Evaluating the Scale-Awareness of Convective Parameterization Using Cloud-Resolving Model Simulation

Description

Current convective parameterization schemes were designed for low-resolution global climate models (GCM). As the model resolution increases, they will likely be inadequate for properly parameterizing subgrid-scale convective processes. To improve the scale-awareness of convective schemes, in this study we evaluate various aspects of the scale-awareness of subgrid-scale convective processes and properties, and their parameterization using cloud resolving model output. A convective parameterization scheme generally consists of three components: trigger, closure and a simple cloud model, each of which can depend on the resolution of the host model. By coarse-graining the CRM output to subdomains of different sizes (equivalent GCM resolutions), we examine the dependence of a number of trigger functions and closures on GCM resolution. We found that all the trigger functions examined are scale-dependent, especially for dCAPE-type triggers, with skill scores dropping precipitously from low resolutions to high resolution. While the CRM data show a decreasing frequency of convection occurrence as GCM resolution increases, the trigger functions either predict rapidly increasing convection frequency or unchanging frequency. Remedies are proposed to reduce these over-prediction biases as the GCM resolution increases. For closure, it is found that in CRM data for the same amount of GCM-scale forcing, cloud-base convective mass flux decreases as GCM resolution increases. This is not built into current convective parameterization schemes. Methods to improve this aspect will be discussed.