CIESIN Reproduced, with permission, from: Kalkstein, L. S., R. E. Davis, J. A. Skindlov, and K. M. Valimont. 1986. The impact of human-induced climatic warming upon human mortality: A New York City case study. In Climate change. Vol. 3 of Effects of changes in stratospheric ozone and global climate, ed. J. Titus, 273-93. Proceedings of the United Nations Environment Programme (UNEP)/Environmental Protection Agency (EPA) International Conference on Health and Environmental Effects of Ozone Modification and Climate Change. Washington, D.C.: U.S. Environmental Protection Agency.


Volume 3: Climate Change

Edited by

James G. Titus

U.S. Environmental Protection Agency

The Impact of Human-Induced Climatic Warming Upon Human Mortality: A New York City Case Study

Laurence S. Kalkstein, Robert E. Davis Jon A. Skindlov, Kathleen M. Valimont Center for Climatic Research University of Delaware Newark, Delaware USA


The goal of this study is to determine if weather has an impact on mortality in New York City and to ascertain whether expected future climatic warming will alter the death rates significantly. Summer weather appears to have a significant impact on New York's present mortality rates, and a "threshold temperature" of 92deg.F was determined, suggesting that mortality increases quite rapidly when the maximum temperature exceeds this value. Days with high minimum temperatures, long periods with temperatures above the threshold, and low relative humidities appear to increase mortality most dramatically. Five climatic scenarios were developed to estimate New York's future weather assuming that warming does occur, and "acclimatized" and "unacclimatized" mortality rates were estimated for each scenario. The unacclimatized rates were computed by using New York's weather/mortality algorithm developed from the historical analysis. Acclimatized rates were computed by selecting present-day "analog cities" which resemble New York's predicted future weather and by developing weather/mortality algorithms for them.

Results indicated that the number of additional deaths at temperatures above the threshold could increase by over tenfold if New Yorkers do not become acclimatized to the warming. The elderly will constitute an increasing proportion of these deaths. However, if full acclimatization occurs, the number of additional deaths above the threshold temperature could be no different than today. No similar relationships were discovered for the winter, and the data suggest that any changes in winter weather will have minimal impact on New York's mortality rates. A preliminary precipitation/mortality study was undertaken, and summer days following a precipitation event had significantly lower mortality rates than summer days without precipitation. In the winter these results were reversed, and days with rain (but no snow) had significantly higher mortality rates than nonprecipitation days.


A procedure has been developed recently to evaluate the impact of longterm climatic warming on inter-regional variations in human mortality. Fifteen cities around the country are presently being evaluated, possible future climatic scenarios are being developed for each, and estimates of possible weather-related changes in mortality are being calculated.

The objective of this report is to describe our procedure and to apply it to one of our fifteen cities, i.e., New York City, New York. The impact of present-day weather on New York's present mortality rates is discussed, and estimates are presented describing the potential impact of climatic warming on New York's future mortality.

Although no previous study has attempted to predict the impact of future weather changes on mortality, there has been considerable work relating to present climate/mortality relationships. For example, studies at the Centers for Disease Control have identified a number of factors that may accelerate the onset of heat stroke, including decreases in use of air conditioning, consumption of fluids, and living in well-shaded residences (Kilbourne et al. 1902). However, other researchers have found that many causes of deaths other than heat stroke increase during extreme weather (Applegate et al. 1981; Jones et al. 1982). In addition, it has been shown that mortality attributed to weather varies considerably with age, sex, and race, although there is disagreement among researchers in defining the most susceptible population group (Oechsli and Buechley 1970; Bridger, Ellis, and Taylor 1976; Lye and Kamal 1977; Jones et al. 1982). The impact of cold weather is less dramatic than hot weather, although mortality increases have been noted during extreme cold waves (Centers for Disease Control 1982; Fitzgerald and Jessop 1982; Gallow, Graham, and Pfeiffer 1984).

This study will incorporate some approaches used in previous studies while offering a new approach to account for potential changes in mortality/weather relationships that might be attributed to acclimatization.


A very detailed mortality data base is presently available from the National Center for Health Statistics (NCHS), which contains records for every person who has died in this country from 1964-present (National Center for Health Statistics 1978). The data contain information such as cause of death, place of death, age of death, date of death, sex, and race. These data were extracted for the New York Standard Metropolitan Statistical Area for 11 years: 1964-66, 1972-78, and 1980 (during intervening years, a sizable amount of information was missing from many records). The number of deaths for each day were tabulated and divided into categories of total deaths and elderly deaths (65 years and older). These daily death totals were standardized to conform to a hypothetical "standardized city," which contains fixed population characteristics (Table 1). The death rates for New York were adapted to the population characteristics of the standardized city to conform to procedures commonly found in the epidemiological literature (Mausner and Bahn 1974; Lilienfeld 1980). The advantages of this standardization procedure are twofold. First, when the study is extended beyond New York, inter-city comparisons will be feasible since demography is kept constant. Second, if a city has grown rapidly during the study period, the bias introduced by the increase in deaths that are due to population growth is eliminated, and changes in mortality attributed to environmental factors can be better assessed.

Apparently weather does have some impact on daily mortality (Figure 1). During the heat wave of late July 1980 in New York, standardized deaths rose by over 50% above normal on the day with the highest maximum temperature. Elderly deaths showed similar increases. In this study, daily changes in mortality were compared to nine different weather elements that might have some influence on death rates (Table 2).

Initial observations of daily standardized deaths vs. maximum temperature suggest that weather has an impact only on the warmest 10%-20% of the days; however, the relationship on those very warm days is impressive (see Figure 2). Figures similar to Figure 2 were developed to compare the maximum temperature on the day of the deaths, as well as one, two, and three days prior to the day of deaths to determine if a time-lag exists between weather and the mortality response. In the case of New York, there is a one-day lag between weather and mortality. In addition, a "threshold temperature," which is the maximum temperature above which mortality increases, can be determined. The threshold temperature can be calculated objectively by using a sum of squares technique (Kalkstein 1986). The threshold temperature for total deaths in New York was 92deg.F; mortality increased dramatically at temperatures above this level. This procedure can be repeated for winter, where the threshold temperature represents the temperature below which mortality increases.

Once the threshold was established, a multiple regression analysis was performed using the weather elements described previously to determine the weather/mortality relationship for days above the threshold temperature. When a statistically meaningful relationship was determined, an algorithm was developed and used to predict the expected increases in mortality at temperatures above the threshold.

The next step was an attempt to estimate changes in New York's mortality that might occur with the predicted climatic warming. In consultation with EPA and the NASA-Goddard Institute for Space Sciences, investigators developed future weather scenarios for New York by adding temperature increments to existing historical New York temperatures. These scenarios were created for the period recorded by adding 1deg., 2deg., 4deg., 5deg., and 7deg.F to the existing weather data. This produced an approximation of what New York's temperature regime could be over the next 100 years. New mortality estimates were created for each of the temperature increments by using the algorithm developed from the historical data evaluation.

When measuring the impact of warming on future mortality, the question of acclimatization had to considered. Will New Yorkers react to heat as they do today, or will their reaction be similar to people who presently live in hotter climates? There is much disagreement in the literature concerning human acclimatization to changing weather. Some research indicates that acclimatization responses are very rapid (Rotton 1983); others think that it is a much slower process (Ellis 1972; Kalkstein and Davis 1985), and a few suggest that virtually no acclimatization occurs at all (Steadman, 1979). It is obvious that the full range of possibilities must be examined in this study. First, the historical algorithm that was developed from the previously described multiple regression procedure was applied to the future weather scenarios with the incorporated incremental increases in temperature. The mortality increases estimated from this procedure imply no acclimatization because an assumption is made that New Yorkers will respond to heat in the future in much the same way that they do today. Second, analog cities for New York were established to account for full acclimatization. For example, by adding the temperature increment to New York's present temperature regime, its weather will approximate another city's present weather in the U.S. A 5deg.F increment added to New York's present summer temperatures will yield a regime approximating that of Norfolk, Virginia, today. Since Norfolk residents are fully acclimatized to this regime, the weather/mortality algorithm developed for Norfolk can be utilized for New York to account for full acclimatization when New York's temperatures rise by 5deg.F.

Present-day analogs to account for full acclimatization were selected for New York for the 1deg., 2deg., 4deg., 5deg., and 7deg.F increments, and mortality models similar to the one described for New York were created for them. The analog cities were determined by computing for the three summer months (June, July, August) mean maximum temperatures, mean minimum temperatures, and mean number of days with maximum temperatures over 90deg.F for over 100 cities in the United States. The city that best duplicated New York's regime was established as an analogue city. This was achieved objectively using a variety of statistics for model evaluation (Willmott et al. 1985).

Figure 3 illustrates the hypothetical differences expected in mortality with full and no acclimatization. It is probable that the warmer analogs will show smaller increases in mortality than the original New York model since residents are already acclimatized to the increased warmth. Thus, for warming scenarios of 7deg. or more, the differences in predicted deaths between full and no acclimatization situations may be very large (area hatched between lines 1 and 2). In certain cases, it is possible that no extra deaths will be predicated for full acclimatization, as residents will be conditioned to hot weather. For example, in Jacksonville, Florida, heat waves appear to produce no extra deaths (see Figure 4). The relationship is so poor that it is almost impossible to determine a threshold temperature.


The multiple regression analysis to determine those weather elements having the greatest impact on present-day mortality in New York produced a surprisingly strong relationship (Table 3). Minimum temperature, maximum dewpoint, and heating degree hours (HDH) were all highly statistically significant and explained almost 66% of the variance in mortality at temperatures above the threshold. The most offending days appeared to possess high minimum temperatures, high HDH values, and low dewpoints, indicating that hot, dry conditions in New York appear most conducive to rises in mortality. The results from the evaluation of the elderly were similar, and the explained variance was slightly higher. Thus, it appears that predictive algorithms can be developed to estimate mortality in New York at temperatures above the threshold. These algorithms were also used to estimate unacclimatized deaths in New York using each of the warming scenarios.

Next, the analog cities were determined, threshold temperatures were calculated, and multiple regressions were developed for each (Table 4). As expected, the relationships became progressively worse for the analog cities representing the warmest scenarios, and the lack of a weather/mortality relationship for Norfolk and Jacksonville indicated that people in those cities were not sensitive to even the warmest temperatures because they are fully acclimatized to the frequent heat. Thus, there would be no expected increase in mortality in New York for the 5deg. and 7deg.F scenarios if the people become fully acclimatized. Note that threshold temperatures were higher for the warmer analog cities, supporting the contention that the impact of weather on mortality is relative on an inter-regional scale.

The number of deaths predicted from the nonacclimatized New York algorithm increased very rapidly with each succeeding warming scenario. One of the reasons for this was the increasing number of days exceeding New York's threshold temperature of 92deg.F for the warmer scenarios (Table 5). At present, the average monthly percentage of days exceeding this threshold is 3.3% in June, 10% in July, and 3.6% in August. Thus, for an average summer season, only 5.7% of the total days exceed the threshold temperature. These percentages increase steadily as the predicted warming increases, and for the 7deg.F scenario, almost half of the days in July and over one-third of the days in the entire summer season exceed the threshold. Obviously the total number of days with heat-related increases in mortality will also increase if there is no acclimatization.

A comparison of expected mortality increases for all age groups with no and with full acclimatization showed dramatic differences (Table 6). At present in New York, the average number of additional standardized deaths that occur on days above the threshold temperature each month is 19 in June, 86 in July, and 25 in August (the raw, unstandardized totals for New York are considerably higher, but these figures should be used with caution). Using the algorithm for no acclimatization, these figures more than doubled with a 2deg.F rise in temperature, and increased by more than tenfold with a 7deg.F rise. Thus, if New Yorkers do not acclimatize to the increasing warmth, it is predicted that the average number of additional standardized deaths will exceed 1300 each summer season if the weather warms by 7deg.F (the raw totals will exceed 3200). The full acclimatization results showed much different trends. The seasonal number of standardized deaths remained virtually constant with a 1deg.F rise (this was calculated using the present-day Indianapolis algorithm, which represents New York's analog city for a 1deg.F rise), and rose only slightly with a 2deg.F increase (Philadelphia's algorithm). However, for the warmer scenarios, the acclimatized deaths dropped sharply, and no additional deaths were predicted at 5deg. and 7deg.F warming. These results reflect the present-day Norfolk and Jacksonville situation, where no additional deaths are noted at temperatures exceeding the threshold.

The actual number of deaths attributed to future warming will fall somewhere between the predicted values for nonacclimatization and full acclimatization, but the precise impact of future acclimatization is obviously unknown. We suggest that a lag in acclimatization to climatic change is likely, and factors such as the physical composition of the city (i.e., building construction designed to accommodate present weather conditions) will delay or prevent full acclimatization. Thus, it is improbable that New Yorkers will become as totally insensitive to hot weather as Jacksonville residents are today, and the decrease in weather-related mortality predicted by the full acclimatization model is highly unlikely.

Predicted mortality increases for the elderly show similar trends (Table 7). Very large increases are noted with no acclimatization (raw unstandardized values are not provided, as they are partially dependent upon demographic information which is unknown, such as the proportion of population in the elderly category when a 7deg.F rise is achieved). However, deaths once again decrease to zero with full acclimatization. There is some indication that the elderly will constitute an increasing proportion of the total mortality as the climate warms (Table 8). At present, the percentage of the standardized mortality that is attributed to the elderly at temperatures above the threshold is 64% in June, 70% in July, and 54% in August. Using the algorithms for no acclimatization, this proportion is predicted to rise significantly as the weather warms, and since the deaths are standardized, this does not assume that the elderly will constitute a larger proportion of the population in the future.

An attempt was made to duplicate the procedure to determine the impact of winter weather on mortality using the same warming scenarios. Winter analog cities were selected, threshold temperatures were determined, and multiple regressions were performed, but the relationships for New York and the analog cities were unimpressive for winter (Table 9). Since their explained variance was low, the models were not robust enough to produce any predictive algorithms. Although findings will probably differ for other evaluated cities, these results suggest that any change in winter weather in the future will have little impact on weather-related mortality in New York.

A final aspect in the New York analysis was an attempt to determine if precipitation has any effect upon mortality. No attempt to estimate future impacts of precipitation was made, and the study concentrated on historical relationships only. It appears that precipitation may have an impact on mortality during both summer and winter (Table 10); however, unlike the temperature relationships, its influence does not appear to increase steadily as precipitation amounts increase. Thus, the precipitation evaluation was limited to comparing mortality rates during periods of precipitation and nonprecipitation and determining if the difference in the mean daily mortality rates was statistically significant between the two periods. It appears that a one-day lag exists between the precipitation episode and the mortality response during all seasons, and that the strongest correlation between these variables occurred in summer. On summer days with precipitation, mortality averaged .135 standard deviation below the mean, but on those days without precipitation, mortality averaged .100 standard deviation above the mean. One possible explanation for this relationship is that summer rain may provide a refreshing, cooling influence which tends to lessen discomfort and therefore, to lower mortality. Some strong winter relationships were also discovered, and rain appeared to have a greater influence on mortality than snowfall. A significant relationship was determined between all precipitation and no precipitation days, but when precipitation was subdivided into rain and snowfall, only the rain relationship proved to be statistically significant. Unlike the summer findings, mortality was significantly higher on days with rainfall, and mean daily mortality rate was almost .200 standard deviation above the mean on those days. Although days with snow falling appeared to have little impact on mortality, days with significant accumulations on the ground did correspond with higher mortality rates. Those days with three or more inches of snow on the ground averaged over .330 standard deviation above the mean.


The objectives of this study were to determine the historical relationships between weather and mortality and to estimate the possible impact of long-term climatic warming on future mortality rates in New York City. During the summer, weather appears to exert a significant influence on mortality in New York, but the future impact is largely dependent on whether New Yorkers will acclimatize to the predicted increasing warmth. If acclimatization is slow or is nonexistent, thousands of additional deaths may occur during each summer season if the mean temperature warms to 7deg.F above present levels. However, changes in winter weather should have little impact on mortality.

This study will be expanded to include fourteen additional cities around the United States. Analog cities will be determined, and inter-regional influences of weather will be examined. In addition, mortality rates will be subdivided by race and additional age categories, and those causes of death that are considered to be weather-related will be isolated and independently evaluated.


This research was supported by the U.S. Environmental Protection Agency under contract number 68-01-7033. The authors thank Mr. Dennis Tirpak, Office of Policy Analysis, EPA, and Dr. Melvyn Tockman, Associate Professor of Environmental Health Sciences, The Johns Hopkins University, for their suggestions and support. Thanks are also extended to the National Oceanic and Atmospheric Administration, for funding our initial climate/mortality work, and to various scientists at the NASA-Goddard Institute for Space Sciences for their interest in our project.


Applegate, W.B., J.W. Runyan, Jr., L. Brasfield, M.L. Williams, C. Konigsberg, and C. Fouche. 1981. Analysis of the 1980 heat wave in Memphis. Journal of the American Geriatrics Society 29:337-42.

Bridger, C.A., F.P. Ellis, and H.L. Taylor. 1976. Mortality in St. Louis, Missouri, during heat waves in 1936, 1953, 1954, 1955 and 1966. Environmental Research 12:38-48.

Centers for Disease Control. 1982. Exposure-related hypothermia deaths--District of Columbia 1972-1982. Morbidity and Mortality Weekly Report December:31-50.

Ellis, F.P. 1972. Mortality from heat illness and heat-aggravated illness in the United States. Environmental Research 5:1-58.

Fitzgerald, F.T., and C. Jessop. 1982. Accidental hypothermia: A report of 22 cases and review of the literature. Advanced Internal Medicine 27: 127-50.

Gallow, D., T.E. Graham, and S. Pfeiffer. 1984. Comparative thermoregulatory responses to acute cold in women of Asian and European descent. Human Biology 56:19-34.

Jones, T.S., A.P. Liang, E.M. Kilbourne, M.R. Griffin, P.A. Patriarca, S.G.G. Wassilak, R.J. Mullan, MD; R.F. Herrick, H.D. Donnel, Jr., K. Choi, and S.B. Thacker. 1982: Morbidity and mortality associated with the July 1980 heat wave in St. Louis and Kansas City, MO. Journal of the American Medical Association 247:3327-30.

Kalkstein, L.S. 1986. Progress Report for Contract #NA85AA-H-AI098: The Impact of Climate Upon Human Mortality. National Oceanic and Atmospheric Administration.

Kalkstein, L.S., and R.E. Davis. 1985. The development of a weather/mortality model for environmental impact assessment. Proceedings of the 7th Conference of Biometeorology and Aerobiology. 334-36.

Kilbourne, E.M., K. Choi, T.S. Jones, S.B. Thacker, and the Field Investigation Team. 1982. Risk factors for heatstroke: A case-control study. Journal of the American Medical Association 247:3332-36.

Lilianfeld, A.M., and D.E. Lilianfeld. 1980. Foundations of epidemiology, Oxford University Press.

Lye, M., and A. Kamal. 1977. The effects of a heat wave on mortality rates in elderly inpatients. The Lancet 1:529-31.

Mausner, J.S., and A.K. Bahn. 1974. Epidemiology: An introductory text, W.B. Saunders Co.

National Center for Health Statistics, 1978. Standardized Micro-Data Tape Transcripts. U.S. Department of Health, Education, and Welfare, DHEW Publication No. (PHS) 78- 1213.

Oechsli, F.W., and R.W. Buechley. 1970. Excess mortality associated with three Los Angeles September hot spells. Environmental Research 3 :277-84.

Rotton, J. 1983. Angry, sad, happy? Blame the weather. U.S. News and World Report 95: 52-3.

Steadman, R.G. 1979. The assessment of sultriness: Part I: A temperature-humidity index based on human physiology and clothing science. Journal of Applied Meteorology 18:861-73.

Willmott, C.J., S.G. Ackleson, R.E. Davis, J.J. Feddema, K.M. Klink, D.R. Legates, J. O'Donnel, and C.M. Rowe. 1985. Statistics for the evaluation and comparison of models. Journal of Geophysical Research 90 :8995-9005.