Concerns over climate change impacts in the United States include the proposition that human death rates will increase (e.g., U.S. National Assessment). These predictions are primarily based upon projections of longer, more intense heat waves, more frequent or more intense isolated hot days, or changes in air mass frequency or characteristics that would enhance heat stress in humans.

The research focus has been on the warm season, when daily deaths in certain areas spike in direct response to hot days or prolonged heat waves. However, in all U.S. cities, total deaths are higher in winter. Since observed temperatures since World War II have predominantly increased in the cold season, their impact on mortality in winter could theoretically be more significant than the warm-season effects. In this research, we examine the variability in the seasonality of temperature and human mortality in 27 U.S. cities.

Counts of daily mortality are compiled from National Center for Health Statistics archives for all available years from 1964–present and are age-adjusted relative to a standard population. Weather data are acquired from first-order surface observations for each city. Data are aggregated on a monthly basis to allow for both intercity comparisons across months (e.g., how does mean daily mortality change from month-to-month between Boston and New Orleans?) and spatial comparisons within months (e.g., how does January or July mortality differ between northeastern, southern, and western cities?).

Preliminary results exhibit a complex pattern of natural monthly variability that is consistent across many cities coupled with variability that is region-specific. In all cities, mortality is higher in winter than in summer and no location has below average mortality from December through March. Similarly, all cities examined exhibit below average death rates from July through September.

Linear regressions between mean monthly mortality anomalies and 7 a.m. average temperature, run separately for each city, identify two categories of cities based upon the slope of the regression line (or the month-to-month mortality variability with temperature). The steepest slopes occur in eight southern or western cities with either mild summers, mild winters, or both. These locations have relatively high winter mortality although their winters are of short duration. The remaining 19 cities have comparable and much flatter regression slopes. In these cities, which typically have a summer mortality spike on hot, humid days, winter mortality is lower per month but winter has a longer duration. For these cities, in summer, there is evidence for a “mortality optimum temperature,” or the temperature beyond which the negative relationship between mean monthly mortality and temperature is lost. For most cities, this is a mean monthly 7 a.m. temperature of approximately 15°C (68°F).

These preliminary results suggest that mortality prediction models based upon climate change must account for geographic, demographic, and climatological factors that vary from city to city. The observation that warm cities experience enhanced mortality during their short winters relative to cities in colder climates suggests that annual mortality rates are largely independent of seasonal climate.