Tornadoes and other severe thunderstorm phenomena frequently cause as much annual property damage in the United States as do hurricanes, and often cause more fatalities (see http:// www .nws .noaa .gov/ om/ hazstats .shtml). In 2008, there were 2176 preliminary tornado reports logged through mid-December, with 1600 "actual counts" (duplicate reports removed) through September, the highest total in the past half century ( Figure 1). The mass media have covered these events extensively, and experts have been deluged with requests for explanations, including possible links to anthropogenic global warming. Although recent research has yielded insight into the connections between global warming and tornado and severe thunderstorm forcing, these relationships remain mostly unexplored, largely because of the challenges in observing and numerically simulating tornadoes. Indeed, a number of questions that have been answered for other climate-related phenomena remain particularly diffi cult for climate and severe weather scientists, including whether there are detectable trends in tornado occurrence and if so, what causes them. This article explores the challenges and opportunities in pursuing these areas of research.
Tornado Trend DetectionThe number of tornadoes reported in the United States per year has been increasing steadily (~14 per year) over the past half century (Figure 2). However, determining whether this is a robust trend in tornado occurrence is diffi cult. In the United States, the historical record is both relatively short (with the offi cial collection of tornado reports beginning in the 1950s) and nonuniform in space and time [see Brooks and Dotzek, 2008]. Further, the observational record of tornadoes is based on a reporting system designed essentially for the verification of forecasts rather than for research-quality climate studies, meaning that maintaining consistency throughout the historical reporting record has not been a specific priority. Reliance on human reports leaves trends subject to external influences such as population growth, although some regions such as the southern Great Plains do not reflect the congruous long-term trends in tornado occurrence and population growth seen throughout the entire United States (Figure 2).Interestingly, the number of tornadoes classifi ed as the most damaging (rated F2-F5 on the Fujita scale) actually appears to have decreased over the past fi ve decades (Figure 2), although this trend could be spuriously infl uenced by "overrating" the severity of tornadoes in the fi rst half of the record [e.g., Verbout et al., 2006]. Trends in tornado intensity are further complicated by the fact that intensity is equated with a humanbased assessment of damage severity and is therefore modulated by the presence or absence of structures, vegetation, or other sources of debris. Other nonphysical factors infl uencing the tornado record include the organization of networks of volunteer "storm spotters," the enhancement of community education and awareness programs, develo...