Assimilating Surface and Rawinsonde Data in WRF Microphysics Simulations of Warm-season Convection for SGP Central Facility

Zewdu Segele CIMMS/University of Oklahoma
Lance Leslie Ohio University
Peter Lamb University of Oklahoma

Category: Modeling

Working Group: Cloud-Aerosol-Precipitation Interaction

Image Thumbnail
Probability distributions of correlations between observed (CMBE) and simulated precipitable water vapor for eight WRF microphysics simulations with (colors) and without (black) data assimilation for 27–31 May 2001 warm-season heavy precipitation event over the SGP. Correlations are computed for the inner nested domain within 7x7 (solid lines) and 35x35 (dashed lines) grid boxes surrounding SGP CF. CNTRL, OBNUD, SFDDA, 3DVAR, and 4DVAR denote no data assimilation, observation nudging, surface analysis nudging, three-dimensional variational data assimilation, and four-dimensional variational data assimilation, respectively.

Several high-resolution (3-km) nested simulations were performed to examine the ability of the Weather Research and Forecasting (WRF) model microphysics schemes to reproduce the observed cloud properties and convection characteristics in the vicinity of the SGP Central Facility for the warm-season heavy precipitation case of May 27–31, 2001. The results showed substantial differences in simulated cloud water content, cloud ice concentration, and reflectivity profiles at the SGP Central Facility for all WRF microphysical parameterizations. To minimize differences in observed and simulated convection onset and cloud microphysical properties, SGP surface and SGP CF upper-air sounding data were assimilated in SGP WRF microphysics experiments. Accordingly, simulations without data assimilation (CNTRL) for eight WRF microphysics schemes were compared to corresponding simulations with surface analysis nudging (SFDDA), upper air observation nudging (OBSNUD), three-dimensional variational data assimilation (3DVAR), and four-dimensional variational data assimilation (4DVAR). Evaluation of the performance of the data assimilation experiments involved direct comparisons of simulated grid point values and observed CMBE and Mace et al.’s (2006) best estimates of SGP CF atmospheric state/cloud properties. To allow for a possible spatial grid point mismatch between simulated and observed variables, model performance also was assessed using the probability distributions of the correlations between observed atmospheric state variables and corresponding simulated values of the inner nested domain within 7x7 and 35x35 grid boxes surrounding SGP CF. The correlation analysis reveals that the 3DVAR/4DVAR experiments showed improved reproduction of the observed precipitable water vapor for Lin et al., WRF Single Moment 6-class (WSM6), WRF Double Moment 5-class (WDM5), and WRF Double Moment 6-class (WDM6) microphysics schemes, with the 4DVAR WDM6 simulation achieving the highest modal correlation exceeding +0.9. Compared to the CNTRL experiment, both the 3DVAR and 4DVAR runs perform poorly for the Eta and Thompson microphysics schemes because of increased convection early in the simulations. Although the SFDDA simulations produced increased LWC/IWC that compared favorably with the observed values, the correlation between observed and simulated PWV showed no improvement for many of the microphysics runs. Results of model sensitivity experiments to radiation schemes also will be presented.

This poster will be displayed at ASR Science Team Meeting.

POSTER in PDF: for proper viewing, it should be viewed with Adobe Acrobat Reader. Download the latest version from the Adobe Reader website.