Abstract
The parameterization of deep convection in the ACME atmosphere model continues to be based upon the Zhang-McFarlane (1995, ZM) parameterization scheme. Only moderate adjustments have been made to this scheme since the CCSM4 version of the NCAR model. In preparation for the v1.0 version of the model we have probed many of the parameters key to the operation of the ZM scheme. Initially, these sensitivity experiments were performed in CAM5.5, a largely similar configuration to v1.0 in it's moist physics representation. A range of parameters that are not frequently used as tuning parameters were investigated including parcel temperature perturbations, downdraft fraction and CAPE microphysical calculation assumptions. The aims of this study were two-fold. To improve South American DJF rainfall biases, which currently have a severe impact on the Amazon land surface due to excessive dryness, and to increase the strength of tropical modes of variability which remain poor from the v0.3 version.
The dominant parameter change that had the greatest impact in the mean climate and topical variability was a parameter value used for the calculation of vertical integrated buoyancy or Convectively Available Potential Energy (CAPE). The current setting of this parameter allows the buoyancy profile to go negative up to 5 times before the top of convection is determined. When only one negative buoyancy region is allowed for the cloud top calculation, there is a significant improvement in mean precipitation biases over the tropics and enhancement of the Madden Julian Oscillation (MJO) wave mode characteristics (strength and phase speed).
It is speculated that this model response is due to a greater sensitivity to dry and stable conditions in the vertical. With this in mind, we contrast this behavior with simulations from the ACME v1.0 prototype atmosphere since the increased vertical resolution (72 vs. 32 levels) has the potential for greatest structure and therefore more stable layers in the vertical. Results will be presented here from equivalent ACME v1. and CAM5.5 experiments.