Meetings > PI / Science Team Meeting > Posters

Improving cloud microphysics simulations for the SGP from four-dimensional variational data assimilation (4DVAR)

 

Authors

Zewdu Tessema Segele — CIMMS/University of Oklahoma
Lance Leslie — Ohio University
Peter J. Lamb — University of Oklahoma

Category

Modeling

Description

Figure 1. Observed (left) and simulated (right) reflectivity and thermodynamic structure for the SGP CF for 2300 UTC 27 May–0100 UTC 28 May 2001. (a) KVNX WSR-88D reflectivity, showing the nested WRF domain and the orientation of transects for eight microphysics schemes. (b) MMCR reflectivity (shading) and rawinsonde θe (contour, K). (c) Saturated moist static energy deviations (hs, contours, K) and buoyancy perturbations (B, shading) from time-averaged (TA) reference profiles (RP). (d) MMCR raw reflectivity, showing complete attenuation. (e) Observed and simulated Ka-band reflectivities for 4DVAR runs and their averages along with those for CNTRL. (f) RH differences between the 4DVAR and CNTRL simulations. (g) Same as (f) except for B. (h) Vertical transects along the lines in (a) of 4DVAR Ka-band reflectivities and θe for selected schemes (top and third rows), and transects of B (shading) from area-averaged RP, and hs (contours) from TARP (second and bottom rows). All vertical transects in (h) are 3-hr averages (arrows in b). Ka-band cloud radar reflectivities were computed according to the particle size distribution specifications of the eight microphysics schemes.

This study evaluates the improvements from a 4DVAR data assimilation of SGP surface and SGP Central Facility (CF) rawinsonde observations in simulating deep convection in eight WRF microphysics schemes, relative to control simulations without data assimilation (CNTRL), for the SGP warm-season heavy precipitation event of May 27–31, 2001. Because convection tends to develop in a region of high water vapor content, the 95 percent of maximum microphysics-simulated water vapor mixing ratio at 5.5 km (level of maximum observed cloud reflectivity) is used to identify objectively, through linear regression, the location and orientation of simulated convective cloud structure (Figure 1, top left) in the vicinity of the SGP CF. Ka-band cloud radar reflectivity was computed by using particle size distributions employed in the eight WRF cloud microphysics schemes. In-cloud thermodynamic parameters—such as saturated moist static energy, buoyancy, equivalent potential temperature (θe), convective available potential energy (CAPE), and convection inhibition (CIN)—were computed, and vertical transects across the SGP CF were compared with observations. To maximize the utility of the observed millimeter-wavelength cloud radar (MMCR) reflectivity, the analysis was performed for the three hours immediately preceding the intense convection at the SGP CF. During the intense convection period, the MMCR suffered severe attenuation (Figure 1).

Compared with CNTRL, the 4DVAR experiments show marked improvement in the simulated Ka-band cloud radar reflectivity in the vicinity of the SGP CF (Figure 1e). This improvement is associated with enhanced lower- and mid-tropospheric buoyancy and higher near-surface and mid-tropospheric relative humidity of 4DVAR compared with CNTRL. Because of drier air at the top of a capping inversion in the 4DVAR simulations, the simulated CAPE values are lower than CNTRL. However, the CNTRL simulations have higher CIN compared with 4DVAR. Although all simulations under-predicted the observed high θe below 3 km at 0000 UTC 28 May, saturated moist static energy deviations from time-averaged reference profiles compared well with similarly computed rawinsonde counterparts, especially for the Thompson microphysics scheme.