Appears in: Proceedings, 1998 Sigma Xi Forum, November 12-13 (Vancouver, BC, Canada), pp. 159-166. [Available from Sigma Xi, The Scientific Research Society, P.O. Box 13975, Research Triangle Park, NC 27709]
Notice:
This is a U.S. Government manuscript and, as such, is public
domain.
Hazardous weather can take many different forms: heat waves, blizzards, tornadoes, hurricanes, floods, ice storms, and fog, to name just a few. Moreover, weather is a contributor to other disasters like avalanches, mudslides, fires, and crop failures. Since weather is global, it can be hazardous virtually anywhere. Thus, to do a comprehensive treatment of the hazards and impacts of weather on human society would require far more effort than I am permitted for this presentation.
Instead, I am going to use tornadoes as a proxy for the whole range of events. This choice is dictated largely by my own interests, rather than because tornadoes are so widespread. In fact, tornadoes are largely considered to be a phenomenon confined to the central plains of the United States; the so-called "tornado alley" region. This perception is not precisely true even within the United States. "Tornado alley" is not a scientific term but rather one promoted mainly by the media. The actual distribution of tornadoes in the United States is reasonably well-known, although it is likely that even now we are missing about half of the actual number of tornadoes; reporting of tornadoes and the impacts of tornadoes is subject to many vagaries. To make the climatological record, a given tornado must be seen by an observer, the observer must recognize it correctly to be a tornado, and must report the observation to an agency that will see to it that it enters the official record. The official record is maintained by the National Climate Data Center of the National Oceanic and Atmospheric Administration (NOAA). However, the actual responsibility for collecting the data about tornadoes falls on the National Weather Service as an extra duty; there are no resources for careful and detailed analysis of individual tornado events. The lack of a systematic process for gathering and analyzing reports of tornadoes leads to considerable concern for the completeness of the record and accuracy in the details of that record. Nevertheless, it is what is available, and that record forms the basis for what I will present here.
Given this situation in the United States, it is quite likely that tornadoes are seriously underreported world-wide. If it is perceived that tornadoes do not occur in some place, then no effort will be made to collect information about tornadoes in that place, leading to a perpetuation of the initial perception. In my own experience, I know that many tornadoes have occurred in Spain, including some that have been captured on videotape. Nevertheless, the Spanish government maintains no official record of tornadoes; hence, appearances that tornadoes simply don't happen in Spain (see Snow and Wyatt 1997) are quite deceptive. Although there can be little doubt that tornadoes are most frequent in the central plains of the United States, they are far more common world-wide than they are thought to be.
The available data about fatalities in the United States show a pretty erratic record (Fig. 1); generally speaking, the "spikes" in the record of the annual fatality count correspond mostly to individual events or to single-day outbreaks of tornadoes. That is, the annual fatality totals tend to be dominated by a small number of events per year. The best example of that is for the year 1925, when what was apparently a single tornado (see Changnon and Semonin 1966; Doswell and Burgess 1988) killed almost 700 people. This single event is the largest death toll ever for a tornado in the United States, although tornadoes in other places around the world have produced more deaths than this. It can be seen from this record that the number of fatalities has been declining.
Fig. 1. Annual fatality total for the United States. Annual values are shown by the solid black dots connected by dashed lines (data for 1998 were incomplete when the figure was done); the solid red line represents a smoothed version of the data, produced by applying first a 3-point median filter followed by a 5-point moving simple average.
If this record is normalized by the population of the United States (Fig. 2), it can be seen that the number of deaths per million of population was roughly constant until 1925, and then it began an exponential decline. Reasons for this decline in the face of an expanding population are not entirely clear, but it may be related to changing demographics; specifically, a declining rural population (López and Holle 1998) and an associated concentration of population in cities. Hazard awareness may also be a factor (Doswell et al. 1999).
Fig. 2. Annual tornado fatality total in the United States, divided by the population (interpolated linearly between census figures). Raw data are shown by the solid black dots connected by dashed lines (data for 1998 were incomplete when the figure was done); the solid red line is smoothed as described in Fig. 1, the solid dark green lines are regression lines fit to the smoothed data through 1925 and from 1925 onward, with the dashed green lines showing the 10th and 90th percentiles about the regression line.
Like the annual fatality count, the annual damage done by tornadoes (adjusted for inflation using the annual consumer price index) is erratic (Fig. 3). Again, most of the damage is associated with a small number of events annually. The reason of this is clear: the total area affected by a tornado is typically rather small, on the order of a few square kilometers. The odds of a populated area being struck are correspondingly low. Even in the peak frequency zones of Kansas, Oklahoma, and Texas, the odds of a particular square kilometer being hit are quite small in any given year: on the order of one chance in a thousand or less.
Figure 3. Annual estimated damage (red solid dots connected by thin black lines); solid red line is after smoothing as described in Fig. 1. Dollar values are in inflation-adjusted 1995 dollars.
Another interesting trend in the tornado hazard can be seen by considering the ratio of the damage costs (again, adjusted for inflation) to the number of fatalities. This ratio is designed to reveal the amount of damage needed to produce casualties, so a large number indicates relatively low hazards, while a small number shows relatively high hazards. It can be seen (Fig. 4) that for a selection of the most significant tornadoes (those producing large fatality counts and/or large damage estimates), this number has been rising. Some of that rise is apparently associated with increasing costs associated with tornado events; in simple terms, as time goes by, citizens have more things in their homes and businesses to damage. However, it appears that an important change took place in 1953, perhaps not coincidentally the most recent year in which a tornado killed 100 or more people (114 in Flint, Michigan on 08 June). The inception of tornado forecasting in the United States was in 1952, but 1953 was the first full year of operational tornado forecasting. We have tested the statistical significance of the discontinuity in the year 1953, and the fitted regression lines including this discontinuity produce the best fit to the data.
Figure 4. Dollar damage-to-fatality ratio for selected tornado events (black dots), using inflation-adjusted 1995 dollars; two noteworthy disasters (the major storm of 1925, called the "Tri-State" event and the 1979 event in Wichita Falls, Texas) are highlighted as red dots). Events were selected from the record by having more than 45 fatalities and/or more than $50 thousand in damage. The solid green line is a regression line fitted to the events up to and including the Flint, Michigan tornado of 08 June 1953, while the magenta line is for events after the Flint tornado of 1953. The curved blue line is a quadratic curve fit to the data; it has been shown that the mean squared deviation from the discontinuous straight line fit is less than that associated with the quadratic fit. Although the Saragosa, Texas event of 1987 fails to qualify (too few fatalities), it is plotted here as a reference event (see text for discussion).
Thus, the trends in tornado fatalities suggest that overall, a number of factors have contributed to a long-term trend toward fewer fatalities. It is difficult to know precisely which of the contributing factors is most important. Given the apparent decline in the fatalities in the face of continuing large damage, it seems that tornadoes are becoming less of a hazard. Events of 1998, including a fatality count that exceeds the 90th percentile about the mean shown in Fig. 3, indicate that, nevertheless, we may well be subject to a continuing hazard, in spite of the trends.
Concentration of population in cities means a lower annual probability of a disaster, but it also carries with it what may be an increasing risk of a substantial number of fatalities in the unlikely event that a populated area is hit. Shown on Fig. 4 is an event that hit the small town of Saragosa, Texas in May of 1987; the event did not qualify for inclusion with the other events of the figure because it caused neither the required number of fatalities nor the required amount of damage. What it does illustrate, however, is that we are still vulnerable to tornadoes and there is no reason to be believe we can't be unlucky in the future. In fact, it can be argued that the United States has been extraordinarily lucky; each year, the number of near-misses is substantial. In 1998, for example, a violent tornado narrowly missed Birmingham, Alabama. Although Nashville, Tennessee was hit directly by a tornado in 1998, it turned out not to be a very strong tornado. Eventually, our luck will run out.
Thus, the record suggests that although we continue to be threatened by tornadoes, considerable progress has been made. Today, more is known about tornadoes and tornadic storms than ever and that understanding is being incorporated into operational forecasting and warning procedures within the National Weather Service, to an unprecedented extent. It's impossible to know precisely how many lives have been saved, but my colleagues and I have estimated that the number is probably on the order of ten thousand lives (Doswell et al. 1999).
In 1994 and 1995, a field project was undertaken to study tornadoes; called Verification of the Origin of Rotation in Tornadoes Experiment (VORTEX), it was the largest tornado-related field project in history (see Rasmussen et al. 1994). The team of scientists involved in VORTEX included a broad spectrum of institutions and included more field observational capability than ever used before to study tornadoes and tornadic storms. The data from that field phase of the project included samples from more tornadic storms than have been looked at by all previous efforts combined. Although the scientific lessons learned from VORTEX are still not entirely clear, one thing has become clear from the project so far: we still have a lot to learn. Many of our ideas about tornadoes and tornadic storms had their origins in numerical simulation models. Although these models have been quite successful in simulating many aspects of these events, our field observations were quite successful in pointing out the shortcomings of our model-derived understanding. I believe this underscores the importance of observations in meteorological science. Although practical applications of the results from VORTEX may be modest at the moment, the potential for forecasting improvements remains tied to the development of increased understanding of the physical processes giving rise to tornadoes. Certainly improved simulation models will point the way to improved forecasting models, and observations provide an important basis for model improvements.
Improved prediction is certainly possible, but the benefits of increased understanding will also depend on having new and improved data about the current state of the atmosphere. The observational infrastructure that underpins the prediction models has been aging and is in need of continuing modernization and investments in new observing capabilities. Modern satellites and radars are expensive and the costs for these improvements may have some negative impacts on the human side of the infrastructure as forecasting agencies seek to keep overall costs down. This is a continuing issue and certainly is not unique to weather forecasting in society; we are all seeking an appropriate mix of investments in humans and in hardware.
Surprising to some is that there are clear indications that mitigation of damage from tornadoes is possible, as well as reductions in casualties. In fact, damage reduction and casualty reduction are linked. The seeming preference of tornadoes for hitting mobile home parks is, in reality, a reflection of the structural integrity of manufactured homes relative to typical frame home construction. Damage, and its associated danger of casualties, increase substantially as structural integrity decreases. This means that anything that can increase the structural integrity of homes and businesses in the path of tornadoes can mitigate damage costs as well as reducing the vulnerability of the inhabitants. Building codes and their enforcement can play a major role in increasing structural integrity. Many rural areas and some urban areas either have no building codes, inadequate building codes, or a lack of enforcement. Tornadic winds vary widely over the total area affected; even in the most violent tornadoes, the fraction of the total area that experiences the most violent windspeeds (up to about 150 meters per second in the most violent events) is typically about one percent. Since the vast majority of the path experiences less than devastating windspeeds, the potential of reducing damage from enhancing structural integrity is quite real.
Yet another factor in hazard mitigation is preparedness. Although tornadoes are rare events in any one place (recall that even in the most tornado-prone areas, the odds of experiencing a tornado are quite low in any given year), nevertheless, there are about one thousand reported tornadoes per year in the United States. The odds don't help you much when you are actually hit! The key to survival and to reducing injuries is preparation. Although the National Weather Service provides both forecasts for tornadoes (called watches) and the local tornado warnings when tornadoes are either underway or imminent, it is up to local communities to develop plans for protection of citizens and, ultimately, it is up to the citizens themselves to accept responsibility for their own safety. It's well-known that those who have planned what they will do in the event of a weather hazard are much more likely to take the right action than those who don't. Thus, being aware of the threat potential and being prepared to act promptly and correctly in the face of an actual threat are the keys to safety.
One thing that is clear from our scientific study of tornadoes is that prevention of tornadoes is not feasible now or is it likely to be in the foreseeable future. Not only is our understanding of the physical processes inadequate to suggest feasible tactics for mitigating the tornadoes themselves, but the growth of that understanding has made it clear that pretty large energies are involved. Tornadic storms can release the energy equivalent of a one-megaton thermonuclear bomb during their lifetimes although, obviously, they spread the release of that energy over a much longer time than does a thermonuclear bomb. What makes storms like tornadoes hazardous is that they concentrate atmospheric energy into relatively small areas. That concentration is what makes the winds so strong and the danger so high, but the energies involved in the whole process are large enough that tactics for modifying that process must command enough energy that a "cure" could be just as devastating as the tornado itself.
I've concentrated on the tornado hazard in the United States in this presentation, but most of what I've said could be adapted to apply to many other hazardous weather events. Severe thunderstorms produce vast amounts of damage from large hail and non-tornadic "straight-line" winds. Heavy rainfalls from thunderstorms can generate flash floods under suitable hydrological conditions, notably in regions with steep terrain and in areas that have experienced antecedent rainfalls such that ground infiltration of the rainfall is minimal. Flash floods typically cause more fatalities year in and year out from storms than any other hazard, with the possible exception of lightning. Moreover, floods and flash floods are arguably the most widespread hazard from storms, world-wide.
Development and recreational use of flood-prone areas puts many more people at risk and creates huge annual costs to society. All societies must confront directly the tough issues tied to mitigation of hazards associated with development and recreational use of hazardous regions in general, including coastal zones (which are subject to accelerated erosion, high winds, and flooding from either tropical cyclones or coastal storms) and regions of steep terrain (which are subject to flash floods, as well as fires in drought years and rock/mud slides in rainy years). Pressures for development and expansion of recreational use must be balanced against the societal costs. Extended episodes of high temperatures and intense cold probably cause more fatalities during the course of several years than storms do during the same period.
Scientific data collection, development of operational infrastructures for hazard mitigation, and investments in basic research are all contingent on favorable national and international priorities. Virtually every science can marshal strong arguments in favor of increased investment in that science and technology, so societies must struggle to find priorities among the many competing uses for what may be a dwindling resource: the national inflation-adjusted budget. Weather hazard mitigation is something that has the potential to affect everyone, but not everyone experiences it directly every year. Moreover, hazard mitigation only pays for itself through reductions in costs and losses; it doesn't generate income directly but, rather, creates profit by reducing outgo when done properly.
It is difficult for meteorologists, in general, to show the real benefits associated with their activities (see Doswell et al. 1999). As already noted, for example, it is not obvious how to identify the number of lives saved as a direct result of tornado watches and warnings, although we can estimate it indirectly. If an estimate of 10, 000 lives since 1953 is of the correct order of magnitude, what is the benefit to society? I don't know of any obvious way to assign a dollar value to a human life but, for the sake of argument, I assume it to be $2 million. If this figure is of the right order of magnitude, then the direct benefit to the nation during that period is roughly $20 billion from tornado forecasts and warnings. I estimate the total cost of the tornado forecasting and warning service provided by the National Weather Service during that same period to be on the order of $5 million per year, for a total of $225 million; this yields an estimated "profit" associated with severe weather forecasting and warnings that approaches ten times the investment. I believe that this crude result is actually representative of the rate of return on investment associated with weather forecasting services, although (as noted) it is difficult to be very confident of the details. Although this result is strictly appropriate only for the United States tornado threat, I conclude that continuing investment in severe weather research and operations is probably a reasonably prudent thing for societies to do.
Changnon, S. A., and R. G. Semonin, 1966: A great tornado disaster in retrospect. Weatherwise, 19, 56-65.
Doswell, C. A. III, and D. W. Burgess, 1988: On some issues of United States tornado climatology. Mon. Wea. Rev., 116, 495-501.
Doswell, C. A. III, A. R. Moller, and H. E. Brooks, 1999: Storm spotting and public awareness since the first tornado forecasts of 1948. Wea. Forecasting, 14, 544-557.
López, R. E., and R. L. Holle, 1998: Changes in the number of lightning deaths in the United States during the twentieth century. J. Clim., 11, 2070-2077.
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.
Snow, J. T., and A. L. Wyatt, 1997: Back to basics: The tornado, Nature's most violent wind: Part I -- World-wide occurrence and categorisation. Weather, 52, 298-304.