Forecast simulations in a multi-scale modeling framework: maximizing the use of high-value observations

Description

The conventional approaches to evaluating global climate models (GCMs) against heavily averaged and sub-sampled observations cannot fully take advantage of high-value intermittent measurements taken during well-instrumented field campaigns or at fixed sites, such as the ARM SGP site. Such composite analysis makes untangling the causes of model deficiencies and assessing simulated variability, timing, and intensity of individual processes very difficult. An alternative approach, following the Cloud-Associated Parameterizations Testbed (CAPT) methodology, offers an improved perspective. The approach applies evaluation techniques previously limited to weather prediction models by using GCMs to simulate short weather periods initialized with observed atmospheric states—forecasts simulations. The CAPT approach is applied here to a super-parameterized GCM, also known as a multi-scale modeling framework (MMF), in which two-dimensional cloud-resolving models (CRMs) are embedded in each grid cell of a GCM to explicitly represent sub-grid convection and replace conventional cloud and boundary layer parameterizations. Over the Central U.S., the MMF in free-running mode yields an improved representation of nocturnal summer organized convection, but the mechanisms that allow this are poorly understood. CAPT forecasts offer a better vantage point to understand advantages of the multi-scale modeling approach at the process level, at the ARM SGP site, to inform future conventional parameterization development. However, running an MMF in forecast mode introduces a new challenge not previously faced by the CAPT community—initializing a high-resolution internal CRM. This poster presents a method to overcome this challenge by spinning the model up in an observationally constrained mode through nudging to generate forecasts’ initial conditions at both the GCM and CRM scales.