Strategy for testing candidate convection schemes at high resolution

1.Poster Title	Strategy for testing candidate convection schemes at high resolution
2.Authors	Wuyin Lin, Shaocheng Xie, Steve Klein, Yuying Zhang, Hsi-Yen Ma, Qi Tang, Erika Roesler , Mark Taylor, Richard Neale, Jin-Ho Yoon (Unlicensed), Phil Rasch (pnl.gov), Vince Larson, Andrew Gettelman (Unlicensed), P. Bogenschutz, Sungsu Park (Unlicensed)
3.Group	Atmosphere
4.Experiment	AMIP, RRM, CAPT hindcasts
5.Poster Category	Early Results
6.Submission Type
7.Poster Link	ACME_15_Results_Poster_Convection_Lin.pdf

Abstract

This work describes a strategy for testing candidate convection schemes at high resolution. Candidate schemes are being tested using a multi-tier infrastructure that is designed to assess model performance under various configurations. Long-term AMIP simulations are commonly used to assess the performance of candidate schemes in capturing observed multi-scale spatiotemporal variability of key physical quantities. For testing at high resolution (ne120), especially for a coupling of physics and configuration that has yet to be well tuned during the development stage, a more economical strategy is needed. One such strategy is to use static regional refinement to test the response of new physics at high resolution for a specific region (see posters by Roesler et al., and Tang et al.). Here a strategy for testing the response at globally uniform high resolution is presented. It relies on the use of Cloud Associated Parameteization Testbed (CAPT)’s short-term hindcasts. The feasibility of this strategy is illustrated by using the model configuration at low-resolution (ne30), with both multi-year AMIP simulations and short-term hindcasts. The analysis of the ne30 simulations shows that the relative performance between different schemes in terms of almost all major metrics is consistent between AMIP simulation and short-term hindcasts. These includes mean state, diurnal characteristics, relationship between physical quantities, and probability distribution of key variables associated with the new physics, except for some variability metrics at scales beyond the hindcast range. The consistency can be even largely maintained with just a small subset of hindcasts for a single season. High-resolution hindcasts for the same subset of initial times are further analyzed to show the effectiveness of this approach. Results for both CLUBB and UNICON are presented. The analysis suggests that the strategy can be an effective setup for further tuning of the model and its coupling with new physics.