Preprint: 18th Conference on Severe Local Storms, Amer. Met. Soc.
San Franciscon, CA, Feburary, 1996

PREDICTION AND SIMULATION OF A SQUALL LINE CASE DURING VORTEX-95

Ming Xue1, Kelvin K. Dreogemeier1,2, Donghai Wang1, and Keith Brewster1,2

Center for Analysis and Prediction of Storms1
School of Meteorology2
University of Oklahoma
Norman, OK, 73019

1. INTRODUCTION

Convective cells are frequently observed to form in lines that can extend hundreds of kilometers in length and last for many hours. Such organized lines of convection are known as squall lines. Squall lines, together with the mesoscale convective complexes (MCC), account for a large portion of the warm season precipitation over the central and eastern United States and severe weather events during the spring and summer season (e.g., Fritsch et al., 1981).

Squall lines have been studied extensively for many years, both observationally (e.g., Newton, 1950; Smull and Houze, 1985) and numerically (e.g., Moncreiff and Miller, 1976; Thorpe et al., 1978; Rotunno et al., 1988). Much understanding has been gained about their structure and dynamics from these studies, and the simulations of such systems in mesoscale numerical models have also been made. Using a nested-grid meoscale model, Zhang et al. (1989) were able to simulate most of the mesoscale features associated with a PRE-STORM squall line, including an accurate prediction of its time and location. However, due to the relatively coarse resolution (25 km on the nested grid), the model could not explicitly resolve convection.

During the past a few years, a model known as the Advanced Regional Prediction System (ARPS) has been developed at the Center for Analysis and Prediction of Storms (ARPS, Xue et al., 1995) at the University of Oklahoma. The eventual goal is for it to serve as a tool for operational storm-scale weather prediction. Towards achieving this goal, ARPS was run daily during the spring 1995 in real time using 3 km maximum resolution, and this was the first attempt to predict convective storms and systems explicitly in real time (Droegemeier et al., 1996; Xue et al., 1996, referred to as X96 hereafter). This experiment was conducted in collaboration with a field experiment known as VORTEX-95 (Rasmussen et al., 1994).

A prominent case during this experiment was that of the May 7, 1995 multi-squall lines. Two intense squall lines occurred on that day, and the operational prediction of ARPS did reasonably well. The 3-km resolution grid, which was one-way nested within a 15-km grid, managed to capture both of these lines. At the end of the hours, both predicted lines were within 50 km of the actual location on radar. However, the model under-predicted the southern portion of the second line and delayed their initiation by as much as 3 hours. This is not surprising considering the model was "cold started" from the 6-h NMC RUC (Rapid Update Cycle) forecast valid at 18 UTC; at that time both lines were already formed and the ARPS had no information on storm-scale structure. The forecast for this case has been summarized in X96 as one of the ARPS/VORTEX forecast examples.

In an attempt to better understand the initiation and the evolution of these two squall lines, and to validate the model, we conducted a number of simulation experiments, and report here the preliminary results.

Figure 1. Analyzed surface dew-point temperature (deg.C, interval 2.0) and surface wind (ms-1) at 18 UTC, May 07, 1995. Negative contours are dashed.

2. MAY 7, 1995 SQUALL LINES

On May 7, 1995, the surface analysis at 18 UTC (Fig. 1) was characterized by a distinct north-south oriented dryline extending from the mid-Texas (TX) panhandle to southwestern TX. This dryline was located on the western TX border 3 hours earlier. Associated with the dryline were strong southeasterly winds on the east side and southwesterly winds on the west. This region lay ahead of a deep upper-level low centered west of the Rockies, with weak troughs moving northeastward from the southwest. Deep moist air was transported northward by the flow east of the dryline, and the flow evolved into a 20 to 25 ms-1 low-level jet by the evening (local time).

Two main lines of thunderstorms occurred during the day. In the early morning, storms which had originated in eastern New Mexico (NM) moved northeastward across the TX and OK panhandles and joined up with a line of cells ahead of the dryline in western TX. By 15 UTC, these two clusters had already become a connected line in western OK and northern TX, and continued to move eastward into central OK. A storm at the southern end of this line broke off, became an isolated supercell later in northern TX, and spawned tornadoes, including a large one near Ardmore, OK between 21 and 21:30 UTC. The remainder of this first line dissipated soon after it moved into Missouri.
Figure 2. Radar reflectivity fields at 18 UTC, May 7, and 00 UTC, May 8, 1995, showing two squall lines in the western Texas panhandle and western Oklahoma.

The second line had more precipitation and was longer lived. Convective cells started to appear along a north-south axis through the middle of OK panhandle, just east of Amarillo, Lubbock, and Midland, TX around 17 UTC. This axis lay slightly east of the surface dryline, and was apparently associated with the convergence zone along the dryline. These cells intensified during the next hour and became a nearly continuous line by 18 UTC (Fig. 2a). The line moving eastward at approximately 10 ms-1, while individual cells in the line moved north-northeastward. A rotating supercell was prominent in the middle portion of this line at around 21 UTC while the line was at the western OK border. This supercell had been targeted and tracked by the VORTEX field operation team but no tornado formed. The line continued to travel eastward and evolved into a classic solid squall line with a trailing stratiform precipitation region that is evident in the radar reflectivity field (Fig. 2b). This squall line extended more than 1000 km and lasted for more than 10 hours while it moved across OK and into MO late in the evening. The dryline stayed within the TX panhandle at all times.

3. SIMULATION EXPERIMENTS

Version 4.0 of the ARPS (Xue et al., 1995) was used to perform the simulation study. The ARPS is a 3-D non-hydrostatic model designed specifically for storm scale weather prediction. Similar to the setup used during the real-time operational prediction experiment in spring 1995 (Droegemeier et al., 1996; X96), two grids were used together in a one-way nested mode. The coarse grid had either a 15-km or 7.5-km horizontal grid spacing, and the fine grid had a 3-km grid spacing. 35 levels were used in the vertical, among which 13 were within the lowest 2 km. The model used an analyzed terrain that is consistent with that of the 60-km resolution RUC. The same physics were used on both grids, including Kessler warm-rain microphysics, 1.5-order TKE-based subgrid scale turbulence, and a soil model coupled with surface energy budget equations. Ground surface temperature and moisture were explicitly predicted and stability and roughness length-dependent surface fluxes were calculated. The ground temperature was initialized by empirically relating it to the initial surface air temperature. The initial volumetric moisture content in the surface soil moisture was assumed to be 70% of the saturated moisture, which depends on the soil type that was analyzed from a surface characteristics database.

Different from the operational experiment, we started the simulation at 15 UTC because the western squall line was initiated at around 17 UTC. Using this earlier start time, we hoped to be able to simulate the storm initiation process. Also different from the operational runs, we used analyzed fields instead of the RUC forecasts as the initial and boundary conditions. The model was again "cold started" from a single time-level analysis and forced at the lateral boundaries using a Davies (1983) type relaxation technique.

An objective analysis system has been developed at CAPS to analyze multiple data sources directly on the ARPS grid. The spatial analysis is based on the Bratseth (1986) scheme, which is an iterative approach to statistical interpolation. The observation correlation was modeled by Gaussian functions of distance in the horizontal and vertical. The NMC RUC (Rapid Update Cycle) operational analyses were used as the background. Data from standard surface hourly observations, the Oklahoma Mesonet and the Wind Profiler Demonstration Network were used. Special mobile rawinsondes from the VORTEX experiment, and WSR-88D Doppler winds are now available, but were not used in these experiments. The surface and near-surface fields were most impacted by the addition of the mesonet data.

Starting at 15 UTC, we ran the ARPS for 9 hours on a 1200x1200 km2 domain shown in Fig. 1. Two resolutions, namely, 15 km and 7.5 km, were used for this domain. We then interpolated the 7.5-km solutions at hourly intervals to a 672x504 km2 3-km resolution grid, and used them to force the 3-km boundary. This 3-km grid was 224x168 km2 in size and was chosen to cover the TX panhandle and western OK, the regions where the squall lines initiated and moved through. Grid spacing of 3 km was chosen as a compromise between using the commonly accepted cloud-resolving resolution (1-2 km) and having a reasonably large domain. It was hoped that this grid could explicitly resolve convective storms. The 15- or 7.5-km grid was used to provide a set of dynamically consistent boundary conditions for the 3-km grid and at the same time resolve meso-[[beta]] scale features over a bigger domain.

4. RESULTS

In the following, we present results at the end of the 9-hour simulation, which corresponds to 00 UTC on May 8 (19:00 CDT on May 7). In Fig. 3, contours of surface dew-point temperature, Td, are plotted as thin lines. The tight gradient in western TX indicates the surface location of the dryline. This dryline was located at the western TX boundary at 15 UTC and moved to the middle portion of the TX panhandle during the first 3 hours (see Fig.1). By 00 UTC, the analyzed dryline had moved very little, except for the northern part that moved from western end of OK panhandle to its center (not shown). The Td gradient in the analysis increased by about one half of its 18 UTC value. At 00 UTC, the southern end of the simulated dryline location agrees well with the analysis but the northern part did not move to the east as much as it should - it stayed at the western end of the OK panhandle (Fig. 3). This error, as will be seen later, seemed to have contributed to the position error of the squall lines in the 3-km runs. Further experiments are being conducted to identify the source of this error.

The vertical velocity, w, at about 500 m AGL is shown by the dark contours in Fig. 3. The w maxima are found in south-central KS and in north-central TX. No connected updraft line is identifiable from the w field between these two maxima, however. Obviously this 15-km run did a poor job capturing the squall line, partly because of the use of explicit microphysics only. Explicit microphysics requires sufficient vertical lifting to produce condensation and possibly convection, but the lifting tends to be weak on a coarse-resolution grid. Adequate parameterization of convection is generally needed at this resolution. In fact, the 15-km operational prediction of ARPS that used the Kuo parameterization scheme performed better on this case, notably by producing most of the precipitation in the west-central OK (X96). Additional sensitivity experiments by Wong et al. (1996) also demonstrate the superiority of the cumulus parameterization over the Kessler (explicit) microphysics, mainly because the former can produce a quicker response to the mesoscale convergence, which seems to be the primary mechanism of squall line initiation in this case.
Figure 3. Predicted surface dew-point temperature (deg.C, thin contours, interval 2.0), wind (ms-1) and the vertical velocity (ms-1, thick dark contours, interval 0.1) at 500 m AGL level on the 15-km grid, valid at 00 UTC, May 08, 1995. Negative contours are dashed.

Figure 4. As Fig.3, except for the 7.5-km grid and the w contour interval is 0.25 ms-1. The rectangular box indicates the location of the nested 3-km grid.

Figure 4 depicts the same fields as Fig. 3 but for the 7.5-km resolution run. The overall surface flow pattern is similar to Fig. 3, but the Td gradient is tighter. The dryline location is also similar. The w field, however, indicates a much better simulation than the 15-km run. More specifically, the convective cells are connected through western OK, extending from south-central KS through western OK to western central TX. Compared to the radar reflectivity fields, this line location is about 100 km too far to the west in OK, but is much more accurate in TX and KS. The surface flow shows convergence along the convective line and the Td field shows two minima along the line, which are the result of dry air descending in the downdrafts. The drying signature is not clear with other convective cells in this line, suggesting that a significant spinup time is needed for the cells to develop full scale downdrafts. Presumably, as a result, the propagation of the line is slowed for the lack of a sufficient driving force from the cold pool.

The first convective line seen in the radar reflectivity field (Fig. 2a) is missing in this and the 15-km simulation may be because that the mesoscale forcing is not well captured in the initial analysis fields. Further, the ARPS was given no information from radar reflectivity or other fields on the storm scale. No clear convergence signature can be found in the initial surface wind analysis. Interestingly, the ARPS 3-km operational prediction that started from the RUC forecast field at 18 UTC (OLAPS analyses were used later as the initial conditions) was able to predict two convective lines. The positions of both lines were close to the observations (X96). (Further investigation of the initial fields is clearly necessary.) In Fig. 4, the convective cells near the northwest corner of the model grid were also observed, and they were associated with a surface cyclone centered at the central-south-west Colorado. The disturbances found at the western TX boundary in Fig. 4 are spurious, however.

The results of the 3-km simulation are presented in Fig. 5. As mentioned earlier, the 3-km grid covers the western TX and OK panhandle areas where the second squall line was initiated, and central and western OK, where the squall lines later moved. Similar to what was found in the ARPS/VORTEX operational experiment, the 3-km resolution resolves individual convective cells much better than the 7.5- or 15-km runs. The use of explicit microphysics is also more appropriate.

Again, the surface dryline position is good, except for its northern end. Figure 5a shows that, by 6 hours, lined convection has formed along the northern portion of the dryline, and it has apparently produced a cold pool between the convective line and the main Td gradient zone to the west. This cold outflow drove the convective line eastward. By 9 hours, this line had moved past the western OK border, while at the same time the storm cells within the line moved northeastward. The high w contours nearly coincide with the gust front depicted by the tight Td contours at the surface. While the initiation mechanism and the propagation speed of this portion of squall line seem to have been correctly simulated, its location is far from accurate. Between 6 and 9 hours, the line is almost 150 km too far to the west. Further, the squall line is about 3 hours too late. As mentioned earlier, the inaccurate position of the northern portion of the dryline may have contributed to part of this, and the delay may also be caused by the spinup time the model requires to initiate the storms. Both of the problems will be addressed in future experiments, by improving the prediction of the dryline location, performing diabatic initialization (using radar and satellite data), and retrieving temperature and winds from sequences of Doppler radar wind data in an assimilation cycle.
Figure 5. Predicted surface dew-point temperature (deg.C, thin contours, interval 2.0), wind (ms-1) and the vertical velocity (ms-1, thick dark contours, interval 0.5) at 500 m AGL level on the 3-km grid, valid at 21 UTC (a) and 00 UTC (b), May 08, 1995. Negative contours are dashed.

It was somewhat unfortunate that the storm initiation regions were located in a relatively data sparse area. The dense network of Oklahoma Mesonet covers only Oklahoma. However, additional sounding and surface data were collected in the VORTEX field experiment, and these data will be included in future analyses. It has also been noticed that the surface winds to the west of the dryline were not well predicted, presumably due to the crude soil model initialization scheme used here. Further validation and turning of parameters for the soil model and surface layer physics are likely to improve the simulation results. As mentioned earlier, this study represents a preliminary work that tries to simulate storm-scale phenomena and as the same time tries to validate certain aspects of the model in a real data setting. The validation of the model against idealized cases has been done and can be found in Xue et al. (1995).

Finally, it should be pointed out that south of this region in TX, the convective line (Fig. 5b) is actually in a good agreement with observations (Fig. 2b). This line was initiated after 6 hours, ahead of the southern portion of the dryline, and propagated eastward in the ARPS at the correct speed. The 3-km grid did not capture the first line either, presumably for the same reasons discussed earlier.

5. SUMMARY

We reported preliminary results from model simulations of the May 7, 1995 squall lines. The model simulated storm initiation along the dryline reasonable well, but the processes were significantly delayed, perhaps due the spin-up required because the radar-observed precipitation was not in the initial conditions. Part of the squall line location was in good agreement with the radar observations, but further improvement is clearly needed for it to be considered a good simulation. Additional data collected by the VORTEX field experiment will be used in future model initialization and validation. The use of two-way interactive nesting available in ARPS (Xue et al., 1993) is also desirable for improving the meso- and storm-scale interaction, and will be used in future experiments.

6. ACKNOWLEDGMENTS

This research was supported by NSF grant ATM91-20009 to the Center for Analysis and Prediction of Storms at the University of Oklahoma. The simulations were made at Pittsburgh Supercomputing Center on their Cray C90 and Cray T3D. Adwait Sathye and Yuhe Liu are acknowledged for their assistance with the Cray T3D.

7. REFERENCES

Bratseth, A.M., 1986: Statistical interpolation by means of successive correction. Tellus, 38A, 439-447.

Davies, H., 1983: Limitations of some common lateral boundary schemes used in regional NWP models. Mon. Wea. Rev. 111, 1002-1012.

Droegemeier, K. K., M. Xue, A. Sathye, K. Brewster, G. Bassett, J. Zhang, Y. Lui, M. Zou, A. Crook, V. Wong, R. Carpenter, and C. Mattocks, 1996: Real-time numerical prediction of storm-scale weather during VORTEX-95. Part I: Goals and methodology. This volume.

Fritsch, J. M., R. A. Maddox and A. G. Barnston, 1981: The character of mesoscale convective complex precipitation and its contribution to warm season rainfall in the U.S. Preprints. Fourth Conference on Hydrometeorology, Reno, Amer. Meteor. Soc., 94-99.

Moncreiff, M. W. and M. J. Miller, 1976: The dynamics and simulation of tropical squall line. Q. J. Roy. Meteor. Soc., 102, 372-394.

Newton, C. W., 1950: Structure and mechanism of the prefrontal squall line. J. Meteor., 7, 210-222.

Rasmussen, E. N., J. M. Straka, R. Davies-Jones, C. A. Doswell III, F. H. Carr, M. D. Eilts and D. R. MacGorman, 1994: Verification of the Origin of Rotation in Tornadoes Experiment (VORTEX). Bull. Amer. Met. Soc., 75, 995-1005.

Rotunno, R., J. Klemp, and M. L. Weisman, 1988: A theory for severe long-lasting squall lines. J. Atmos. Sci., 45, 463-485.

Smull, B. F., and R. A. Houze, Jr., 1985: A midlatitude squall line with a trailing region of stratiform rain: Radar and satellite observations. Mon. Wea. Rev. 113, 117-133.

Thorpe, A. J., M. J. Miller and M. W. Moncrieff, 1982: Two-dimensional convection in non-constant shear: a model of midlatitue squall line. Q. J. Roy. Meteor. Soc., 108, 739-762.

Wong , V., M. Xue, K. Droegemeier, Y. Liu, X. Song, J. Zhang and L. Zhao, 1996: Impact of physics on the development of severe storms during VORTEX-95. This volume.

Xue, M., K. K. Droegemeier, V. Wong, A. Shapiro, K. Brewster, 1995: ARPS Version 4.0 User's Guide. 380pp. Available from the Center for Analysis and Prediction of Storms, University of Oklahoma, Norman, OK, 73019.

Xue, M., K. Brewster, Droegemeier , K. K., F. Carr, V. Wong, Y. Liu, A. Sathye, G. Bassett, P. Janish, J. Levit, P. Bothwell, 1996 (X96): Real time prediction of storm-scale weather during VORTEX-95, Part II: Operation summary and example cases. This volume.

Xue, M., K. K. Droegemeier, and P. Woodward, 1993: Simulation of tornado vortices within a supercell storm using an adaptive grid refinement technique. Preprints, 17th Conference on Severe Local Storms. St. Louis, MO, Amer. Met. Soc., 362-365.

Zhang, D. L., K. Gao, and D. B. Parsons, 1989: Numerical simulation of an intense squall line during 10-11 June 1985 PRE-STORM. Part I: Model Verification, Mon. Wea. Rev., 117, 960-994.