The Path to A Well-Tuned High-Resolution ACME V1 Atmosphere Model and Initial Results

Monday, June 5, 2017 - 15:30
Add to Calendar

The atmosphere group has delivered a reasonably well-tuned high-resolution ACME V1 atmosphere model. This presentation is to document the pathway from the best well-tuned low resolution configuration, which however exhibits large biases at high resolution if without further tuning. A strategy heavily relying on the Cloud Associated Parameterization Testbed (CAPT) framework is adopted for the tuning at high resolution. As described in previous report, it is efficient in terms of computational cost and effective in terms of identifying and reducing model biases. The physical parameter tuning is guided by the Perturbed Parameter Experiments (PPE) at low-resolution, with successive fine tuning to narrow the perturbation range at high resolution using CAPT-type short-term hindcasts. High number of short-term hindcasts were performed, involving one or more parameters at a time, to gauge the sensitivity response. The best configuration is selected after taking into account overall model performance in various metrics at both global and regional scales.  It produces consistent performance in both short-term hindcasts and climate simulations, with respect to other configurations. Initial results in cloud, precipitation, radiation, and large-scale thermodynamics will be shown, in comparison with observations and low-resolution simulations. The performance of high-resolution model is comparable to that of low-resolution in all essential metrics, with some notable improvements. The CAPT based tuning strategy can be extended to include more critical metrics, such as applicable variability characteristics and important regional features, and automated in continuing development of the ACME model to speed up the process and optimize the performance.