The research conducted under this fellowship (1 September 1996 - 31 August 1999; no-cost extension from 1 September 1999 - 31 October 2000) focused on the evolution of linear mesoscale convective systems (MCSs) that often produce severe weather in the Midwest, or more specifically, the Mid-Mississippi Valley region. The research was conducted with the cooperation of the St. Louis National Weather Service (NWS) office in Weldon Spring, Missouri (KLSX) under the direction of the Science and Operations Officer, Mr. Ron Przybylinski, and the Department of Earth and Atmospheric Sciences at Saint Louis University (SLU). Dr. Yeong-Jer Lin served as the advisor to Mr. Jim O'Sullivan at SLU during the course of this research.
Linear MCSs, or squall lines, are important phenomena to understand from a forecasting perspective in this region of the country. They are responsible for large swaths of damaging straight-line winds, hail, and often non-supercell tornadoes. There are a number of storm outflow interactions that occur with squall lines. The warning process of forecasting these systems if very challenging. Therefore, one of the major goals of the research was to increase the operational forecasters' understanding of these events. Most of the previous research done in this area has been performed during warm-season events. Most cases of linear MCSs do occur during the late spring and early summer months (May through July). However, this project was to primarily focus on cooler season events, from October through April.
Doppler radar data are an important tool in the real-time forecasting of squall lines. Therefore, it was to play an important role in examining the features of the linear MCSs that occurred within the county warning area (CWA) of the St. Louis NWS office, and several adjoining offices, such as Paducah, Kentucky (KPAH). Data from both the WSR-88D Principal User Processor (PUP) and the WSR-88D Algorithm Testing and Display Systems (WATADS) was planned to be used in the study. The former is the standard display that the forecasters in the NWS have to use operationally. This is the device that had been used (until recently) during any severe weather events. Examining the events on the PUP gauged what the operational forecaster used to develop their forecasts during the events chosen. The latter software had been standard issue at NWS offices for some time at the onset of this project, and was installed on the workstations at SLU. It uses archive Level II data from the National Severe Storms Laboratory (NSSL), where the package was also developed.
An important tool that was used in this research was the incorporation of a mesoscale model into the forecasting and warning procedures at the St. Louis NWS office. The Pennsylvania State / National Center for Atmospheric Research (NCAR) Mesoscale Model (MM5) was chosen as the model to be used during the course of this investigation. The Mesoscale and Microscale Meteorology Division of NCAR maintains the model, but MM5 is a public model used by many institutions (academic, private, and government) around the world. There are many options, which users can adjust for their specific research, which makes the model highly adaptable to a variety of conditions. The MM5 had been used in many successful studied before the start of this research, and it was felt that this model could best assist in the forecasting of linear MCSs. Also, the model would be used as a diagnostic tool to investigate the kinematical and dynamical structure of the linear MCSs investigated during the course of this project. Another great benefit of using the model at the onset of the study was its portability to a number of different platforms, including various workstations that could (in 1996) be found in universities and NWS offices.
Collectively, the goals of the research was to gain a better understanding of the squall line structure, mesoscale airflow and vortex signatures, and to aid the forecasters by identifying the preferred regions within the overall squall line which may produce damaging straight-line winds and non-supercell tornadoes.
Much of the early interactions between the forecasters at the NWS office and the researchers at SLU dealt with the continuing investigations into mid-level mesoscale interactions within squall lines, begun before the start of this project. Identifying Mid-Altitude Radial Convergence (MARC) signatures within linear MCSs has shown to be a good predictor in the now-casting of severe weather that is associated with these events. Examinations of several cases pre-selected in this study (listed below) were performed to detect MARC values in order to collect more data on this important signature.
Much of the initial research and model runs were done at KLSX during 1997 and early 1998. The Central Region of the NWS obtained a workstation for the primary purpose of running the MM5 model and to store WATADS datasets for this project. One of the first goals of the investigation was to get the MM5 operational at the office. This proved to take longer than anticipated, and was delayed for a short time in order to get more disk space to allow the machine to both run the model daily, and to keep archives of model and radar data. A great amount of assistance was given during these early months by Mr. David Bright of the NWS office in Tucson, Arizona, who provided several personal programs developed to assist the preprocessing of the model on the Hewlett-Packard machines used by the NWS. It was soon discovered that the model could only be run in an operational mode using one domain of 27-km resolution due to the speed of the Hewlett-Packard system, which were found to be on the slow end of benchmark testing for the MM5. However, this setup did provide useful for several linear MCS events that occurred from 1998-2000. Also, since it was run each day (except on those days when the Early Eta data did not arrive on time), the model provided another dataset that forecasters used in everyday operations. All MM5 output was transferred to the workstation in the operational area of the office, and could be accessed by a forecaster using the GARP software package, which makes use of GEMPAK for the analyses.
During the latter months of the project, the operational model at the NWS office in St. Louis was transferred to a multi-processor computer, which could run the model faster and more efficiently that the older workstation. Also, the output from MM5 is now shown using the newer, integrated AWIPS workstations that have become standard at NWS offices.
When the project began, there were several events that were already chosen for inclusion in this investigation. These included including 15 April 1994, 21 November 1994, 19 April 1996, 5 May 1996, and 27 May 1996. During the course of the research, several other events occurred, on 28 March 1998, 11 February 1999, and 27 May 2000. Each case was examined closely. Certain cases were initially focused upon due to the availability and quality of the radar data for the event. In particular, the 15 April 1994, 27 May 1996, 11 February 1999, and 27 May 2000 cases looked very appealing for further examination and detailed numerical modeling. Each case mentioned above was simulated using the MM5. The initialization data for the model runs came from the Early Eta model produced by the National Centers for Environmental Prediction (NCEP). The exception to this were the two 1994 cases. The Early Eta model's usefulness to create boundary conditions for a mesoscale model at this time was somewhat limited by its relatively course (80 km) grid spacing. For these events, the initialization fields and boundary conditions were obtained using the NCEP Global Tropospheric Analyses. These are hemispheric datasets available for 0000 and 1200 UTC daily. These datasets are available for download by NCAR account users. However, not all the cases could be successfully modeled. The primary reason for the failure of several cases to be successfully predicted (and often simulated) was the time of initialization of the system in the model's forecast. It had been hoped that the locally-run model could be used to forecast events, which began up to 18 hours after model initialization. However, this soon proved to be slightly beyond the capability of the model. Without additional mesoscale data incorporated into the model's early forecast hours (which is beyond the scope of this project), the model was not able to maintain the proper mesoscale forcing needed to sustain (and in some cases develop) the liner MCS system. The one notable exception to this discovery is the 11 February 1999, where the model was able to do a fair job of simulating the linear aspect of the system. However, the placement and timing the modeled event were incorrect. In this case, the simulation of the event was decent, but the model's ability to forecast the timing and location of the event was not as good as expected.
It appears that the useful limit to the forecasting of the initiation of an organized linear MCS system with the parameters used in the local running of the model would be 12 hours. Beyond this limit, the MM5 model's ability to organize the proper mesoscale conditions from its larger-scale environment is decreased significantly. With this fact in mind, and with the faster computing now available at NWS offices, model runs will be started more than once a day. This will provide the forecasters with more updated short-term mesoscale predictions to use in their operations.
The MM5 model does a very good job at predicting the necessary larger-scale environment that is favorable to the initiation of linear MCSs. Also, on many of the cases investigated, these forecasts were very similar to the actual observations for the events. The 15 April 1994 case is an excellent example of how well the MM5 model can simulate a squall line given only minimum, initial conditions. The event also "began" only 6-8 hours into the forecast, so the simulation was within the mesoscale "window of opportunity."
The pre-convective environment was very similar to the observations. The necessary moisture levels was present over the area of interest, and the model developed the larger-scale frontal zone that was seen for this date. While the model was about two hours late in developing the system, the forecast is a very good one. Model simulated reflectivity values during the lifetime of the model-generated squall line were compared to Doppler radar observations examined with the aid of the NWS staff present during the event on 15 April 1994. The agreement is excellent. The model is successfully able to simulate two distinct line segments associated with the leading edge of the system. This agrees with what was observed at the time of the event.
The model's simulation of the squall line's southern segment is astonishing. The model was able to create and maintain the signature of a line-echo wave pattern (LEWP) caused by the strong, descending mid-level winds forcing an acceleration of the low-level and near-surface winds behind the leading edge of the system. The result is a bowing of the southern segment. This was seen in the radar observations of KLSX. Plan views of low- and mid-levels over this segment show the model is able to predict several regions of intense updrafts behind the leading edge. Several of these updrafts are seen to be rotating based on vertical vorticity values. Two such rotating updrafts at elevations of 5 km exceed mescocyclone criteria. The agreement with the examined radar data is quite striking. While the model simulations used for this project could not reveal more detailed structure, it is possible that given the correct special and temporal resolution, it is possible that greater signatures for possibly imminent severe weather can be seen. Cross sections of the southern line segment reveal that the model creates a region of very rapid multicellular evolution along the northern flank, and more vertical (but not supercellular) evolution along the leading edge of the system. This also agrees with the radar observations. Overall, the MM5 model performed a very accurate prediction of this event.
There are several other cases studied which will warrant further examination. There are plans to begin a more detailed study of the 27 May 1996 and 27 May 2000 cases this summer or fall using the resources at SLU and/or the St. Louis NWS office. The 27 May 1996 case is of particular interest since it was one of the initial cases examined for MARC signatures during the early stages of the project. Also, the 15 April 1994 event will be modeled with more detail than was possible during the course of this research project. Greater temporal resolution (20-30 minutes) and special resolution (down from 3 km to 1 km) will be given to the simulation. The results should provide the researchers with more a more detailed dynamical and kinematic structure of what is already a finely-predicted system with the MM5, and will provide more information on the evolution of linear MCSs that occur in the Mid-Mississippi Valley during the cool season and spring and early summer. It is hoped that this research will begin late in the summer of 2001 with the aid of other graduate students at SLU and the staff of the NWS office.
The research between SLU and the St. Louis NWS office staff has already produced several papers at local and national conferences. A two-part publication on the observations and model simulation of the 15 April 1994 case is being prepared at this time, and it is anticipated that a third article on forecasting aspects of the model forecast of the 11 February 1999 event will be written in the near future.
The work between the NWS office and SLU conducted under the auspices of this project (as well as several indirect collaborations stemming from this research) has provided the forecasters with more detailed information concerning the prediction of linear MCSs using a mesoscale model. The MM5 model will continue to be maintained at the office, under the supervision of the forecasting staff. In addition, the researchers at SLU have gained more insight into the basic structure of squall lines form the model and radar output. The radar data examined also has allowed the forecasters at the NWS to collect more statistics on the frequency and timing of severe weather associated with these events. Future work being planned will continue the findings and conclusions from this project.