine how the different modes of handling convection lead to such improvements. We are currently analyzing Amazon Basin rainfall by low-level wind regime to determine how the different rainfall rate distributions affect diabatic heating profiles. Furthermore, one-month runs with hourly model output will be analyzed to further elucidate the convection-environment interactions. Time permitting, attention will also be given to the West Pacific warm pool region, where the CRM runs tend to grossly exaggerate rainfall amounts.