Lab: Climate and Health

I. Summary

Among the indirect health impacts of climate change are the effects that increased ambient temperatures and incident sunlight can have on ground-level ozone (O3) concentrations, particularly in the summer months. Ozone is a powerful oxidizing agent and lung irritant, and numerous studies have linked O3 exposures to increases in the number of hospital admissions or emergency room visits for respiratory illnesses, diminished lung function, and exacerbation of asthma symptoms (Brunkereef and Holgate 2002, Kinney 1999). Recent studies have also linked short-term O3 exposures to increased risk of death. It is estimated that more than 100 million Americans currently live in areas that exceed the EPA’s health-based O3 standard and that even low levels of tropospheric O3 are associated with increased risk of premature mortality (Bell et al. 2006). However, greenhouse gas regulations which decrease fossil fuel combustion could provide public health "co-benefits" by diminishing current-day O3 (and particulate matter) exposures as well as limiting future adverse effects from a range of climate-health impacts. It has been estimated that if greenhouse gas mitigation measures were initiated now, the cumulative number of PM- and O3-associated deaths avoided in New York City in the next 20 years could exceed 9,000 (Cifuentes et al. 2001).

Regional Ozone Modeling

Climate change can influence the concentration and distribution of air pollutants through a variety of processes, including the modification of biogenic emissions, the change of chemical reaction rates, mixed-layer heights that affect vertical mixing of pollutants, and modifications of synoptic flow patterns that govern pollutant transport. For example, warmer temperatures can result in increased concentrations of photochemical oxidants, while many past studies have revealed the impact of changing meteorological conditions on high ozone episodes. Tropospheric ozone is a secondary pollutant (i.e., not directly emitted) that is formed via complex chemical reactions involving nitrogen oxides, reactive hydrocarbons (also known as volatile organic compounds or VOCs), and sunlight. Key sources of nitrogen oxide O3 precursors include vehicle emissions, and vegetation is a source of biogenic VOCs. Many cities like New York experience appreciable inputs of nitrogen oxides (NOx) from commuter and commercial vehicle emissions, but typically have a lesser degree of VOC inputs from vegetation and are often referred to as "VOC-limited" areas. The complex, non-linear chemistry of O3 formation means that in some cases increased NOx emissions in urban core areas like NYC can react with O3 and locally lower O3 concentrations. This effect is called titration (Knowlton et al. 2004). Because of the importance of solar radiation and temperature in ozone photochemistry, significant concentrations of O3 in the New York region occur in the warmer months, i.e., May through October. The current national ambient air quality standard for O3 is an eight hour average concentration of 80 parts per billion, a level often exceeded in the New York region.

In order to project how global climate change in the future could affect local meteorology and air quality in the New York metropolitan region, the New York Climate and Health Project (NYCHP) developed a "downscaled" model system that linked global and regional climate models from the NASA/Goddard Institute for Space Studies (GISS), New York, NY, with regional air quality model simulations from researchers at the NY State Department of Environmental Conservation/University at Albany, NY (see Figure 1). The GISS GCM/Mesoscale Model 5 linked model (Lynn et al. 2004, Russell et al. 1995, Grell et al. 1994) provided the meteorological inputs needed for the air quality simulations, which used the Community Multiscale Air Quality (CMAQ) model (Byun and Ching 1999). The outputs from the air quality simulations have been used to evaluate the modeling system against observed O3 data, to project future O3 concentrations throughout the 36 km eastern us modeling domain under different climate scenarios, and for assessing potential public health impacts based within the New York metro region (Hogrefe et al. 2004a,b; Knowlton et al. 2004).

We will do a lab using O3 simulation files created by Christian Hogrefe of the NY State Department of Environmental Conservation/University at Albany. We will apply a health risk assessment model to ozone simulations of current vs. future conditions in the New York metropolitan region, in order to evaluate how climate change effects on local ozone concentrations could impact ozone-related mortality.

Metro NY study area

Figure 1. Map of Metropolitan NY study area for the NY Climate & Health Project:
CMAQ modeling domain on left, 31-county NY metro region on right (from Figure 1, Hogrefe et al. 2004a).

References

Bell M., R.D. Peng, F. Dominici. 2006. The exposure-response curve for ozone and risk of mortality and the adequacy of current ozone regulations. Environ Health Perspect 114:532-536.

Brunkereef B., S.T. Holgate. 2002. Air pollution and health. Lancet 360:1233-1242.

Byun D.W., J.K.S. Ching (eds.). 1999. Science Algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling Systems. EPA-600/R-99/030. Research Triangle Park,NC:US EPA Office of Research and Development.

Cifuentes L., V.H. Borja-Aburto, N. Gouveia, G. Thurston, D.L. Davis. 2001. Assessing the health benefits of urban air pollution reductions associated with climate change mitigation (2000-2020) : Santiago, Sao Paulo, Mexico City, and New York City. Environ Health Perspect 109(suppl 3):419-425.

Grell G.A., J. Dudhia, D. Stauffer. 1994. A description of the fifth-generation Penn State/NCAR Mesoscale Model (MM5). NCAR Technical Note, TN-398+STR. Boulder, CO: National Center for Atmospheric Research.

Hogrefe C.H., J. Biswas, B. Lynn, K. Civerolo, J.-Y. Ku, J. Rosenthal, C. Rosenzweig, R. Goldberg, P.L. Kinney. 2004a. Simulating regional-scale ozone climatology over the eastern United States: model evaluation results. Atmos Environ 38:2627-2638.

Hogrefe C.H., B. Lynn, K. Civerolo, J.-Y. Ku, J. Rosenthal, C. Rosenzweig, R. Goldberg, S. Gaffin, K. Knowlton, P.L. Kinney. 2004b. Simulating changes in regional air pollution over the eastern United States due to changes in global and regional climate and emissions. J Geophys Res 109, D22301, doi:10.1029/2004JD004690.

Kinney P.L. 1999. The pulmonary effects of outdoor ozone and particle air pollution. Sem Resp Crit Care Med 20:601-607.

Knowlton K., J.E. Rosenthal, C. Hogrefe, B. Lynn, S. Gaffin, R. Goldberg, C. Rosenzweig, K. Civerolo, J.-Y. Ku, P.L. Kinney. 2004. Assessing ozone-related health impacts under a changing climate. Environ Health Perspect 112:1557-1563.

Lynn B.H., L. Druyan, C. Hogrefe, J. Dudhia, C. Rosenzweig, R. Goldberg, D. Rind, R. Healy, J. Rosenthal, P.L.Kinney. 2004. Sensitivity of present and future surface temperatures to precipitation characteristics. Climate Res 28:53-65.

Russell, G.L., J.R. Miller, D. Rind. 1995. A coupled atmosphere-ocean model for transient climate change studies. Atmos Ocean 33:683-730.

Thurston, G.D., K. Ito. 2001. Ozone and premature mortality. J Expo Anal Environ Epidemiol 11(4):286-294.

Contributors

Stuart Gaffin, NASA/GISS Center for Climate Systems Research

Christian Hogrefe, NY State Department of Environmental Conservation, and University at Albany, Albany, NY

Patrick L. Kinney, Department of Environmental Health Sciences, Mailman School of Public Health, Columbia University, New York, NY

Cynthia Rosenzweig, NASA/GISS Center for Climate Systems Research, New York, NY [see the “Climate Change Information Resources” website at: http://ccir.ciesin.columbia.edu/nyc/]

Source Files

OZONE_1hr_DailyMax_1990s_2050s.xls (look at individual worksheets for different datasets and information referenced below)

II. Lab: Assessing Ozone-Related Mortality Risks Under a Changing Climate

Modeling Regional Ozone Air Pollution: Geographic and Temporal Variations

The worksheets in the Excel spreadsheet titled OZONE_1hr_DailyMax_1990s_2050s.xls contain 1-hr daily maximum O3 concentrations, simulated on a 36 km horizontal grid resolution by the NYCHP linked climate/air quality regional modeling system. To compare summertime O3 concentrations typical of the present to those projected under future climate-change conditions, daily CMAQ simulations for five summers in the 1990s (1993-1997) and for five summers in the 2050s (2053-2057) are included. Note that "summer" to climate and air quality modelers means June, July and August (or "JJA," which covers 92 days). Concentrations in ppb are given for the geographic "centroids" of the 31 counties in the study region (CMAQ model results were interpolated to these locations). Daily maximum concentrations averaged over the 31 counties are also provided, in the second column.

Reading from left to right through the worksheet tabs, tab 1 contains a "READ ME" file. On tabs 2 through 11, the names of each county are given in the first line. Starting in line 2, the first column contains the date, the second column contains the average daily maximum O3 concentrations over the 31 county region, and columns 3-33 contain the daily maximum concentrations at each centroid. All concentrations are given in ppb. 1-hr and 8-hr daily maximum concentrations are each contained in different worksheets.

All simulations use EPA's 1996 National Emission Trends inventory, i.e. all variability is coming from the MM5 meteorology (no changes in O3 precursor emissions), and the O3 concentrations for the 2050's reflect the climate effects of the IPCC SRES "A2" scenario only which describes a heterogeneous future world with continued growth in greenhouse gas emissions and world population through the 21st century (a fuller description of the IPCC SRES A2 scenario can be found at: http://sedac.ciesin.org/ddc/sres/index.html

Tab 12 contains the compiled daily simulations of all five summers for the 1990s vs. 2050s, for the 31-county average. This will be used in the 2nd section of the lab. On tab 13 ("CountyGeoCentroids_UrbNon") is the latitude and longitude of each county centroid, as well as whether each county can be classified as "urban" versus "non-urban" using a 10,000 person per square mile threshold criteria.

Part 1. Geographic Variations in O3

  1. Open the Excel file OZONE_1hr_DailyMax_1990s_2050s.xls.
  2. Go to worksheet named "1-hr 1993." Compute some simple descriptive statistics for daily summer 1-hr ozone concentrations in an urban county (New York, NY i.e. Manhattan) vs.  a non-urban county (Dutchess County, NY):

    3. [If you would like to select other urban vs. non-urban counties, go to the worksheet labeled, "County GeoCentroids_UrbNon" (it can be found on tab 13 to the right of "1-hr 2057"). It classifies all 31 of the counties according to their population density, which is one surrogate of urban vs. non-urban character.]

  3. Which county has a lower mean summer 1-hr O3 concentration, the urban or the non-urban? Lower maximum? Given what you know about different sources and types of O3 precursor emissions, can you propose reasons why there might be a difference in rural vs. urban concentrations?
  4. Plot the time-series of daily O3 concentrations for the 2 counties on the same axes, to get a visual comparison of the time trends in daily 1-hour maximum O3 concentrations over the course of a typical summer of the 1990s.
  5. Now repeat steps 2 through 5 for the same two counties, but use O3 concentrations on the worksheet named "1-hr 2053" i.e. looking at the future projections from CMAQ and simulations over the course of a typical summer of the 2050s.
  6. Which county has a lower mean summer concentration? Lower maximum? Have the comparative relationships between the 2 urban vs. non-urban counties changed over time?

Part 2. Temporal Variations in O3

  1. Now go to the "1990s_v_2050s_31 County Average" worksheet on Tab 12.
  2. Compute the descriptive statistics (again, the mean, minimum, maximum and standard deviation) for the 1990s vs. 2050s summers, averaged across the NY metro region, and plot the time-series.
  3. Which decade has a lower mean summer concentration? Lower standard deviation? Lower minimum? Lower maximum? Consider that by using the "31 County Average" values, you have removed the improvement in model resolution afforded by CMAQ's 36 km model grid, and are in effect applying one grid-box value across the entire study area. This is similar to the grid resolution of some of the GCMs in use globally which use grid boxes that cover 4x5° of latitude and longitude i.e. hundreds of kilometers on a side.

Part 3. Health Risk Assessment: Calculating Ozone-Related Mortality

In public health, there are two widely-used modeling methods that can be applied to evaluate potential climate-health links. One method includes epidemiological analyses which use empirical data on environmental exposures and human health responses to derive models that estimate the relationship between a past or present environmental exposure and potentially associated health outcomes. Typically, epidemiological models derive effect estimates that describe the change in the “relative risk” of a particular health effect (death, hospitalization, etc.) per unit change in a particular environmental exposure. When expressed as the percentage change in outcome per unit change in exposure, these estimates are often referred to as exposure-risk coefficients (ERCs).

A second method, health risk assessment, is a way to project future risks to health from estimated changes in environmental exposures. Health risk assessment applies the effect estimates or ERCs from epidemiological studies into computer model simulations of changes in future vs. present environmental conditions, and can project the number of additional impacts stemming from those changing conditions (see Figure 2). Risk assessment has been in wide use in environmental regulation since the 1980s.

Health Risk Assessment Model

The four components on the left side of Figure 2 have been estimated for you in the calculations on tab 15 "Health RA Formula" in OZONE_1hr_DailyMax_1990s_2050s.xls.

For each decade, county-level mortality impacts were computed as:

(P/100,000) * B*E * ERC = M

where

P is the estimated population county population during the time period of interest, and A2-consistent population growth projections for the year 2055, as well as Census 2000 county population data, are given on tab 14;

B is the estimated baseline county-level daily summer mortality rate (per 100,000 population), also given on tab 15;

E is the mean daily 1-hr maximum summer O3 concentration in each county, which has been derived from the CMAQ model simulations on tabs 2-11;

ERC is the exposure-response coefficient, which quantifies the magnitude of the proportional change in daily mortality that would be expected in response to a given daily O3 concentration, based on results from the epidemiological literature of 1.056 per 100 ppb increase in daily maximum O3 levels (Thurston & Ito 2001); and

M is the estimated number of summer seasonal deaths attributable to O3 concentrations, which you will calculate using the Excel spreadsheet on tab 16.

For the purposes of this lab, the Excel spreadsheet will perform the calculation on the mean daily summer 1-hr maximum O3 values for each county in each decade, and multiply that result by 92, which is the number of days per summer. The formula you will see in the Excel spreadsheet is:

(H7/100000) * I7 *92 * [(EXP(0.00054488*J10))-1],

which translates to:

(Population/100,000) * Daily mortality rate per 100,000 pop * 92 days per summer * [(eβ * Δenvironment)-1
= change in mortality from a change in exposure

(Population/100,000) * Daily mortality rate per 100,000 pop * 92 days per summer* [incremental change in RR of mortality]
= change in mortality from change in exposure

There is an exponential term in the ERC component of the risk assessment because ozone has a log-linear mortality response function: for a linear change in ozone concentrations there is a logarithmic change in mortality.

The β value is derived from the Relative Risk of 1.056, because:

β = [ln(Relative Risk)]/Δenvironment = [ln(1.056)]/Δ100ppb = 0.054488/100 = 0.00054488

In a complete iteration of the O3-mortality health risk assessment, each day's specific O3–mortality contribution would be estimated using a statistical software package like SAS, and the daily contributions across the 92 days of the summer would be added to arrive at the seasonal summer O3-related mortality burden typical of a particular decade. Another feature of a complete health risk assessment would be applying the 95% confidence interval around the ERC. For example, the lower and upper 95% confidence limits of the relative risk from Thurston and Ito 2001 were 1.032 - 1.081, meaning that if ozone concentrations increase by 100 ppb, we can say with 95% confidence that the relative risk of death from that change lies between 1.032 and 1.081. This is a method for describing uncertainty in risk assessments. For expediency in your calculations you will only apply the central effect estimate for Relative Risk, or RR=1.056.

By comparing the decadal summer O3-mortality totals, climate-related changes under present-day vs. future climate-changed conditions can be evaluated.

Geographic & Temporal Variations in O3-Related Mortality

  1. In spreadsheet OZONE_DailyMax_1990s_2050s.xls, open the worksheet "Health RA Formula" on tab 16.
  2. In cell G7, examine the formula for calculating the 1990s county-specific summer O3-related mortality burden, given the CMAQ-simulated mean daily 1-hour maximum O3 concentration for that decade’s summers.
  3. Copy that formula through cell G37 to evaluate the other 31 counties for the 1990s. Calculate the regional total number of O3-related deaths in a typical 1990s summer. What is the range of county-specific values for summer O3-related deaths (the minimum value & the maximum value from individual counties)? Which types counties show the highest numbers of O3-related deaths? What are the primary factors affecting these totals?
  4. Repeat steps 11-12, but for the "A2" 2050s, by applying similar steps in column L, and answer the same questions.
  5. In column M, you will calculate the difference between your estimate of mortality in the 1990s vs. your projection of mortality in the A2 2050s. Copy the formula in cell M7 down through cell M37. Then total the differences calculated for each county to sum the aggregated regional change in summer O3-related mortality under a changing climate. What is the percentage change across the entire region by the 2050s, using the 1990s as the denominator (for example, M7/G7)? Looking at the individual county values, in what county is the percentage increase greatest? What is the range of absolute values, and in which counties are the absolute numbers greatest and least?
  6. Are the aggregated regional changes in O3-related mortality different from what you expected? If so, describe and propose one possible reason.

Lab Report Instructions

Write a lab report (as per the Lab Report Format) summarizing the major findings of your investigation. Consider the following questions as you write your report.

Updated October 2, 2007
To report problems, email webmaster.