Appeared as: paper 2.2 in Preprints, 14th Conf. on Weather Forecasting and Analysis (15-20 January 1995), Dallas, TX, American Meteorological Society, 23-28.
Corresponding author address: Dr. Charles A. Doswell III, National Severe Storms Laboratory, 1313 Halley Circle, Norman, OK 73069 e-mail: email@example.com.
During the spring of 1994, an experiment was conducted called the Verification of the Origins of Rotation in Tornadoes EXperiment (VORTEX-94). This experiment has been described in some detail by Rasmussen et al. (1994). Forecasting for the field operations was a component of the overall experiment, of course, but some experimental forecasting was done that was not necessarily an integral part of the field operation. This dual role for forecasting has been typical of experiments conducted in and near Oklahoma during the past several years (see, e.g., Doswell and Flueck 1989, Jincai et al. 1992).
As discussed in Rasmussen et al. (1994, hereinafter referred to as R94), the primary objectives of VORTEX-94 are all related to tornadoes and tornadic storms. The goals of the forecasting component associated with VORTEX-94 include:
Moreover, given the observing platforms available to us during the experiment (as described in R94), it is clear that opportunities would arise to observe severe thunderstorms and their environments in considerable detail. Phenomena heretofore undocumented or poorly observed with previous data sets present a special challenge to forecasters. Several such opportunities arose during the experiment, as anticipated, and it is these which provide much of the motivation for this report. Interesting forecast problems can arise even in years when the weather during the experiment fails to produce many of the desired events. Indeed, 1994 was not characterized by very much tornadic severe weather compared to climatological norms. Nevertheless, forecasting challenges were common so that, in some sense, the forecasting components of field experiments virtually cannot "fail" even if no target events occur.
Fulfillment of the specific forecasting objectives described above will require a considerable effort at verification, and that effort is just now beginning, to be reported upon in the future. Nevertheless, the events during VORTEX-94 offer a chance to point out some of the important issues that will have an impact on the National Weather Service (NWS) Modernization that is presently underway.
In what follows, we shall describe a selection of topics derived from our forecasting experiences in VORTEX-94. These will by no means exhaust the forecasting issues that arose during the experiment, but should suffice to represent the forecasters' concerns. Future publications based on the experimental data will explore these and other topics in more detail than we can do at this early stage of the research following the first collection of field observations. VORTEX will include another field phase during the spring of 1995, so that the experiment should provide a rich spectrum of events for study.
2. PREVIOUSLY UNOBSERVED EVENTS
Owing to the new observational systems available to the forecasters during VORTEX-94, it was inevitable that we would see events that we have little or no experience with in the past. It should be clear that this presents a particularly vexing challenge to forecasters: it is difficult to imagine dealing successfully with meteorological phenomena about which little or nothing is known.
Perhaps the most striking of such events arose on 14 April, when a dryline was intruding into the western part of the state. Although the soundings in the moist layer revealed considerable potential instability, it was clear that there was substantial negative area to be overcome if any deep surface-based convection was to develop (Fig. 1). With considerable surface heating and the approaching dryline, it was felt that the chances for severe weather merited field operations. As it turned out, the negative area was never overcome, and only towering cumulus clouds developed in central Oklahoma late in the afternoon.
Fig. 1. Skew-T plot of sounding from 14 April 1994, showing the capping stable layer.
Fig. 2. Surface map from 14 April 1994, showing the complex
dryline structure. Solid lines are isodrosotherms at 5 C intervals.
However, as described in Brooks et al. (1995), as the afternoon wore on, the dryline developed considerable structure, especially as evidenced by the Oklahoma Mesonetwork. The complex structure seen in the dryline (e.g., Fig. 2) is such that it cannot be sampled well by a surface network with station spacing on the order of 100 km. If such an event is sampled by the regular surface reporting stations, it is quite likely to cause confusion and uncertainty in the forecast. Indeed, even with the mesonetwork, this complex dryline behavior certainly caused our forecasters confusion and uncertainty, simply because it is unclear what meteorological processes are operating to create such complex dryline structure. Some past studies (e.g., Koch and McCarthy 1982; Davies-Jones and Zacharias 1988; Sanders and Blanchard 1993) have hinted at "waves" along the dryline, but this event is not obviously oscillatory.
Another type of interesting event can be found on 03 May when some rather unexpected supercells developed in southeastern Colorado. As is typical for the time of the year, moisture and instability were present as seen in the 0000 UTC sounding near the storm at Dodge City, KS (not shown), but it was not obvious to the forecasters that enough wind shear of the appropriate character would be available to any storms that developed. The wind profiler at Granada, Colorado (Fig. 3) revealed that a favorable wind profile had developed nearby the storm but this wind structure was not obviously present earlier in the day, and appeared to be confined to the vicinity of the storm. The origins of such features in the wind field are quite obscure, and certainly represent a serious challenge to forecasters.
Fig. 3. Time series of hourly wind profiles at Granada,
Colorado during the evening of 03 May.
3. NWP MODEL-RELATED PROBLEMS
A particularly frustrating episode for the forecasters arose during the days leading up to the evening of 25 April. The long range NWP model forecasts for 500 mb depicted a significant trough that would pass through Oklahoma on the evening of 25 April (Fig. 4). As the date approached, the "optimistic" forecast continued to be maintained by the models, lending further credence to that forecast. Indeed, as it turned out, the forecast was quite good, with the observed trough at 0000 UTC on 26 April seemingly ideally positioned for severe weather in Oklahoma (Fig. 5).
Fig. 4. 240-h 500 mb height and vorticity forecast valid at 00 UTC on 26 April.
Fig. 5. Observed 500 mb chart at 00 UTC on 26 April. Thick
solid lines are height contours at the standard 60 m intervals, while
the thin solid lines are contours at intermediate 30 m intervals, for
comparison with Fig. 3.
However, in spite of this excellent series of 500 mb forecasts, the expected major outbreak of severe weather failed to materialize. The apparent cause for this failure is the development of convection during the morning of 25 April. This early convection swept rapidly across much of Oklahoma, stabilizing the stratification and leaving an inhibiting stable layer that virtually prevented significant developments in Oklahoma. What severe weather that did occur was south of the Red River, in Texas, and while those storms were tornadic and otherwise intense, the day did not produce the anticipated major outbreak of severe weather.
Another model-related problem was the apparent tendency of all the NWP models to diminish peak wind values aloft, even over periods as short as 12-h. Late in spring, it becomes ever more difficult for the atmosphere to develop strong enough wind speeds aloft (at 300 mb or higher) to maintain intense convection. During the period of 23-30 May the flow over the VORTEX area of operations was predominantly northwesterly. Minor short wave troughs and wind maxima aloft were embedded within that flow. The likelihood of severe weather as evaluated by our forecasters depended critically day-to-day on the forecast windspeeds aloft. With few exceptions, the flow aloft depicted on the PC-Grids diagnostic program were consistently weaker (by as much as 20-25 kt!) than the observed peak windspeeds aloft. [As of this writing, full documentation of this problem is not complete; this will be reported upon in detail in a future publication.] This persistent underforecasting of the windspeeds aloft led to many forecasts underestimating the potential for severe weather until late in the period, when forecasters began to recognize and account for this bias.
A somewhat puzzling problem was the presence of apparent flaws in some of the PC-Grids algorithms (see Petersen 1992) for calculating some of the important diagnostic parameters from model forecast data. It appears that the Convective Available Potential Energy (CAPE) values depicted by the program from the model data were in error by a factor of two, being consistently twice as large as the model forecasts really indicated. This began to become apparent only well into the experiment, and the nature of this seeming "bug" in the CAPE calculations remains unknown as of this writing.
4. ANOMALOUS STORM BEHAVIOR
A not entirely unexpected finding during VORTEX-94 was a number of storms that behaved in ways that do not fit accepted knowledge of how the environment influences the observed severe weather. In view of the relatively obvious fact that we do not know all there is to know about how storm behavior is affected by environmental conditions, events of this sort were anticipated. The atmosphere did not disappoint these expectations, but this is small consolation to forecasters and the field crews when forecasts of storm behavior go awry.
One such event occurred on 27 April, when considerable CAPE was present and storms developed on an intense thermal boundary near the Red River. It was observed by the field teams that the cumulonimbus towers on this day were rather slow-rising and "fuzzy" in visual appearance, indicative of rather modest updrafts. Some marginally severe hailfalls were reported with these storms. Given the large CAPE on the warm side of boundary, it is difficult to reconcile such storm behavior; it is unlikely that the updrafts were being fueled by the near-surface air equatorward of the intense surface boundary. Under somewhat similar conditions, quite severe storms had developed the previous day, including a tornadic storm near Gainesville, Texas, and it is not at all obvious how a forecaster would have been able to anticipate the relative weakness of the 27April convective storms.
In contrast, on 26 May, a supercell developed near Lubbock, Texas that was quite well sampled by the field teams, including some special soundings in the near-storm environment. The soundings exhibited relatively modest CAPE values of 783 J kg-1. Although supercells have been observed with small CAPE values (Johns et al. 1993), it is interesting that this storm produced large hailstones (stones at least 4.5 cm in diameter were observed). It is not entirely clear how a storm occurring in such an environment can produce large hail. Of course, it is possible that the "dynamic" contribution to updraft strength can produce updrafts well beyond those predicted from pure buoyancy alone. However, most supercells in low-CAPE environments are not typically producers of large hail, so some other parameter besides updraft strength might be important in the development of such hailfalls.
In the preceding, we have indicated some of the challenges VORTEX-94 forecasters encountered. These are exciting findings; unexpected observations and new phenomena are the grist for a scientist's mill. Exploration of these topics can lead to new understanding of atmospheric processes, revision of scientific models, and (ultimately) to improved forecasts. However, it would be misleading to ignore the frustration forecasters experience due to such unexpected events. Forecasters do not like to be wrong, even if their job virtually requires them to be wrong frequently; any forecaster who accepted such "defeats" without wincing is probably unsuited to be a forecaster.
The modernization of the NWS is associated with a considerable growth in observing capability. As we have indicated, a direct consequence of that enhancement to our observations is the discovery of heretofore unobserved phenomena. It would be naive in the extreme to believe that scientific understanding is already in place to deal with everything that we are going to see in the next decade as new observing systems come on line. How are forecasters to cope with this scientific bonanza? It is all well and good to go on about how exciting these discoveries will be, but each one is going to be associated with a lot of busted forecasts!
The only way to deal with this situation, in our opinion, is for operational forecasters to recognize their obligation to take on what may seem to many to be an unfamiliar role: the forecaster as a scientist. Whereas it might have been possible (arguably!) in the recent past to treat forecasting as a more or less "technical" job involving the application of science, this seems increasingly a barren and short-sighted attitude for forecasters to adopt. Like it or not, the interests of forecasters and researchers are going to be more coincident in the next decade than in any comparable time period since the 1940s, with the proliferation of operational radiosondes around the world. The history of our science has suggested that research and operations interactions, leading to the development of new understanding peak during such periods when new operational data sets are of research interest (see Doswell et al. 1981). Since this condition is going to exist for the next decade or so, it is now a time to encourage forecasters to take a more active role in research. They observe the weather every day and will be most likely to see new phenomena when they occur under the close scrutiny of the new observing systems.
Furthermore, the forecaster has a great deal of vested interest in the rapid development of new understanding applicable to forecasting. If researchers are unwilling to pursue their discoveries as applications, which is generally the case, then it should be obvious that forecasters must assume a larger role in applications development than they do at present. This in turn has implications for what sort of person is likely to be successful as a forecaster; it is quite inconceivable that a Bachelor of Science degree is going to be adequate in the future for a forecaster, irrespective of one's opinion about its adequacy at present.
The continuing development of NWP models is assured, of course. However, as they grow in complexity, it appears that there will be diminishing returns from the approaches that have been successful in the past. As Brooks et al. (1992) have suggested, it is unlikely that explicit predictions of convective phenomena are going to be as successful as the synoptic scale NWP models were. Moreover, Brooks and Doswell (1993) have indicated that increasing technological capability should not be channeled exclusively into enhancing model resolution. Whatever course the so-called ensemble forecasting takes, it appears that humans will be needed to interpret the results of the ensemble approach. For a number of reasons, it is becoming apparent that the dream of a totally automated forecasting system is not likely to be a very good forecasting system. Although the juggernaut of automation is yet to run its course, there are many signs that a viable role for humans in weather forecasting will exist into the far future. But that role is not going to be as "technological" as it has become at present. A forecaster is going to need ever greater science orientation with time, and this means that forecasters must begin to see themselves as scientists and not "simple forecasters." They must recognize the meanings inherent in new observations, knowing when a new observation contradicts existing paradigms and knowing how to accomplish a substantive test of a scientific hypothesis, as new hypotheses are created from the new observations.
In its implementation, the NWS Modernization effectively has assumed that (a) science is going to be accomplished mainly outside the NWS, (b) scientific understanding sufficient to take advantage of the new observations already exists, and (c) it is not the responsibility of the operational forecasting agencies to ensure that the new data sets are archived and available at reasonable cost for scientific study. These assumptions are, we contend, manifestly wrong and threaten the very success of the modernization.
Of course, it might be possible simply to declare the modernization a success, effectively emulating the fable about the emperor's new clothes. We do not believe this will be an acceptable answer, in view of the challenges associated with using the new observations. New diagnostic tools will have to be developed and debugged in order to apply the new scientific concepts growing out of the new data sets. New ways to look at the data to replace superannuated techniques currently employed will be required. The complexity of linking software systems continues to plague our society as a whole, resulting in huge cost overruns and costly delays in implementation of new systems, threatening the modernization in a purely technical way. We in meteorology can be leaders in the solution of these problems, or we can simply thrash about ineffectually while society as a whole struggles to cope with technological complexity. The NWS Modernization represents sufficient resources to attract public scrutiny of the process, unlike much of our meteorological past, when our whole program was "small potatoes." We cannot afford to ignore the important challenges that these new data represent.
Finally, we hope that the foregoing presentations and discussion suggest that field projects like VORTEX and operational weather forecasting share enough common ground that there is a real benefit for future interactions. If a research scientist has an interest in successful forecasts, then to that extent, that researcher has a common goal with forecasters. When the data of research interest include large amounts of operationally-collected data, then that collaboration is of even greater value. Forecasting problems are problems to the field project researcher, and vice-versa. This presents us with a real opportunity to expand on the interaction between research and operations. It is not clear that the NWS is postured to participate in this sort of interaction. The Science and Operations Officer (SOO) position is arguably a start, but the level of on-site scientific expertise cannot be confined to an overburdened SOO, who must also take on some administrative and tutorial duties, as envisioned in the modernization plan. The NWS has relatively little experience with and understanding of the scientific research process. Perhaps a substantive commitment to the Experimental Forecast Facility (EFF) concept within the NWS is a viable plan for research in the operational arena, but current levels of commitment appear to fall far short of a "critical mass."
Management of scientific research is largely an alien concept to the NWS management teams, who have virtually no experience in basic or applied research. The open-ended character of real scientific investigations typically leads to discomfort among NWS managers, who feel most at home with well-defined tasks and accomplishment timetables. This situation needs to change, if the system as a whole is to move efficiently into the era implied by the modernization. It seems contradictory that research, a process quite definitely not characterized by constraints such as timetables for accomplishment, is needed to make the modernization a reality. However, the research is necessary to take the maximum advantage of the gains represented by the new observing and diagnostic systems being introduced by the modernization. Science may not be straightforward, but its accomplishments are the only rational basis for successful application of new observations to forecasting. If the advantage associated with a scientific approach is not pursued vigorously, then the "learning curve" for efficient utilization of the new data streams and computing capability will be extended for many years; the history of meteorology has shown this repeatedly. The choices made now to pursue, or not to pursue, a scientific approach will reverberate in our profession for many decades to come. It is our hope that real support (and not pro forma exercises for cosmetic purposes) for science by the operational forecasting agencies will be forthcoming.
Acknowledgments. The forecasting in VORTEX-94 would not have been possible
without the voluntary participation of a number of individuals. Lead
forecasters during VORTEX-94 included: Mike Branick (NWSFO Norman,
OK), Don Burgess (WSR-88D OSF), Chuck Doswell (NSSL), Jack Hales
(NSSFC), Bob Johns (NSSFC), Larry Ruthi (NWSFO Norman, OK), and Steve
Weiss (NSSFC). Assistant forecasters included: Arnold Ashton (AES),
Phillip Bothwell (SPC), Harold Brooks (NSSL), John Cortinas (CIMMS),
Dennis Dudley (AES), Roger Edwards (NSSFC), John Hart (NSSFC), Paul
Janish (NSSL), Rob Kuhn (AES), Mike Leduc (AES), Norm Paulsen (AES),
Isabel Ruddick (AES), Ria Reesor (AES), and Glenn Vickers (AES). We
are very grateful to their respective agencies for their support of
Brooks, H.E., C.A. Doswell III, 1993: New technology and numerical weather prediction - A wasted opportunity? Weather, 48, 173-177.
_______, ______ and R.A. Maddox, 1992: On the use of mesoscale and cloud-scale models in operational forecasting. Wea. Forecasting, 7, 120-132.
_______, _______, E.N. Rasmussen and S. Lasher-Trapp, 1995: Detailed observations of complex dryline structure in Oklahoma on 14 April 1994. Preprints, 14th Conf. Wea. Analysis and Forecasting, Dallas, Texas, Amer. Meteor. Soc., 62-67.
Davies-Jones, R., and D. Zacharias, 1988: Contributing factors in the 10 May 1985 tornado outbreak in northwest Kansas. Preprints, 15th Conf. Severe Local Storms, Baltimore, Maryland, Amer. Meteor. Soc., 284-287.
Doswell, C.A. III, L.R. Lemon, and R.A. Maddox, 1981: Forecaster training -- A review and analysis. Bull. Amer. Meteor. Soc., 62, 983-988.
______, and J.A. Flueck, 1989: Forecasting and verifying in a field research project: DOPLIGHT '87. Wea. Forecasting, 4, 97-109.
Jincai, D., C.A. Doswell III, D.W. Burgess, M.P. Foster and M.L. Branick, 1992: Verification of mesoscale forecasts made during MAP "88 and MAP '89. Wea. Forecasting, 7, 468-479.
Johns, R. H., J.M. Davies and P.W. Leftwich, 1993: Some wind and instability parameters associated with strong and violent tornadoes. 2: Variations in the combination of wind and instability parameters. The Tornado: Its Structure, Dynamics, Prediction, and Hazards (Church et al., eds), Geophys. Monogr. 79, Amer. Geophys. Union, 583-590.
Koch, S.E., and J. McCarthy, 1982: The evolution of an Oklahoma dryline. Part II: Boundary-layer forcing of mesoconvective system. J. Atmos. Sci., 39, 237-257.
Petersen, R.A., 1992: A PC-based system for the display of gridded RAFS data. Proceedings, WMO Tech. Conf. Tropical Aeronautical Meteor. (TECTAM-92), Geneva, Switzerland, World Meteorological Organization, 1-6.
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 origins of rotation in tornadoes experiment: VORTEX. Bull. Amer. Meteor. Soc., 75, 995-1006,
Sanders, F., and D.O. Blanchard, 1993: The origin of a severe thunderstorm on 10 June 1985. Mon. Wea. Rev., 121, 133-149.