Weather Research and Forecasting Model simulations of extended warm-season heavy precipitation episode over the US Southern Great Plains: data assimilation and microphysics sensitivity experiments

Abstract

This study examines eight microphysics schemes (Lin, WSM5, Eta, WSM6, Goddard, Thompson, WDM5, WDM6) in the Advanced Research Weather Research and Forecasting Model (WRF-ARW) for their reproduction of observed strong convection over the US Southern Great Plains (SGP) for three heavy precipitation events of 27–31 May 2001. It also assesses how observational analysis nudging (OBNUD), three-dimensional (3DVAR) and four-dimensional variational (4DVAR) data assimilation (DA) affect simulated cloud properties relative to simulations with no DA (CNTRL). Primary evaluation data were cloud radar reflectivity measurements by the millimetre cloud radar (MMCR) at the Central Facility (CF) of the SGP site of the ARM Climate Research Facility (ACRF). All WRF-ARW microphysics simulations reproduce the intensity and vertical structure of the first two major MMCR-observed storms, although the first simulated storm initiates a few hours earlier than observed. Of three organised convective events, the model best identifies the timing and vertical structure of the second storm more than 50 hours into the simulation. For this well-simulated cloud structure, simulated reflectivities are close to the observed counterparts in the mid- and upper troposphere, and only overestimate observed cloud radar reflectivity in the lower troposphere by less than 10 dBZ. Based on relative measures of skill, no single microphysics scheme excels in all aspects, although the WDM schemes show much-improved frequency bias scores (FBSs) in the lower troposphere for a range of reflectivity thresholds. The WDM6 scheme has improved FBSs and high simulated-observed reflectivity correlations in the lower troposphere, likely due to its large production of liquid water immediately below the melting level. Of all the DA experiments, 3DVAR has the lowest mean errors (MEs) and root mean-squared errors (RMSEs), although both the 3DVAR and 4DVAR simulations reduced noticeably the MEs for seven of eight microphysics schemes relative to CNTRL. Lower-tropospheric θ e and convective available potential energy (CAPE) also are closer to the observations for the 4DVAR than CNTRL simulations.

Keywords

• data assimilation
• microphysics sensitivity
• WRF evaluation
• MMCR-equivalent reflectivity simulation
• Southern Great Plains