Extreme statistics: a new toolkit helps profile the far ends of weather
Atmospheric outliers—a snowstorm in Guadalajara, say, or a deadly heat wave in England—often get left behind in statistical analysis. NCAR’s Environmental and Societal Impacts Group (ESIG) now provides online tools for researchers looking at extreme events, the ones that loom largest in their effects on Earth and people.
On 8 May 2003, a tornado rated F3 on the Fujita-Simpson scale slammed into some of the same parts of Moore, Oklahoma, that had been raked by an F5 tornado on 3 May 1999. (Photo by Carlye Calvin.)
ESIG scientist Rick Katz launched the Statistics of Weather and Climate Extremes site as a way to reach out from statistical specialists to the larger group of atmospheric researchers who seek better ways to account for extremes in their analyses. “Unfortunately,” says Katz, “many researchers are unaware that the statistical methods appropriate for analyzing extremes differ fundamentally from the more familiar ones for averages.” The good news, he adds, is that “the statistical theory of extreme values has been developed for quite a while. With the computational advances and software developed in recent years, the application of this theory to weather and climate has become relatively straightforward.”
The site includes a short history of extreme-value theory and some of its founders; a set of references and links, including relevant software; and the Extremes Toolkit, a package originated by Gregory Young, formerly of ESIG. The toolkit, completed and now maintained by NCAR’s Eric Gilleland, consists of functions written in an open-source programming language called “R” and designed to perform extreme value analysis. Many of these functions are provided courtesy of Stuart Coles (University of Bristol). Knowledge of R is not required, as a graphical user interface is provided. A tutorial explains how the toolkit can be used to treat weather and climate extremes in a more accurate manner.
Katz and colleagues developed the Web site with NSF support through NCAR’s Geophysical Statistics Project, as well as the center’s Weather and Climate Impact Assessment Science Initiative, which aims to improve the treatment of extremes and uncertainties in climate-change research.
“Although much discussion in recent years has been devoted to the importance of extreme events in climate change, it is quite rare that modern statistical theory for extremes is applied,” says Katz. Such theory, he adds, “would allow for more systematic study of extremes, as opposed to the present ad hoc treatment of a few somewhat arbitrary events.”