Performance of the CAM Simulations through a Stochastic Mulitcloud/Unified Shallow-Deep Convection Scheme over the Tropics

In this study, we focus on the performance of the National Center for Atmospheric Research (NCAR) Community Atmosphere Model (CAM) version 5 over the tropics, using a stochastic modification of the Zhang and McFarlane cumulus parameterization that also unifies the representation of deep and shallow convection. The results from a 15-year simulation of the unified-stochastic CAM (USM-CAM) are examined and compared with the CAM simulations using the default Zhang and McFarlane parameterization (referred to here as the control, CTRL for short). The Stochastic Multi-cloud model (SMCM) is based on an interacting particle system on a lattice whose microscopic configuration represents various cloud types (namely, shallow cumulus, cumulus congestus, deep, and stratiform) as they interact with each other and with the environment. The SMCM’s stochasticity and equilibrium measure are by design dependent on a set of “transition timescale” parameters that can be inferred from radar data using machine learning. Here, three sets of experiments have been conducted based on three different sets of transition timescale parameters, one of which is inferred from the Dynamics of the MJO (DYNAMO) field experiment. Although, the SMCM itself is highly dependent on the transition time-scale parameters, overall all three USM-CAM simulations perform reasonably well in capturing the mean state climate. This improvement might be due in part to the overall improvement in the ability of the USM-CAM simulations to produce realistic proportions of large/stratiform versus convective rainfall while the CTRL simulation is dominated by convective precipitation—the Zhang McFarlane scheme seems to rain too much and too often.