1.Poster Title	Strategy for Tuning High-Resolution ACME V1 Model
2.Authors	Wuyin Lin, Shaocheng Xie, Po-Lun Ma, Phil Rasch (pnl.gov) (Unlicensed), Yun Qian, Hui Wan, Steve Klein, Chris Golaz, Philip Cameron-Smith (Unlicensed), Qi Tang, Hailong Wang, Kai Zhang, Susannah Burrows, Julio Bacmeister (Unlicensed), Richard Neale, Cecile Hannay, Vince Larson, Andrew Gettelman (Unlicensed)
3.Group	Atmosphere
4.Experiment
5.Poster Category	Early Results
6.Submission Type	poster
7.Poster Link
8.Lightning Talk Slide	WLIN_ACME_One-Slide_Poster_Presentation_Research.pptx

Abstract

This work describes an effective strategy for tuning the ACME V1alpha model at high resolution. ACME v1alpha model at high resolution (0.25°), when assuming the parameters considered well-tuned at coarse (1°) resolution, tends to always exhibit large biases in spatial structure as well as global mean statistics in cloud, precipitation and energy budget. A long list of parameters over ranges are found to affect the simulation significantly, as observed from coarse-resolution tuning efforts (see Ma et al.’s poster) and Perturbed Parameters Ensemble (PPE) experiments (see Qian et al.’s poster). Tuning a multitude of parameters at high resolution is computationally expensive, particularly if overly relying on long-term AMIP simulations, though AMIP-type simulations are commonly used and eventually necessary to assess the overall model performance. A more economical tuning strategy is to extensively employ the short-term hindcasts using the Cloud Associated Parameterization Testbed (CAPT) framework. The CAPT tests have been found to be effective in numerous previous studies in identifying model biases due to parameterized physics. It has been proposed by the ACME atmosphere team as a major framework for testing new physics and tuning new model configurations. For the ACME v1alpha at 0.25° resolution, it is confirmed that, given the same parameters, the major biases in global mean statistics are consistent between AMIP-type free simulations and CAPT-type hindcasts, even with just a small number of short-term simulations over the corresponding season. The use of CAPT to find parameter choice for the reduction of large model biases dramatically improves the turnaround time for high resolution tuning. An iterative CAPT-AMIP simulation is therefore adopted during each major tuning cycle, with the former to survey the likely responses and the latter for confirmation along with assessment in greater detail once an educated set of parameter choice is selected . Results from the coarse-resolution tunings and PPE experiments are used as guidance for the high-resolution CAPT then AMIP tunings in terms of the sensitivity responses in direction and magnitude.

Browser not supported