Trends in U.S. Extreme Heat Indices
FY99 Office of Oceanic and Atmospheric Research Operating Plan Milestone Report
Submitted by Dian Gaffen, Air Resources Laboratory
Milestone Details
Operating Plan Objective 3: Ensure a Long-term Climate RecordSubject: Climate Monitoring
Q3: Determine long-term trend in heat index as experienced by the U.S. population. (ARL, D. Gaffen)
Q3: Analyze trends in the frequency of extreme heat events for the period 1949-95 using NWS hourly surface meteorological observations as archived by NESDIS/NCDC. The analysis will examine: (1) extremes of both temperature and apparent temperature (factoring in humidity) relative to local climatology to reflect differences in human acclimatization and (2) trends in the frequency of heat waves. The study should contribute to analyses of potential threats to public health, particularly among the elderly. (ARL, D. Gaffen)
Background
Among the possible effects of global warming associated with increased atmospheric greenhouse gas concentrations is an increase in the frequency of extremely hot weather events. Because extremes of summertime heat are thought to have a greater impact on human health than any other form of severe weather in the U.S. (Changnon et al., 1996), more frequent occurrences of extreme heat would have important public health implications. Heat waves can lead to heat stress, exacerbated illness, and death, and the elderly are at greatest risk. The notable July 1995 Chicago heat wave, for example, is thought to have caused 465 deaths (WHO, 1996).Discussion
Previous analyses of U.S. climatological records during the 20th century have revealed increases in mean surface temperature, with greater warming at night than day (e.g., Easterling et al., 1997). During the past several decades, the warming evident in monthly mean temperatures has been accompanied by an increase in monthly mean atmospheric humidity, which is also more pronounced at night (Gaffen and Ross, 1999). These surface humidity trends are broadly consistent with trends in lower-tropospheric water vapor evident in radiosonde data (Ross and Elliott 1996).It is natural to ask whether these trends in monthly-mean conditions are associated with trends in the frequency of days with extremely high temperature and humidity, as would be expected if the probability distribution functions of both variables maintain the same shape but shift towards higher mean values in the mean. Because people are acclimatized to local conditions, extremes that effect human health vary from one city to another (WHO, 1996) and so must be defined locally. Furthermore, the human response to summertime heat depends not only on temperature but also on ambient humidity. The apparent temperature, also known as the heat index, measures the combined effects of temperature and humidity on human comfort (Steadman, 1984).
Accomplishments
The following accomplishments represent completion of this milestone.1. Dataset development and distribution
We have developed and analyzed a dataset, based on hourly surface meteorological observations from National Weather Service first-order stations (and archived by NCDC), for the period 1948-95, for the purpose of determining trends in extreme heat events. The data are available on the Air Resources Laboratory web site (www.arl.noaa.gov/ss/climate).To define extremes, threshold values of temperature (T) and apparent temperature (A) were computed based on the climatological data. The thresholds are values that are exceeded on only 15% of days in July and August and are therefore indicative of local extreme summertime conditions. The number of days on which the thresholds were exceeded each year was used to delineate trends. Separate thresholds for A and T, and for daily-average, daily-maximum, and daily-minimum conditions were evaluated. The thresholds for daily-average A (shown in Fig. 1) reveal an expected pattern in which summertime heat is more severe in the southeast and south-central U.S., and in the desert southwest.
Fig. 1. Threshold values of daily-average apparent temperature in the U.S. Days on which the daily-average apparent temperature exceeded these thresholds were considered extreme. The thresholds are the 85th percentile values of daily-average apparent temperature computed from thirty years of July and August data. Similar thresholds were computed for daily minima and daily maxima, and for temperature as well as apparent temperature.
These thresholds, based on climatological data, are very closely associated with thresholds independently determined to be associated with increased mortality in various U.S. cities, as shown Fig. 2.
Fig. 2. A comparison of the Gaffen and Ross thresholds of daily-maximum temperature and daily-average apparent temperature with independently determined thresholds based on analysis of human mortality data by L. Kalkstein and co-workers. The latter are thresholds at which excess deaths due to heat were identified. Each point represents a different U.S. city. The strong correlation between the two thresholds indicates the utility of the climatology-based thresholds for understanding human health impacts, even at locations for which mortality-based thresholds are not available.
2. Trend analysis
For each of 113 stations with complete data records, 1949-95 trends in the frequency of extreme heat stress were computed. Both single-day events and multi-day heat waves were considered. Trends were upward over most of the U.S. Trends in nighttime extremes were generally larger than for daytime, consistent with trends in mean conditions. Apparent temperature trends were generally larger than temperature trends because of the influence of humidity increases. As shown in Figs. 3, 4, and 5, trends were largest and most statistically significant in the eastern and western thirds of the country. The increases represent an approximate doubling of the frequency of extreme heat events over the past half century.
Fig. 3. Map of station decadal trends in the frequency of extreme daily-minimum apparent temperature for 1949-95. The sign of the trend is indicated by the orientation of the triangle (apex up for positive trends, down for negative), and the magnitude is indicated by the size and color. Filled in symbols represent trends that are statistically significant at the 95% confidence level. Decadal trends of about 2 days/year per decade represent about an 80% increase over the study period, relative to mean frequencies of about 12 days/year.
Fig. 4 Time series of regional averages of extreme daily-minimum apparent temperature. Note the stronger trends in the East and West compared with the Central US.
Fig. 5 Time series of the regional averages of the number of three day heat waves. The trends indicate an approximate doubling over the study period.
3. Publication and communication
These results were published in Nature magazine (Gaffen and Ross, 1998). As a result of press releases from Nature and NOAA Public Affairs, the publication generated considerable media attention. Newspaper stories appeared in the New York Times (front page), Washington Post, USA Today, Miami Herald, Boston Globe, and numerous papers that ran articles by Associated Press and Reuters. Radio interviews were broadcast by Voice of America, WWDB radio (Philadelphia), and the Great Lakes Radio Consortium (of National Public Radio stations).Next Steps
A comparable analysis of surface meteorological data from China is underway. Trends in extreme heat events there may have even more serious implications than in the U.S. because air conditioning is not as widespread. We are also investigating U.S. trends in the frequency of extreme cold events, which may be related to growing season length and therefore to agricultural production and ecosystem health.Significance
A small increase in mean surface temperature is one predicted effect of an enhanced atmospheric greenhouse effect, and its detection is important for identifying the nature and causes of climate change. However, human comfort and health, as well as ecosystems, are more directly influenced by extreme than mean conditions. Until now, little was known about the multi-decadal variations in extreme summertime heat and humidity over a large region. This study contributes to further understanding the nature of observed climate changes and their impact on society.References
Changnon, S.A., Kunkel, K.E., Reinke, B.C., 1996: Impacts and responses to the 1995 heat wave: A call to action. Bull. Amer. Meteor. Soc., 77, 1497-1506.Easterling, D.R., B. Horton, P.D. Jones, T.C. Peterson, T.R. Karl, D.E. Parker, M.J. Salinger, V. Razuyev, N. Plummer, P. Jamason, C.K. Folland, 1997: Maximum and minimum temperature for the globe. Science, 277, 364-366.
Gaffen, D.J., and R.J. Ross, 1998: Increased summertime heat stress in the US. Nature, 396, 529-530.
Gaffen, D.J., and R.J. Ross, 1999: Climatology and trends in U.S. surface humidity and temperature, J. Climate, 12, in press.
Ross, R.J., and W.P. Elliott, 1996: Tropospheric water vapor climatology and trends over North America: 1973-1993. J. Climate, 9, 3561-3574.
Steadman, R.G., 1984: A universal scale of apparent temperature. J. Clim. Appl. Meteor., 23, 1674-1687.
World Health Organization, 1996: Climate Change and Human Health, A.J. McMichael et al., eds., WHO, Geneva, WHO/EHG/96.7, 297 pp.
Prepared: 6 January 1999