Storm Data for Societal Impacts: An Immodest Proposal


Chuck Doswell

Posted: 11 April 2009 Updated: whenever

As is typical, the following represents my opinions. If you wish to comment on or dispute my views, feel free to email me at (use the hyperlink or cut and paste after changing the # to @). If you're not willing to have your opinions added to this webpage, however, don't waste my time and yours.

The official data base for information about storms is the publication Storm Data, available from the National Climate Data Center (NCDC - another part of NOAA). This is considered to be the ultimate "official" source for information about storms. However, the Storm Data data base has a number of problems. Over the years, many aspects of the data have been called into question in refereed articles about the climatology of severe convective storms - I've been involved with several scientific papers that have called attention to these problems (see the Bibliography). However, this is not the intention of this essay. Although those problems continue to be of concern for anyone trying to deduce useful information about storm climatologies, the recent focus on issues of how storms affect society and how best to use information about storms to the benefit of all has raised some new questions about the data that we have about storms.

1. Casualties

The term "casualty" here means the aggregate of fatalities and injuries. For severe convective storm-related casualties, there are two types: direct and indirect. According to a recent revision of the procedures for Storm Data (as of August 2007), these are defined as:

A direct fatality or injury is defined as a fatality or injury directly attributable to the hydro-meteorological event itself, or impact by airborne/falling/moving debris, i.e., missiles generated by wind, water, ice, lightning, tornado, etc. In these cases, the weather event was an “active” agent or generated debris which became an active agent.

Fatalities and injuries, occurring in the vicinity of a hydrometeorological event, or after it has ended, but not directly caused by impact or debris from the event (weather event was a passive entity), are classified as indirect.

A little bit of thought can produce examples of fatalities that might be difficult to classify... for example, suppose the wind picks up a victim who becomes so terrified by the experience that a heart attack ensues. A medical examination might determine that any injuries sustained as a result of being lofted into the air were not sufficient to be deemed a likely cause of death. So is this direct or indirect? In any case, Storm Data is to be lauded for attempting, at least in principle, to keep track of both direct and indirect fatalities. For tornadoes, I would guess that ambiguous cases are relatively rare. [They might become more common for other types of weather.] I note that this attempt to track both direct and indirect fatalities is of quite recent vintage and this vast majority of the historical content of Storm Data has made no attempt to document indirect fatalities - only direct fatalities were to be counted. Hence, it will take the passage of time to begin to assess the impact of indirect fatalities. Moreover, I'm confident that although the historical record prior to August 2007 was limited to direct fatalities only, I suspect some of the official fatality counts for past tornadoes include a few indirect fatalities that crept into that record. It's probably not a big deal, but users should be aware of this issue.

Another concern is the follow-up on people who may die days, weeks, or even months following an event, as a result of injuries sustained from a storm. Although I believe that the current situation in this regard is probably being done with reasonable attention to this potential problem, I'm suspicious that such delayed fatalies likely have not been counted in at least many past events. The time and effort necessary for an accurate count of direct and indirect fatalities may not always be available for the local preparers of Storm Data. Generally, the responsibility for collecting information to be submitted to Storm Data currently is being done by the so-called Warning Coordination Meteorologist (WCM) at each office in the NWS. Since this is one person - who may or may not be able to obtain assistance in collecting the necessary information) and who has a number of other responsibilities in addition to preparing Storm Data - it's reasonable to assume that occasional mistakes of omission occur. The total number of such omissions likely would represent only a small part of the overall totals for the nation, but they do represent another quality concern for the users of Storm Data.

I should point out that in the past, Storm Data preparation was the duty of state climatologists employed by the Weather Bureau. Of course, they also were lone individuals in local weather offices who were certainly capable of errors and they certainly had additional duties. When the state climatologist positions in the National Weather Service (formerly the Weather Bureau) were abolished in the early 1970s, the responsibility for collection of Storm Data information fell to the local weather offices and eventually to the WCMs, (when WCM positions were created nationwide). I believe there are many good reasons to move that responsibility away from the NWS offices (e.g., as discussed here) and have the data collection done by people who are independent of the NWS. Later in this essay, I'll be discussing the topic of the resources necessary to do the task properly, but for the moment I'll suggest that the task of Storm Data preparation is underfunded and understaffed.

it should be evident that having an accurate count of fatalities is important for assessing societal impacts. Fatality counts are probably in reasonably good shape.

Whereas I believe fatality counts are likely to be reasonably accurate, I have serious doubts about the injury counts. According to the recently-revised Storm Data instructions

A directly-related weather injury is one that requires treatment by a first-responder or subsequent treatment at a medical facility. Injured persons who deny medical treatment also may be included. Persons who are not considered injured but who are affected by the phenomenon may be discussed in the event narrative.

I wonder to what extent first responders keep track of and report upon the number of people they treat. I assume that medical facilities keep records that can be retrieved and used to develop counts. The instructions don't seem to make clear just what they mean by "affected by the phenomenon". Does this include psychological trauma as a result of the "phenomenon"? It's also unclear just how things have been done in the past. Injuries are an importan societal impact, and the number of injuries is typically at least 10 times the number of fatalities in major events. Some events only cause injuries and no fatalities, of course. And there's the problem of indirect injuries. If someone breaks an ankle during the cleanup after a tornado, is that likely to be included in the records? I'd like to think so, but I have my doubts.

Again, accurate injury counts are needed to assess societal impacts, and it seems likely to me that these may not be as reliable as fatality counts. Especially when we consider the psychological problems. In my work with a project to study the "Tri-State" tornado of 18 March 1925, it was clear from interviewing the remaining living eyewitnesses that this event was still clear in their minds more than 80 years later. They told stories about how this influenced their lives for many decades after the event. We have anecdotal information, or information from single-event studies (like ours on the Tri-State tornado), but apparently not a comprehensive data base upon which to build an accurate picture of these impacts.


2. Damage

I've already written about this in my blog entry for 15 February 2009 - repeated here:

Lately, I've been wondering about the significance of the dollar damage estimates for tornadoes provided by, say, Storm Data. These numbers seem to come out within a few days of an event, and I'm pretty uncertain about their real meaning. I'm guessing that someone (it's unclear just who) has estimated the cost of structures damaged or destroyed by a tornado. Apart from uncertainties associated with these estimates (and I know little about those uncertainties), this seems to be a vastly oversimplified way to understand the economic impact of tornadoes.

Of late, we've been seeing a fascinating impact story on TV about Greensburg, Kansas (virtually wiped off the map by a tornado in 2007): the "greening" of Greensburg. This had been a small town whose main claim to fame was the world's largest hand-dug well. Now, it seems that the town is being rebuilt in a way to make it a model for an environmentally-conscious community. In the long run, the tornado might have had a positive impact on Greensburg! This is not to minimize the significance of the losses, but to suggest that, at least in economic terms, Greensburg might turn out to be substantially better off than they were before. The jury is still out on this, of course, but there's cause for some optimism.

In 1999, the tornado outbreak of 3 May in Kansas and Oklahoma hit the town of Stroud very hard. Its main industries included an outlet mall on the Interstate through town, a factory, and a hospital. Apparently, after being heavily damaged, the outlet mall and factory were not rebuilt, costing the town jobs and income. Where does that impact show on the ledger for that tornado? This is an obvious economic impact that isn't accounted for at all in the dollar damage estimates after the fact.

It seems to me that economics is another example of a complex, highly nonlinear system (as is the weather itself). Deep understanding is required to do a comprehensive analysis of the economic impact of a devastating natural hazard event. There are all sorts of costs that aren't accounted for in the damage done to structures. The cost of the physical and mental trauma, the loss of income when people aren't able to go to work, the business losses by being shut down along with inventory damaged or destroyed. Plus, there are unaccounted-for benefits for some elements within a community - hardware store business might experience a large positive surge in sales of the means by which damage can be repaired, construction businesses experiencing a boom, replacement of obsolete or worn-out infrastructure (personal or commercial) with the help of insurance, and so on. Infusion of insurance, as well as state and federal funds can offset some of the damage costs. There are short-term and long-term impacts, the latter of which certainly are rarely, if ever, considered and definitely are not included in Storm Data.

I'm sure there are many elements of a community economy about which I'm unaware - I'm not an economist, after all - so I'm confident that a complete accounting of any specific tornado's impact has never really been done, to say nothing of the impact of the 1000 or so tornadoes that occur in the USA every year. Therefore, this topic seems ripe for a collaboration with some real economists. Unfortunately, the small number of economists I've mentioned this to aren't really looking for additional projects to fill up their copious spare time! I can understand their reluctance to add something new to their table, but it's disappointing.

If we want to have a clear understanding of the economic impacts of tornadoes, and I see many reasons why we meteorologists should want this information, then this is a major project just waiting to be done. It seems likely I won't be able to be a part of it, but I sure hope someone can make this happen!

I have very little confidence in the numbers we have in Storm Data regarding the economic costs/losses from severe storms. Damage to structures is just a part of a much more complex picture. At present, I don't believe this is being given anything close to the comprehensive treatment it deserves. In the absence of knowledge about these complex economic impacts, it's hard to know just what to do with the current figures assigned to individual storm events in Storm Data. It's as if we were trying to understand the science of meteorology using only precipitation measurements at the Earth's surface.


3. Discussion

I accept the premise that meteorologists on their own are ill-equipped to assess societal impacts. We haven't the education and training in the methods that are needed to do a comprehensive analysis of societal impacts. But even more disturbing is that it might be quite difficult to obtain the data necessary to even begin doing such a wide-ranging analysis. It's common, whenever someone asks for more resources, to be challenged to show that the cost of those additional resources is less than the benefits that will accrue when those resources are provided. With more than 1,000 tornadoes every year, there can be no doubt that this represents an important phenomenon with considerable societal impacts. To develop a truly thorough analysis of those impacts would require considerably more resources invested in Storm Data collection than is happening now. Without having such a deep understanding at hand, unfortunately, it's difficult to know just what we could do to mitigate the most of the costs. Much of the physical damage (certainly not all) now caused by tornadoes can't be prevented. Thus, most of our attention would have to be devoted to minimizing casualties from tornadoes. As discussed here, it seems evident that improved warning lead times are not the only answer to the challenge of reducing casualties. Thus, a purely meteorological line of research is not going to be very fruitful, barring some unforeseen breakthrough.

As is often the case when it comes to weather, we have precious little information that could be used to justify additional costs for such a program as I'm describing. Given that Storm Data information is collected by the WCMs at local weather offices, it's possible to come up with some estimates of how many FTEs are associated with this task. There are roughly 120 local NWS offices, and I'm going to guess that an average involvement time with collecting information for Storm Data is 10% of an FTE per office. Obviously, some individual WCMs might spend more or less time than this, but I think it's a plausible guess. This is the rough equivalent of 12 people engaged in this task full-time. I could make an argument that if this was someone's only task, they likely would be better at it and perhaps more efficient than someone who can only do the job when time permits the chance to get away from other duties.

Let's suppose we were to create a group outside of the NWS whose sole responsibility was to collect information for a data base about storms that was as comprehensive as I've imagined. How many more FTEs would that require and what sort of people would be needed? Such a team would include meteorologists, of course, but would also need the services of sociologists, economists, psychologists, geographers, etc. And it would be important to have more than one person from each discipline - otherwise, the prejudices of a single individual could contaminate the program results - that is, the data base. I'm guessing that a team of professionals (presumably, most would have Ph.D. - level education) would require 20-25 FTEs, plus support staff (at least an additional 10-15 FTEs) to do the job. It might turn out that this is an underestimate, of course. Such a program would consume considerable travel support, in order to collect the information - many of their professionals have to be on the road a significant fraction of the time, doing data collection in the wake of storm events. To some extent, this program would be comparable to the FAA's aircraft accident investigating team, but with a much broader focus, requiring more than just damage investigation professionals. Some activities would have to take place before storms, to establish baseline pre-storm information. They'd also need a physical facility with sufficient IT capability (hardware plus IT staffing) to manage what rapidly would become a large data base.

What I'm describing sounds like the equivalent of an entire NOAA Research laboratory. Of course, NCDC currently "houses" the Storm Data program, but they have only minimal staffing to deal with it at the moment - the job is being done for them, for the most part, by the WCMs in the local offices. In effect, what I'm describing would, for all practical purposes, be the creation of a new NOAA Research lab devoted entirely to collection and management of comprehensive data about storms in the USA. In practical terms, the odds of creating such an entity any time soon are vanishingly small, especially in the current state of economic collapse in the USA.

I discussed this concept in a context limited to tornadoes here, under the assumption that simply saving lives could be used to justify an additional expenditure of hundreds of millions of dollars annually. Presumably, a program to address the full diversity of storm casualties could support on the order of a billion dollars additional annual expenditures on behalf of data collection and research into mitigating societal impacts from storms. That amount would buy a pretty formidable NOAA research laboratory! Research into societal impacts from storms would virtually be certain to reveal important new ways to reduce the costs to society caused by storms. We currently don't know about those cost reduction measures because we don't have the interdisciplinary knowledge to grasp the true costs to our society that result from storms, and we don't even have the data to do the job properly.

In my opinion, the major thing lacking is the vision within the bureaucracy and the political will to do something about the costs to our society represented by storms. In order to do something that would be effective, this data collection and research based upon it would be necessary and quite likely to be easily justified on the basis of results leading to benefits far in excess of costs in a short time.


(Most publications with me as an author are available here)

Kelly, D.L., J. T. Schaefer, R.P. McNulty, C.A. Doswell III, and R.F. Abbey, Jr., 1978: An augmented tornado climatology, Mon. Wea. Rev., 106, 1172-1183.

Kelly, D.L., J.T. Schaefer, and C.A. Doswell III, 1985: Climatology of nontornadic severe thunderstorm events in the United States. Mon. Wea. Rev., 113, 1997-2014.

Doswell, C.A. III, and D.W. Burgess, 1988: On some issues of Unites States tornado climatology. Mon. Wea. Rev., 116, 495-501.

Doswell, C.A. III and H.E. Brooks, 1998: Budget cutting and value of weather services. Wea. Forecasting, 13, 206-212.

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.

Brooks, H.E., and C.A. Doswell III, 2002: Deaths in the 3 May 1999 Oklahoma City tornado from a historical perspective. Wea. Forecasting, 17, 354-361.

Brooks, H.E., C.A. Doswell III, and M.P. Kay, 2003: Climatological estimates of local daily tornado probability for the United States. Wea. Forecasting, 18, 626-640.

Doswell, C.A. III, 2003: Societal impacts of severe thunderstorms and tornadoes: Lessons learned and implications for Europe. Atmos. Res., 67-68 ,135-152.

Doswell, C.A. III, 2005: Progress toward developing a practical societal response to severe convection (2005 EGU Sergei Soloviev Medal Lecture). Nat. Haz. Earth Sys. Sci., 5, 691-702.

Doswell, C.A. III, H.E. Brooks, and M.P. Kay, 2005: Climatological estimates of daily nontornadic severe thunderstorm probability for the United States. Wea. Forecasting, 20, 577-595.

Doswell, C.A. III, 2007: Small sample size and data quality issues illustrated using tornado occurrence data. Electronic J. Severe Storms Meteor., 2 (5), 1–16.

Doswell, C.A. III, H.E. Brooks, and N. Dotzek, 2008: On the implementation of the enhanced Fujita scale in the USA. Atmos. Res., doi:10.1016/j.atmosres.2008.11.003 [in press]