A comparison of biometeorological comfort indices and human mortality during heat waves in the United States.

Numerous biometeorological comfort indices have been developed to examine the relative or absolute stress placed on individuals via their interaction with their atmospheric environment. Although most of these indices are founded upon consistent physical principles related to heat and moisture fluxes from the body, the indices nevertheless vary widely based upon underlying assumptions, differing moisture parameterizations, etc. Our goal is to compare a suite of these comfort indices with respect to how well they identify high mortality heat waves in a sample of U.S. cities.

Mortality data include daily all-causes mortality (adjusted to a standard population) from several U.S. metropolitan areas where a demonstrable summer mortality response to heat has been identified from 1964—1998. Hourly surface weather observations are used as input to calculate the following indices: apparent temperature, predicted mean vote, physiological equivalent temperature, perceived temperature, relative strain, and standard effective temperature. Air temperature and relative humidity are also included for comparison purposes to examine the "value added" by each index.

Two types of heat waves are identified: absolute heat waves, in which a fixed threshold index value is exceeded; and relative heat waves, which are identified by departures from normal. Each heat wave type is then characterized as either a "killer" or "non-killer" heat wave based upon the total mortality associated with each event.

For each index, total heat wave mortality is plotted versus class intervals of the dependent (index) variable to determine the threshold beyond which mortality increases significantly. The ability of the indices to identify "killer heat waves" is compared using contingency analysis via a suite of accuracy measures and skill scores. Furthermore, indices are compared on the highest mortality days to determine the extent to which they identify summer mortality events.

Most of the comfort indices successfully identify high mortality heat waves, particularly the absolute heat waves the typically occur in July and August. As most of these indices are highly influenced by the combination of air temperature and moisture, the differences between indices are dependent upon the marginal utility gained from the addition of other parameters related to radiation and wind as well as different assumptions and parameterizations. Clearly relative humidity alone is a poor index, although simple air temperature competes reasonably well with some of the more complex approaches. A more detailed analysis attempting to identify the root cause of the generally small differences in skill between the indices has thus far been inconclusive.

Much more work is required to examine the spatial variability of these results to determine if they are consistent in different climatic regions. Furthermore, it remains important to identify precisely why some indices outperform others in some situations so biometeorologists can improve these models. This final result might remain elusive, however, because of inherent noise in the mortality signal that arises from non-climatic factors.