Utilizing the Atlantic Tradewind Ocean-Atmosphere Mesoscale Interaction Campaign for Assessing Marine Shallow Cumulus Simulations in the Unified Forecast System: Insights from the Common Community Physics Package Single Column Model

I-Kuan Hu

CIRES CU Boulder - NOAA Physical Sciences Laboratory

Tuesday, May 21, 2024, 2:00 pm MT
DSRC Room GC402

JOIN THE WEBINAR:

Abstract

Employing the observational data from the Atlantic Tradewind Ocean-Atmosphere Mesoscale Interaction Campaign (ATOMIC), seven test cases are constructed for evaluating the simulations of marine shallow cumuli in the trades using various physics suites in single-column models (SCMs). These test cases are selected based on the collocation of NOAA’s research ship Ronald H. Brown (RHB) and research aircraft WP-3D (P3) Orion, which provide observations for boundary conditions (i.e., sea surface temperature and surface fluxes) and verification of the initial conditions and large-scale advective tendencies ("forcings") constructed from the fifth generation ECMWF atmospheric reanalysis (ERA5) data for driving SCMs. Three physics suites in the Common Community Physics Package Single Column Model (CCPP SCM), which includes all the physics suites available in the Unified Forecast System, are evaluated: Global Forecast System (GFS) with minor and major updates (GFSv16+ and GFSv17HR2, respectively) and the Rapid Refresh Forecast System (RRFS). The corresponding Large Eddy Simulations (LES) are conducted for bridging the gap of observational and CCPP SCM output data. The model biases are presented through a statistical lens, and the simulated cases are categorized into cloudy and clear-sky groups by a simple metric based on the observed cloud fraction. All models exhibit the “drizzling” bias (i.e. overestimation of light precipitation occurrence) particularly in the cloudy group. Regarding cloud fraction, which governs the surface radiation, the LES outperforms all the CCPP SCM runs. The comparison of the GFSv16+ and GFSv17HR2 runs implies that the zero-or-one bimodal cloud fraction in the GFSv16+ run stems from the omission of convective condensates in cloud fraction calculations, and the overestimation of samples with cloud fraction being equal to one in the RRFS runs is partially caused by an assumed boosting effect of high relative humidity on cloud fraction. In terms of the levels of turbulent kinetic energy peaks, the LES and CCPP SCMs mismatch the observations differently in the cloudy group, while the LES agrees fairly well with the observations in the clear-sky group. Moreover, in all the CCPP SCM runs the surface heat fluxes are overestimated primarily due to a cold- and dry-biased boundary layer. Prescribing the surface heat fluxes from the observations worsens the cold and dry biases of the boundary layer and marginally changes other aspects of the results, implying that the interactive surface heat fluxes are more indicative of responses rather than stimuli in the CCPP SCM simulations. Noteworthily, the model biases exhibit variations among cases, with biases in certain cases differing from their respective cloudy or clear-sky statistics. This offers opportunities to explore the differences in performance and reasons behind the simulations of SCMs and LES across different marine shallow convective patterns in the future.