Close Window

North Carolina State University (Devin Kramer, COMET graduate student fellow): "Real-time mesoscale modeling of Carolina frontal weather"

Final Report

(Excerpted from the report)

1. Evolution of Project Goals

This final report describes the evolution over the two year study period of the original goals set forth in our COMET/NWS fellowship proposal submitted 1 March 1995:

Some modifications were made to the above goals in an attempt to make a more substantive contribution to previous and on-going research at the Raleigh NWSFO and address specific forecasting issue in the region. The initial goals of the study provided a general framework for the introduction and use of a real-time mesoscale modeling system at the Raleigh NWSFO. However, once the modeling system was put in place and running reliably, we focused the remainder of the study on addressing the possible benefits that mesoscale model guidance could have on operational forecasting in the Southeast Cold Air Damming (CAD) environment.

CAD events are of extreme importance to the operational forecasters, as they result in some of the most serious winter weather across the Southeast as well as along the entire Eastern Seaboard. They also present some of the most complex mesoscale forecasting challenges to operational forecasters and are often times noted in the literature and by operational forecasters as being poorly handled by synoptic scale model guidance. CAD environments are also oftentimes associated with complex frontal development, which is unique to the Eastern U.S. and particularly the Carolinas. These features are primarily forced by the complex topographical variations that exist in this region. For these reasons, the original goals of predicting convective storm type, developing predictor fields for tornadic activity, and comparing model frontal structure to those observed by the WSR-88D were not specifically pursued in this study. The focused examination of CAD episodes also seemed prudent given the substantial amount of research by Raleigh NWSFO staff and by others in this region addressing winter season forecasting issues. It is important, however, to note that as a result of the development of a stable quasi-operational real-time mesoscale modeling system by this study, several completed and ongoing studies by NCSU researchers, Raleigh NWSFO forecasters, and others have made use of the real-time simulations in regards to both summer and winter season forecasting issues. Furthermore, MASS was used during the warm season this last year to produce experimental convective nowcasts and a mesoscale discussion.

2. Progress to Date

2.1 Real Time Modeling System

Much of the initial year of our study was spent developing a mesoscale modeling system at NCSU, as planned. With substantial contributions from NCSU researchers, Raleigh NWSFO staff, and the staff of MESO Inc., a limited set of MASS model forecast fields began to be delivered to the Raleigh NWSFO on a daily basis as early as September 1995. We felt that by providing the model on a daily basis, rather than on an event driven basis as originally planned, the forecasters would be able to assess the model's performance during a variety of weather events as well as provide them with more opportunities for making comparisons to the operational model guidance.

A major part of the initial regional modeling system development was focused on providing forecasters with the most flexible means by which to use the new model guidance. During my work with the NWSFO, I quickly learned that operational forecasters are inundated with information and because they have only a finite amount of time by which to formulate their forecast, each one approaches it with their own streamlined methodology. Therefore, to successfully introduce new techniques, data, or models to the forecast process, they must show significant added value to what is already available to the forecaster. In the present case, the value added is primarily based on the model performance; however, it is also influenced by the way in which the model guidance is presented, which we endeavored to optimize.

Incorporating both the new NWS Science Applications Computer (SAC) and the relatively new N-AWIPS software package we were able to provide, not only a standardized set of model forecast products, but also a variety of specialized products on a seasonal basis. Use of the SAC hardware and software also allowed the MASS forecast fields to be delivered more quickly and in a more graphically robust format. Later, with the efforts of several of the Raleigh NWSFO forecasters, staff, and an improved version of the N-AWIPS software, we were able to provide the NWSFO with pure digital model output. The output could then be manipulated to produce both standardized and specialized forecast fields on demand, rather than forcing the forecaster to rely on a "canned" set of products generated beforehand. Besides increasing the usefulness of the model forecasts, this also improved our ability to deliver the model output in a timely fashion and at much higher temporal resolution. During this stage, we also developed a real time MASS model web page. The web provided the sole means for forecasters in the region apart from the Raleigh NWSFO to examine the model forecasts. The real-time modeling system is still being actively supported at NCSU and as of this report remains in experimental use at the Raleigh NWSFO.

The regional modeling efforts were also used to develop and test several new local-forecasting tools. This was arguably the greatest benefit gained from a project joining operational forecasters, modelers, and researchers. On several occasions we developed and participated in workshops, as part of the Raleigh NWSFO seminar series. Several of these workshops were devoted to introducing new model products and in some cases they also focused on operationally significant lessons in mesoscale dynamics. In many ways the mesoscale model itself was a new tool to the operational forecasters; therefore, some of the workshops were used simply to discuss the MASS models configuration, physical parameterizations and model assumptions, and to point out differences between MASS and other synoptic and mesoscale models. In my opinion, the workshops were as valuable to the forecasters as they were to the researchers. In several cases it also gave forecasters an opportunity to provide some ground truth evidence of the model's benefits and weaknesses to date.

In an attempt to supplement locally developed forecast tools and possibly generate new ones, several new model products were developed over the course of two years. Two examples followed closely with the subsequent CAD model evaluation. One such product was a statistical tool for forecasting precipitation type based on the 1000-850 and 850-700 mb partial thickness values. This nomogram-based tool was incorporated into the daily suite of MASS forecast products during periods in which forecasters were concerned with the possibility of frozen or mixed precipitation.

A second product attempted to more succinctly evaluate the strength of the damming environment by calculating the Froude number at a specific series of stations located east of the Appalachian Mountains. Previous CAD studies have made reference to the Froude Number as a "numeric" indicator of the strength of the damming event. The Froude number is most simply described as the ratio of the kinetic energy in a fluid flow to the amount of potential energy needed for the flow to surmount a given barrier. Therefore, by evaluating the Froude number one may be able to assess the potential for an easterly flow regime to be blocked by the northeast-southwest oriented Appalachian Mountain chain (i.e. damming). During our study some weak correlation was seen between the forecast Froude numbers and subsequent CAD events. However an accurate critical value or range of values indicating damming was not apparent. It should, however, be noted that as of yet no formal evaluation or verification was specifically made for either of these products. Further, the MASS project has put in place a valuable tool for both forecasters and researchers who wish to study mesoscale processes.

2.2 CAD Forecasting with the MASS Model

In my M.S. thesis study ("A Real-Time Mesoscale Model Evaluation During Appalachian Cold Air Damming"), a 13-month survey of Appalachian CAD events was examined from the perspective of the real time MASS model forecast. This served as the first formal evaluation of the model's performance in its real-time configuration. The recent focus on operational mesoscale models by NCEP also prompted a concurrent evaluation of the 49-km Early ETA model forecasts. Evaluation of the ETA forecast also provided a means of assessing the needs for regional mesoscale modeling.

Overall the study found that both the ETA and MASS had difficulties capturing the damming environment. Further, at no time did MASS outperform the ETA in its surface temperature, pressure, and wind speed guidance over the damming region. Both models were particularly poor in CAD cases that exhibited secondary "Type B" cyclogenesis. This is an extremely important issue to operational forecasters because secondary cyclogenesis often results in some of the most serious winter weather across the eastern U. S. Specifically, MASS frequently forecast the initial position and track of the secondary cyclone to be too far to the north and west of the observed track. The errors caused the strongest cyclogenesis to be located directly over the shallow surface cold pool, which forms along the eastern side of the mountains. In every case this misforecast prematurely eroded the cold air and disrupted the southeasterly flow regime needed to help support the damming environment.

These results were then used to formulate an experimental design for testing tentative hypotheses for why CAD events were prematurely eroded by MASS in some cases. Five sensitivity tests were developed and performed on a single CAD event (18-19 December 1995). These tests were used as a method for possibly isolating the cause of the systematic bias found in the 13-month survey and also to provide a means of improving the MASS model configuration for future winter season forecasting. The results of the sensitivity tests indicated that the poor cyclone forecast by the MASS model was tied to the models diabatic processes. The results of the sensitivity tests did not however, reveal a method for improving the current model configuration; rather they only emphasized the need for further research into the models microphysical and convective parameterization schemes. Further, one of the sensitivity experiments examining the impact of improved microphysical parameterizations revealed an apparent coding error in the new prognostic mixed-phase microphysical parameterization packages. The model developers were made aware of the problems and are currently working with researchers to resolve it. Once this is resolved, a more definitive evaluation can be obtained of the role of explicit microphysical parameterization in correctly modeling the damming process in real-time.

These results elucidate both a benefit as well as a problem with regional modeling efforts being conducted at universities across the country, which are collected with NWS offices, in most cases under COMET support. The problem stems from a lack of operational testing and evaluation of what have historically been research models (MASS, MM5, etc.) as they begin to be used as experimental operational tools. The benefit, however, is the ability to find, understand, and possibly improve upon the models' performance through cooperative studies. Accessibility to the model is the key to this benefit, which is not easily possible with the NCEP models.

This study was by no means an exhaustive set of sensitivity simulations, however, without corrections, the MASS may provide little utility to operational forecasters in the damming region during CAD events which are accompanied by secondary cyclogenesis. The fact that the NCEP 49 km ETA model provided similar, and in most respects, improved guidance as compared to the MASS may also cause one to question even the necessity for local modeling efforts in this region. However, it must be emphasized that this study examined only one of many forecasting issues in the Southeast which could be addressed with a local model, and if improvements can be made the MASS forecasts may prove useful during future CAD events. In fact, the experience over the summer season in 1997 with using MASS to produce experimental convective nowcasts has suggested significant value in terms of its ability to refine the location and timing of mesoscale precipitation events associated with often subtle forcing mechanisms.