4.6 Changing Heat Wave Sensitivity in U.S. Cities

Tuesday, 21 June 2005: 4:00 PM
South Ballroom (Hilton DeSoto)
Robert E. Davis, Univ. of Virginia, Charlottesville, VA; and P. C. Knappenberger, P. J. Michaels, and W. M. Novicoff

Mortality associated with heat accounts for most of the weather-related deaths in the United States. Most of these deaths occur in a few isolated heat wave events during the late spring and summer. Thus, a careful examination of the linkages between heat wave properties and human mortality is required, including temporal changes in the heat wave climatology and possible changes in heat-wave impacts.

Prior and ongoing work by our research group has identified statistically significant declines in heat-related mortality in many “heat-sensitive” U.S. cities despite increasing heat and humidity. This apparent contradiction suggests that society has (actively or passively) adapted to increasingly stressful climatic conditions. Preliminary research reported last year for a single metropolitan area likewise indicates that mortality during prolonged heat events has shown a similar response.

This research provides a more extensive analysis of heat-wave mortality in a variety of climate zones in five large U.S. cities: Philadelphia, St. Louis, Chicago, Seattle, and Tampa. We hypothesize that heat-wave related mortality has not changed or perhaps even delined over time despite increasing strength and/or intensity of heat waves in these cities.

Our data includes daily all-causes mortality acquired from National Center for Health Statistics archives from 1964–1998, excluding the period from 1967–1972 when the date of death was not retained. Daily mortality totals are standardized to a base population and long-term trends and seasonality are removed via a 60-day moving average filter. Data are collected for the counties associated with the Metropolitan Statistical Area (MSA) for each city. Weather data from the nearest first-order station in each MSA are acquired, and 7 a.m. LST apparent temperature (AT) is calculated for use as the independent variable in all analyses.

Following a method established in our earlier work, “heat events” are defined as two or more consecutive day runs of above normal ATs (with normal defined by the 30-day moving average). Then, each heat event is categorized into four possible categories: killer heat waves, non-killer heat waves, high mortality normal heat, and normal mortality normal heat.

Preliminary results show that mortality during “killer heat waves” has declined significantly over the period of record in our “heat-sensitive” MSAs: Philadelphia, St. Louis, and Chicago. In Seattle and Tampa, heat waves have less of an impact on daily mortality and the results are less consistent. Furthermore, there is also evidence of within-season adaptation in the mortality data. In years with multiple heat waves, the first or second heat wave has the greatest mortality impact regardless of the month in which it occurred. In addition, short heat events were not typically associated with high mortality. On average, events needed to be at least four days in duration to be classified as “killer heat waves.”

A further preliminary finding of interest is that high mortality is more closely linked to absolute rather than relative AT departures. This runs counter to current orthodoxy on weather-mortality relationships that proposes that relative discomfort is more critical than absolute temperature/humidity conditions in determining adverse responses to extreme heat.

Future research will continue in an effort to provide a comprehensive heat-wave and mortality climatology for all major U.S. metropolitan areas using the methods developed herein.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner