Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian climate centre general circulation model

Abstract A simplified cumulus parameterization scheme, suitable for use in GCMs, is presented. This parameterization is based on a plume ensemble concept similar to that originally proposed by Arakawa and Schubert (1974). However, it employs three assumptions which significantly simplify the formulation and implementation of the scheme. It is assumed that an ensemble of convective‐scale updrafts with associated saturated downdrafts may exist when the atmosphere is locally conditionally unstable in the lower troposphere. However, the updraft ensemble is comprised only of those plumes which are sufficiently buoyant to penetrate through this unstable layer. It is assumed that all such plumes have the same upward mass flux at the base of the convective layer. The third assumption is that moist convection, which occurs only when there is convective available potential energy (CAPE) for reversible ascent of an undiluted parcel from the sub‐cloud layer, acts to remove CAPE at an exponential rate with a specified adjustment time scale. The performance of the scheme and its sensitivity to choices of disposable parameters is illustrated by presenting results from a series of idealized single‐column model tests. These tests demonstrate that the scheme permits establishment of a quasi‐equilibrium between large‐scale forcing and convective response. However, it is also shown that the strength of convective downdrafts is an important factor in determining the nature of the equilibrium state. Relatively strong down‐drafts give rise to an unsteady irregularly fluctuating state characterized by alternate periods of deep and shallow convection. The effect of using the scheme for GCM climate simulations is illustrated by presenting selected results of a multi‐year simulation carried out with the Canadian Climate Centre GCM using the new parameterization (the CONV simulation). Comparison of these results with those for a climate simulation made with the standard model (the CONTROL simulation, as documented by McFarlane et al., 1992) reveals the importance of other parameterized processes in determining the ultimate effect of introducing the new convective scheme. The radiative response to changes in the cloudiness regime is particularly important in this regard.