The resolution dependence of model physics: Illustrations from nonhydrostatic model experiments

Abstract

The goal of this paper is to gain insight into the resolution dependence of model physics, the parameterization of moist convection in particular, which is required for accurately predicting large-scale features of the atmosphere. To achieve this goal, experiments using a two-dimensional nonhydrostatic model with different resolutions are conducted under various idealized tropical conditions. For control experiments (CONTROL), the model is run as a cloud-system-resolving model (CSRM). Next, a "large-scale dynamics model" (LSDM) is introduced as a diagnostic tool, which is a coarser-resolution version of the same model but with only partial or no physics. Then, the LSDM is applied to an ensemble of realizations selected from CONTROL and a "required parameterized source" (RPS) is identified for the results of the LSDM to become consistent with CONTROL as far as the resolvable scales are concerned. The analysis of RPS diagnosed in this way confirms that RPS is highly resolution dependent in the range of typical resolutions of mesoscale models even in ensemble/space averages, while "real source" (RS) is not. The time interval of implementing model physics also matters for RPS. It is emphasized that model physics in future prediction models should automatically produce these resolution dependencies so that the need for retuning parameterizations as resolution changes can be minimized.