Evaluation of Global Model Physics Advancements using a Hierarchical Framework

The Global Model Test Bed (GMTB) was established in 2015 within the Developmental Testbed Center (DTC) in response to the request from NWS NGGPS (Next Generation Global Prediction System) Program Office. One of the initial foci of GMTB is to develop a common infrastructure for testing physics advancements for future use in operational NWP using state-of-the-art verification tools. A key aspect of the GMTB infrastructure is a hierarchical testing workflow. GMTB has developed a single column model for testing the physics response to prescribed forcing, and has adopted the NCEP Global Forecast System’s (GFS) workflow, which is an end-to-end system including preprocessing, the Global Data Assimilation System (GDAS), the forecast model [currently the NOAA Environmental Modeling System (NEMS)-based Global Spectral Model which is being transitioned to the NEMS-based FV3], and post-processing, as well as the ability to pull archived operational datasets and other input files from the NOAA’s High-Performance Storage System (HPSS). The GMTB workflow is also equipped with generating forecast graphics, performing statistical assessment of forecasts using the DTC’s Model Evaluation Tools (MET), and process-oriented diagnostic tools to identify the model’s response in aspects of cloud, radiation, precipitation, global water budget and tropical cyclogenesis forecasts to changes in model physics. Through the GMTB physics testing framework, model physics developers can evaluate new developments in a near-operational environment consequently producing testing results that are more relevant for consideration during the decision-making processes at operational centers.

This presentation will showcase the capabilities of the GMTB physics testing framework through an experiment designed to assess the performance of the GFS using an alternate convective scheme. The single column model was configured to run two cases including maritime deep convection and continental deep convection, respectively. Global retrospective runs were conducted to compare the GFS FY17 configuration, which employed the scale-aware Simplified Arakawa-Schubert (Han et al. 2017), against an experimental configuration using the untuned Grell and Freitas (Grell and Freitas, 2014) scale- and aerosol-aware convective parameterization. Cycled data assimilation was employed to produce initial conditions consistent with the model physics. Additionally, a set of GF cold-started runs was produced using the GFS FY17 for an assessment of the impact of cycling. Given the limited computational resources available to GMTB and consistent with the concept of hierarchical testing, the global experiments were run at a resolution of T574 (~34 km) and restricted to two weeks (June 1-15, 2016). This presentation will summarize the results of the single column model, and focus on the objective verification of the forecast performance of the three global configurations and some in-depth diagnostics of a case study initialized at 00 UTC on 10 June 2016.