lable at ScienceDirect

Environmental Pollution 185 (2014) 322e332

Contents lists avai

Environmental Pollution

journal homepage: www.elsevier .com/locate/envpol

Risk assessment for cardiovascular and respiratory mortality due to airpollution and synoptic meteorology in 10 Canadian citiesq

Jennifer K. Vanos a,b, Christopher Hebbern a, Sabit Cakmak a,*

aHealth Canada, Environmental Health Science and Research Bureau, Population Studies Division, 50 Columbine Driveway, Ottawa, ON K1A 0K9, CanadabDepartment of Geosciences, Texas Tech University, 2500 Broadway, St. Lubbock, TX 79401, USA

a r t i c l e i n f o

Article history:Received 26 July 2013Received in revised form7 October 2013Accepted 10 November 2013

Keywords:MortalityRespiratoryCardiovascularRelative riskSpatial synoptic classificationAir pollution

q This is an open-access article distributed undeCommons Attribution-NonCommercial-ShareAlike Lcommercial use, distribution, and reproduction inoriginal author and source are credited.* Corresponding author.

E-mail address: sabit_cakmak@hc-sc.gc.ca (S. Cakm

0269-7491/$ e see front matter Crown Copyright � 2http://dx.doi.org/10.1016/j.envpol.2013.11.007

a b s t r a c t

Synoptic weather and ambient air quality synergistically influence human health. We report the relativerisk of mortality from all non-accidental, respiratory-, and cardiovascular-related causes, associated withexposure to four air pollutants, by weather type and season, in 10 major Canadian cities for 1981 through1999. We conducted this multi-city time-series study using Poisson generalized linear models stratifiedby season and each of six distinctive synoptic weather types. Statistically significant relationships ofmortality due to short-term exposure to carbon monoxide, nitrogen dioxide, sulphur dioxide, and ozonewere found, with significant modifications of risk by weather type, season, and mortality cause. In total,61% of the respiratory-related mortality relative risk estimates were significantly higher than forcardiovascular-related mortality. The combined effect of weather and air pollution is greatest whentropical-type weather is present in the spring or summer.

Crown Copyright � 2013 Published by Elsevier Ltd. All rights reserved.

1. Introduction

Weather and ambient air pollution are known influences onhuman mortality and morbidity, but their covariance results in ahigh likelihood that the effect of one modifies the effect of theother. Due to these and other potential confounding factors, healthrisk models for air pollution often include weather-related vari-ables as controls (Anderson et al., 2001; Cakmak et al., 2006; Daleset al., 2006; Roberts, 2004; Zanobetti and Schwartz, 2005). Morerecently, studies examining acute health endpoints have reportedthe simultaneous effects of both weather variables and air pollut-ants on health. Many temperatureemortality studies fail to accountfor air pollution when modelling (Basu and Samet, 2002), whichmay cause the models to overestimate mortality associated withweather (Rainham and Smoyer-Tomic, 2003). Much of the envi-ronmental epidemiological literature associates temperature withmortality, and uses air pollution as a confounder, an effect modifier,or both (Basu, 2009). Other variables, such as additional meteoro-logical attributes, time trends, and air pollution interactions, have

r the terms of the Creativeicense, which permits non-any medium, provided the

ak).

013 Published by Elsevier Ltd. All

also been investigated as effect modifiers or confounders (Filleulet al., 2006; Ren and Tong, 2006; Ren et al., 2006; Smoyer-Tomicet al., 2003), yet teasing apart the independent effect of each isdifficult (Basu, 2009).

For example, in a study examining deaths in nine cities duringthe 2003 heat wave in France (Filleul et al., 2006), ozone andtemperature each made a statistically significant contribution tothe number of daily deaths observed. Studies have controlled forweather using synoptic weather type-based approaches (Chenget al., 2009; Hanna et al., 2011; Rainham et al., 2005; Samet et al.,1998; Vaneckova et al., 2008). Cheng et al. (2009) used synopticweather typing to differentiate between the combined impacts ofextreme temperatures and air pollution on human mortality in fivesouth-central Canadian cities. On extreme weather and air pollu-tion days, the annual mean mortality was significantly abovebaseline (e.g., Ottawa ¼ 462 (CI 438e486)); however, 80% of thismortality was attributable to air pollution. Further, seasonal riskestimates of mortality have been shown to differ yearly, with sta-tistically significant associations between airborne particles anddaily death for both summer and winter seasons in 10 U.S. citieswith varying climates (Schwartz, 2000).

The risks of mortality differ when associating a specific cause ofdeath to the ambient environment, yet there are relatively fewstudies (Choi et al., 1997; Hales and Salmond, 2000; Ren and Tong,2006; Ren et al., 2007, 2006) that examine cause-specific mortality.Ren et al. (2006) found that airborne particles significantly

rights reserved.

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332 323

modified the effects of air temperature on respiratory and cardio-vascular hospital admissions in Brisbane, Australia. Ozone alone hasbeen associated with higher levels of hospital admissions, respi-ratory symptoms, impaired lung development, and mortality,among other adverse health responses (Anderson et al., 2004; Bellet al., 2006; Dockery and Pope, 1994; Hanna et al., 2011; Levy et al.,2005; Stieb et al., 2003). Increasing sunlight and temperatures alsoincrease tropospheric ozone production due to the photochemicalnature of this secondary pollutant. According to HYPERLINKBellet al. (2007a,b), projections of a warming climate will add to thecurrent health burden caused by ozone, which is enhanced on hotdays. Ozone has been found to be closely associated with furtherweather variables (Bell et al., 2006; Davis et al., 2010; Hanna et al.,2011), with lower ozone levels found during moderate or coolweather, as well as moist (USEPA, 2004). During summer, ozone hasbeen shown to modify the human health effects of temperature invarious regions of the United States (Ren et al., 2006). Stable, dry,and hot conditions result in the greatest pollution build up andozone production, as abundant sunlight and heat is present forphotochemical reactions (Davis et al., 2010).

The environmental and health effects of air pollutants areextensive as demonstrated by hundreds of published research pa-pers in many areas of the world (see Confalonieri et al. (2007) andSujaritpong et al. (2013) for extensive reviews). Some key negativehealth outcomes due to increased exposures to air pollutioninclude elevated mortality with PM10 (Ren and Tong, 2006;Stafoggia et al., 2008), increases in hospital admissions due toaeroallergens (Dales et al., 2004) and gaseous pollutants (Daleset al., 2006), and increased morbidity (asthma and myocardialinfarction) with ozone on hot, dry days (Hanna et al., 2011). A studyspecific to Canada (Burnett et al., 1998a) found statistically signifi-cant positive associations between daily fluctuations in mortalityand ambient levels of carbon monoxide (CO), nitrogen dioxide(NO2), sulphur dioxide (SO2), coefficient of haze, total suspendedparticulates (TSP), sulphates, and estimated particulates in Toronto(1980e1994). Burnett et al. (1998b) emphasized that all air pollu-tion from vehicular sources, and CO in particular, should beconsidered as a potential cause of increased mortality in urbanpopulations in Canada.

However, there are inconclusive results from many studies,leaving the magnitude of mortality effect modifications due to airpollution or weather type unclear (Basu, 2009; Rainham andSmoyer-Tomic, 2003). Accordingly, the goal of this study is toexamine the effect modification of synoptic weather types on therelative risk of mortality (RR) associated with four air pollutants:CO, NO2, O3, and SO2. We subdivide the analysis into all non-accidental, respiratory-, and cardiovascular-related mortality, andcombine the results of 10 Canadian cities to estimate an overallcountrywide effect of each pollutant on each mortality type, for theperiod 1981e1999. A spatial synoptic climatological procedure isused to develop a daily series of prevailing air masses (or surfaceweather types). The risk of mortality due to exposure to each airpollutant is determined for each weather type. Additionally, sea-sonality has been found to explain a significant amount of vari-ability in pollution andweather-relatedmortality studies (Pope andKalkstein, 1996; Sheridan and Kalkstein, 2010); hence, we furthersegregate the analysis by season.

2. Methods

2.1. Spatial synoptic classification

The current study accounts for weather modification through identifyingweather types using the Spatial Synoptic Classification (SSC) system (Sheridan,2002). The SSC identifies a daily weather type based on four diurnal surface mea-surements of six meteorological parameters. The weather types include: dry mod-erate (DM), dry polar (DP), dry tropical (DT), moist moderate (MM), moist polar

(MP), moist tropical (MT), moist tropical plus (MTþ), and a transitional (TR) cate-gory. The MTþ designation is an extreme subset of MT, in which the morning andafternoon apparent temperatures are above the corresponding MT means for thegiven location (Sheridan, 2002). Detailed descriptions relevant to Canada can befound in Vanos and Cakmak (2013). The SSC has been implemented in many humanhealth-climate studies (Hajat and Kosatky, 2010; Hanna et al., 2011; Rainham et al.,2005; Sheridan et al., 2009). The SSC accounts for relative temporal and spatialvariability, important in Canadian cites due to their wide-ranging climate regimes.For example, spatially, a summer DT day in Toronto is warmer than a summer DT dayin Vancouver. Temporally, July MT air in Toronto is warmer than in March (e.g., 28versus 15 �C for Toronto, respectively).

2.2. Mortality and air pollution data

Dailymortality and air pollution data from 1981 through 1999 were obtained forten cities across Canada: Saint John on the east coast; Toronto, Montreal, Ottawa,Windsor, and Quebec City in the Great Lakes/St Lawrence river region; Calgary,Edmonton, andWinnipeg in the Prairies, and Vancouver on thewest coast. Mortalitydata were accessed from the Canadian Institute for Health Information (CIHI)database, including all non-accidental (ICD9 < 800) deaths. National Air PollutionSurveillance (NAPS) data were provided by Environment Canada, consisting of meandaily ambient concentrations of carbon monoxide (CO, ppm), nitrogen dioxide (NO2,ppb), sulphur dioxide (SO2, ppb) and ground-level ozone (O3, ppb) measured eitherdowntown or at city airports located within 27 km of downtown. Hourly weatherdata from city airports were downloaded from the National Climate Data and In-formation Archive. Particulate matter (PM10 and PM2.5) was not used in this studydue to brief data records in Canada.

2.3. Time-Series modelling

A piece-wise Poisson general linear model (GLM) was applied to empirical airpollution measurements of CO, NO2, O3, and SO2 and daily all non-accidental, res-piratory, and cardiovascular mortality counts, assuming a linear association betweenair pollutants and mortality on the logarithmic scale varying at random betweencities. Analysis was completed for full year and by season, stratified by weather type.Each time-seriesmodel for each city included indicator variables for the day of week,and was adjusted for temporal trends using natural spline functions for day of study,with a knot for each of 30, 90, 180, 270, and 365 days of observation. The optimalmodel was selected based on the number of knots that either minimized theAkaike’s Information Criteria (AIC) (representing enhanced model strength), orminimized evidence that the model residuals demonstrate departures from themodel assumptions, or structure that may not be accounted for in the model. Thelatter was examined after fitting models with natural spline functions to test forserial correlation in the residuals, which was completed using Bartlett’s test (seeSupplementary Figures for an example of a residual plot). The above steps wereimplemented separately for each city.

Finally, each air pollutant was added to themodel containing natural splines andindicator variables. Lagging times of 0,1, 2, 3, 4, and 5 days were examined for the airpollutants, selecting the lag that optimized the effect size. Once the final model wasselected, the confidence intervals (CI) for RR were generated across each weathertype for each city. The city-specific estimates were pooled using a random-effectsmodel to calculate the overall influence of air pollution on mortality. Thisapproach weights the effect estimates by the inverse sum of within- and between-city variance, thus accounting for any heterogeneity among the cities in the pooledeffect estimates. It also gives greater statistical power, and a final estimation of anoverall countrywide effect.

The model is summarized as follows:

LogEðYt=XtÞ ¼ bXt�l þ DOWt þ ns ðtime; df Þ (1)

where Yt is the daily count of cause-specific mortality; b is the regression coefficientlinking pollution to daily mortality; Xt-l is the pollution level on day twith 0e5 daysof lag; DOWt indicates the day of the week on day t; ns(time, df) is the natural splineof calendar timewith degrees of freedom corresponding to a knot at 30, 90,180, 270,and 365 days of observation. The effect estimates for each season and weather typewere obtained by pollutant � season/weather type interaction terms. For example,season specific effect estimates can be found by replacing b bybWIW þ bspIsp þ bauIau þ bsuIsu, where Iw, Isp, Iau, and Isu are indicators of winter,spring, autumn and summer respectively. The model also allows for temporal trendsto be changed by season by replacing ns(time, df) by

nsðtime; df Þ ¼ nsðtime; df ÞIWþ nsðtime; df ÞIspþ &.g (2)

Pooled relative risk estimates were obtained for the concentration of airpollutant equivalent to the population weighted mean (PWM) level for all cities ineach weather type, and their 95% confidence intervals (CI).

Details of the above statistical analyses are given in Cakmak et al. (2006). Sig-nificance testing for models was completed using t-tests (p < 0.001). All statisticalanalyses were completed using R 2.10.1 (The R Foundation for Statistical Computing,2008).

Table 1Weather type frequencies and average ten-city mean yearly descriptive statistics by weather type for air pollution concentrations, population weighted mean (PWM) for eachcause of morality, and air temperature (1981e1999).

Dry Dry Dry Moist Moist Moist Transition Mean

Moderate Polar Tropical Moderate Polar Tropical All weather types

Frequency (%) 23.3 25.1 1.8 16.7 18.0 4.1 10.9 e

Air Temperature (�C) 9.7 1.4 15.6 9.4 3.8 14.6 5.9 8.9 � 5.4

Mean � SD Mean � SD Mean � SD Mean � SD Mean � SD Mean � SD Mean � SD Mean � SD

Air PollutionCO (ppm) 0.9 � 0.26 1.0 � 0.29 1.0 � 0.31 1.0 � 0.27 0.9 � 0.24 1.0 � 0.35 0.8 � 0.23 0.9 � 0.08NO2 (ppb) 22.1 � 5.78 21.8 � 5.76 25.0* � 6.02 21.1 � 5.56 19.8 � 5.25 22.0 � 6.42 19.5 � 5.12 21.1 � 1.14SO2 (ppb) 5.4 � 2.86 5.2 � 2.29 6.0* � 3.68 5.0 � 2.79 5.0 � 2.69 5.0 � 2.83 4.8 � 2.28 5.1 � 0.21O3 (ppb) 19.0 � 2.39 15.6 � 2.98 30.3* � 5.03 15.3 � 3.29 13.8 � 4.08 23.0* � 4.76 17.8 � 2.70 19.3 � 5.73MortalityAll-Cause 18.1 � 0.09 18.7 � 0.20 17.5 � 0.44 18.5 � 0.11 18.9 � 0.14 18.4 � 0.63 19.0 � 0.18 18.4 � 0.50Cardiovascular 7.5 � 0.05 8.0 � 0.10 7.3 � 0.26 7.8 � 0.06 8.0 � 0.07 7.4 � 0.30 8.0 � 0.10 7.7 � 0.31Respiratory 1.5 � 0.02 1.7 � 0.05 1.5 � 0.12 1.7 � 0.03 1.7 � 0.04 1.6 � 0.23 1.7 � 0.05 1.6 � 0.10

*Indicates values significantly greater than the all-weather type mean for given variable, as shown in the far right column of mean values, for all air pollutants and all non-accidental mortality each season (p < 0.05).Indicates values as significantly lower than the all-weather type mean for given variable.

012345678910

05101520253035404550

J F M A M J J A S O N D

Mean

Co

ncen

tratio

n S

O2

(p

pb

), C

O

(p

pm

)

Me

an

Co

nc

en

tra

tio

n N

O2, O

3(p

pb

)

Month

NO2 O3 SO2 CO

Fig. 1. Average mean monthly concentrations of four air pollutants (left-axis: NO2 and O3, in ppb); right-axis: SO2 (ppb) and CO (ppm)), for 10 Canadian cities for the period 1981e1999 inclusive.

0

5

10

15

20

25

Mor

talit

y (P

WM

)

Winter Spring

DM DP DT MM MP MT TR

Fall

0

5

10

15

20

25

DM DP DT MM MP MT TR

Mor

tait

y (P

WM

)

Summer

All Non-Accidental Cardiovascular Respiratory

Fig. 2. Average mortality for 10 Canadian cities calculated by population weighted mean (PWM), depicting all-non-accidental, cardiovascular, and respiratory mortality.

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332324

Table 2All weather-typea mean relative mortality risk (RR) estimates at population-weighted mean, pooled across ten cities, with standard deviation (SD) for four pollutants (1981e1999).

CO NO2 SO2 O3

RR SD RR SD RR SD RR SD

WinterCardiovascular 1.034 �0.009 1.032 �0.007 1.022 �0.014 1.048 �0.016Respiratory 1.065 �0.010 1.074 �0.025 1.069 �0.011 1.067 �0.026Non-accidental 1.016 �0.018 1.020 �0.012 1.023 �0.012 1.029 �0.015SpringCardiovascular 1.061 �0.008 1.052 �0.020 1.026 �0.018 1.034 �0.016Respiratory 1.132 �0.055 1.118 �0.028 1.068 �0.048 1.093 �0.034Non-accidental 1.037 �0.013 1.040 �0.015 1.022 �0.023 1.038 �0.024SummerCardiovascular 1.058 �0.018 1.045 �0.006 1.043 �0.035 1.060 �0.020Respiratory 1.134 �0.074 1.103 �0.034 1.066 �0.046 1.083 �0.019Non-accidental 1.043 �0.021 1.043 �0.018 1.032 �0.032 1.033 �0.008FallCardiovascular 1.035 �0.011 1.043 �0.019 1.032 �0.006 1.021 �0.010Respiratory 1.063 �0.019 1.087 �0.017 1.066 �0.016 1.065 �0.015Non-accidental 1.024 �0.007 1.038 �0.008 1.039 �0.005 1.015 �0.006

a This excludes the transitional days.

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332 325

3. Results

3.1. Atmospheric conditions and mortality by weather type

The dry polar (DP) and dry moderate (DM) weather types arethe most frequent in the summer for all cities combined (i.e., 25.1%and 23.3%, respectively (Table 1)). In the hot and dry weather typeDT, concentrations of NO2, SO2, and O3 are significantly greaterthan the all-weather type mean, as are the O3 levels in the MTweather type. Significantly lower concentrations of O3 are presentin the DP, MM, and MP weather types, as well as NO2 in the MPweather type. Of the four pollutants, CO levels demonstrate theleast change in average monthly concentrations for 1981e1999(Fig. 1), and the least change by season. Ozone and NO2 display thelargest concentration range relative to the yearly mean between

Table 3Weather types categorized based on increased risk of causing respiratory or cardiovascucurrent study results (low < 5% increase in RR < medium 5e10%, high>10%). Weather typ

Season HighestaverageRR

Low (<5% increase) Medium (5e1

CO NO2 SO2 O3 CO N

CVD

Winter DryModerate

DM, DPMM, MP

DM, DPMM, MP

DM, DPMM, MP

DP MM,MP

Spring DryTropical

MM DP MM,MT

DM, DPMM, MPMT

DM, DPMM, MPMT

DM, DP,DT MP,MT

DM

Summer DryTropical

DM, MT DM, DP,DT MM,MP

DM, DPMM, MPMT

DM, MM,MP

DP, DTMM, MP

M

Fall NA DM, DP,MM, MP

DM, DP,MM, MP

DM, DP,MM, MP

DM, DP,MM, MPMT

M

RespiratoryWinter Dry

ModerateMM DM, DP,

MM, MPD

Spring DryTropical,MoistTropical

DM, DP,MP

DM, MP D

Summer MoistPolar

MT, MM DM, MM DM

Fall Dry Polar MP DM, DP,MM, MP

DM

seasons (17 ppb), with O3 demonstrating the highest concentra-tions in the warm season, and NO2 and SO2 doing so in the coldseason.

The seasonal PWM of all non-accidental mortality for all cities(Fig. 2) is the highest in the winter season (20.6 deaths per day).Similarly, mean cardiovascular- and respiratory-related mortality ishighest in the winter season (8.7 and 2.1 deaths per day, respec-tively). Summertime DP and MP weather types are associated withsignificantly lower mortality than the all-weather type mean. Thehighest mean mortality when divided into weather type occurs inwinter, coinciding with the highest concentrations of CO, NO2, andSO2. High springtime mortality coincides with significantly high O3concentrations, relative to the seasons. City-specific pollution,climate, and mortality information is provided in theSupplementary Tables.

lar morality due to the four air pollutants for four seasons. Categories are relative toes defined as oppressive in Canada and linked to heat-related mortality are bolded.

0% increase) High (>10% increase)

O2 SO2 O3 CO NO2 SO2 O3

DM

M, DTP

DT DT

T DT DP, DT

T

M, MP DM, DP,MM, MP

DP, MP DM

M, DP MM DP, MM,MP

DP, DTMM, MT

DP, DTMM, MT

DT, MT DT, MT

M, MM,T

DM, MM DM, DP,MM, MT

DP, MP,MT

DP, MT MP MP

M, MM,P

DM, DP, MM DM, DP,MM, MP

DP

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332326

3.2. Relative risks of mortality due to air pollution

We report statistically significant relationships between mor-tality and short-term exposure to air pollution for all non-accidental, respiratory-, and cardiovascular-related mortality.Exposure to the four air pollutants individually (CO, NO2, SO2, O3)results in overall significantly higher relative risk (RR) estimates forrespiratory-related mortality than for cardiovascular mortality,with 61% of the estimates significantly greater due to respiratorymortality This is also 5e10% greater than the all non-accidentalmortality estimates, as depicted in the averages in Table 2. Allnon-accidental mortality demonstrates the highest PWMmortality(Fig. 2); however, the corresponding RR estimates are lower thanthat of respiratory-related deaths. Cardiovascular-related mortalityRR estimates are not significantly different to that of all non-accidental mortality, consistently within �3% difference due toany air pollutant.

Table 3 summarizes the RR estimates for both causes of death(respiratory or CVD) due to exposure to each pollutant in eachseason and weather type. Relative risk estimates are grouped into‘low’ (<5% increase in RR), ‘medium’ (5e10% increase) and ‘high’risks (>10% increase). The warmest weather types in each seasonconsistently present the highest risks for each mortality cause (i.e.,DM in winter, DT in summer, MT in spring and fall), although thehighest risk of respiratory mortality, and the highest overall in-crease in RR for both mortality causes, is found within the MPweather type (see Table 4).

Table 4Relative risk of mortalitya (RR) due to air pollution within synoptic weather type pooled apopulation-weighted means, with lower and upper 95% confidence intervals (CI) displayerespect to season and mortality type (CVD or Respiratory) for each pollutant.

CO NO2

RR 95% CI RR 95%

SpringCardiovascular DM 1.061 (1.038, 1.084) 1.065 (1.0

DP 1.064 (1.036, 1.093) 1.039y (1.0DT 1.066 (1.017, 1.118) 1.081 (1.0MM 1.049 (1.022, 1.077) 1.033y (1.0MP 1.052 (1.024, 1.08) 1.060 (1.0MT 1.071 (1.018, 1.126) 1.034y (0.9TR 1.031y (1.009, 1.053) 1.034y (1.0

Respiratory DM 1.079y (1.032, 1.127) 1.093 (1.0DP 1.115 (1.061, 1.172) 1.079 (1.0DT 1.107 (0.984, 1.245) 1.147 (1.0MM 1.188 (1.095, 1.290) 1.120 (1.0MP 1.088y (1.015, 1.165) 1.120 (1.0MT 1.212 (1.091, 1.345) 1.148 (1.0TR 1.086y (1.037, 1.136) 1.070y (1.0

SummerCardiovascular DM 1.043 (1.024, 1.062) 1.047 (1.0

DP 1.060 (1.029, 1.091) 1.041 (1.0DT 1.079 (0.950, 1.225) 1.037 (0.8MM 1.053 (1.021, 1.085) 1.040 (1.0MP 1.079 (1.039, 1.122) 1.049 (1.0MT 1.036y (0.996, 1.077) 1.053 (1.0TR 1.060y (1.021, 1.100) 1.073y (1.0

Respiratory DM 1.087y (1.040, 1.135) 1.080 (1.0DP 1.123 (1.038, 1.215) 1.123 (1.0DT e (e, e) e (e,MM 1.090y (1.020, 1.164) 1.091 (1.0MP 1.107 (1.107, 1.107) 1.151 (1.0MT 1.263 (1.067, 1.495) 1.069y (1.0TR 1.043 (0.921, 1.181) 1.146y (1.0

y Denotes values that are significantly less than bolded estimates when compared withidifferences occur (e.g., Respiratory-related mortality due to O3 in spring), corresponding

a Statistically significant RR found when 95% lower CI > 1.0 (p < 0.001).b Only spring and summer seasons presented as fall and winter did not show significa

3.3. Respiratory-related mortality

Table 2 displays the all-weather type average RR estimates foreach season, while Figs. 3e6 display each combined model esti-mate (and 95% CI) by air pollutant, season, and weather type. Allrelative risk estimates for respiratory disease are significantly above1.0, for all combinations of season, pollutant, and weather type.Further, the average yearly respiratory-related RR (not shown)demonstrates the strongest associations with CO and NO2 for allweather types combined, with a 10.7% and 13.3% increase,respectively.

The highest RR estimates are found in the springtime, with NO2and CO exposure resulting in the greatest overall risk (Figs. 3 and 4).The most harmful springtime air pollutant is CO, associated with a10.0e21.0% increase in mortality for all weather types excludingDM and MP. Although lower in magnitude, the DM and MP relativerisk estimates are also significant, and demonstrate 8.0 and 9.0%increases in mortality risk due to CO exposure, respectively.Respiratory-related risks due to CO are higher in the MT weathertype (RR¼ 1.212 (95% CI 1.091e1.345)). This estimate is significantlygreater than the DM andMPweather types (Table 4). NO2 exposurein the springtime also poses high risk of respiratory-related mor-tality, which is greatest in the DT and MT weather types (i.e.,RR ¼ 1.147 (95% CI 1.027e1.280)) and RR ¼ 1.148 (95% CI ¼ 1.048e1.256), respectively, yet these are not significantly greater than theremaining four weather types. However, as shown in Table 4, therisk estimates within the DT and MT weather types due to SO2 and

cross ten cities for spring and summer seasonsb (1981e1999). Analysis completed atd. Statistically significant differences between the weather types were assessed with

SO2 O3

CI RR 95% CI RR 95% CI

42, 1.087) 1.021y (1.005, 1.037) 1.020y (1.013, 1.038)19, 1.059) 1.021y (1.004, 1.037) 1.021y (1.004, 1.039)38, 1.126) 1.061 (1.018, 1.105) 1.061 (1.019, 1.105)07, 1.06) 1.018y (1.006, 1.03) 1.020y (1.001, 1.038)35, 1.086) 1.008y (1.001, 1.015) 1.039 (1.020, 1.058)95, 1.074) 1.025y (0.995, 1.055) 1.036 (1.001, 1.072)11, 1.057) 1.023y (1.009, 1.036) 1.051y (1.008, 1.096)50, 1.137) 1.023y (1.004, 1.043) 1.053y* (1.017, 1.091)34, 1.126) 1.034y (1.009, 1.059) 1.083 (1.040, 1.127)27, 1.280) 1.110 (0.996, 1.238) 1.124^ (1.001, 1.262)62, 1.181) 1.077 (1.030, 1.127) 1.067y* (1.017, 1.120)64, 1.179) 1.026y (1.007, 1.045) 1.088 (1.044, 1.134)48, 1.256) 1.138 (1.054, 1.228) 1.143* (1.054, 1.240)21, 1.121) 1.074y (1.039, 1.109) 1.114y (1.053, 1.177)

30, 1.064) 1.034y (1.022, 1.045) 1.048 (1.033, 1.064)17, 1.065) 1.028y (1.008, 1.048) 1.081 (1.051, 1.112)80, 1.222) 1.115 (0.917, 1.357) 1.089 (0.892, 1.328)16, 1.063) 1.029y (1.012, 1.047) 1.041y (1.019, 1.064)21, 1.079) 1.024y (1.005, 1.044) 1.043y (1.009, 1.077)22, 1.085) 1.028y (1.006, 1.051) 1.059 (1.031, 1.087)31, 1.116) 1.040y (1.014, 1.067) 1.054y (1.022, 1.086)40, 1.121) 1.069 (1.033, 1.106) 1.070 (1.039, 1.103)58, 1.193) 1.089 (1.089, 1.089) 1.059 (1.013, 1.107)e) e (e, e) e (e, e)40, 1.146) 1.031y (1.007, 1.056) 1.096 (1.052, 1.143)66, 1.242) 1.129 (1.068, 1.194) 1.106 (1.024, 1.194)03, 1.139) 1.012y (1.012, 1.012) 1.085 (1.020, 1.155)59, 1.241) 1.079 (1.018, 1.145) 1.165y (1.081, 1.256)

n the same season and mortality category. When more than one pair of significantpairs contain same symbol ( or *), with bolder values being significantly greater.

nt differences by weather type. Data presented in Figs. 3e6.

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332 327

O3 are significantly greater than other weather types, highlightingthe influences of weather on risk.

Further, in the tropical weather types (DT and MT) in spring,there is a >10% increase in respiratory-related mortality risk for allpollutants (Table 3). This trend extends across the MT weather typefor all pollutants in springtime, with a 16.0% increase in mortalityrisk from baseline mortality. The MT respiratory-related estimatesare also significantly higher than the remaining two mortalitycategories for any air pollutant.

Summertime respiratory-related RR estimates closely reflectthat of springtime; however, the model was not able to predictrespiratory RR on DT days in the summer. This is due to a lownumber of DT days (average frequency of 2.1% for the summerseason), along with very few respiratory disease-related deathsoccurring on these days. CO and NO2 demonstrate the highest as-sociations with respiratory mortality in summer, with CO riskshighest in the MT weather type (1.263 (95% CI 1.067e1.495)). Thisestimate is significantly higher than DM and MM. Alternatively,NO2 risks are the highest in cold DP and MP weather types (1.123(95% 1.058e1.193)) and 1.151 (95% CI 1.066e1.242)), respectively.

Winter season respiratory mortality risk estimates are lowerthan that of spring and summer, and comparable to the fall. There isno significant difference between air pollution types, and althoughsome differences can be found between weather types, these arerarely significant. The dry-weather subsets (DM and DP) are of mostconcern, with each exhibiting an average increase of 7.8% in res-piratory mortality due to any air pollutant exposure. The DM

Fig. 3. Pooled relative risks of cardiovascular e (CVD) and respiratory-related mortality (at PAsterisk (*) indicates the RR estimate of respiratory (Resp) mortality to be significantly gre

weather type contains the highest average RR, and a 10.2% increasein risk due to O3 exposure. In the fall, risk estimates tend to be lowerthan in the remaining three seasons; however, in many cases theyare still statistical significant, with individual air pollutants asso-ciated with an average increase of 5e10% in respiratory relatedmortality (Table 3).

3.4. Cardiovascular-related mortality

Overall, the risk of cardiovascular-related mortality due to airpollution increased significantly from baseline for 86% of the casesmodelled. In the majority of cases, the increase in RR is 5% or lessdue to the individual air pollutants, and none greater than 10%(Table 3). The greatest differences between cardiovascular andrespiratory risk estimates are found in the spring seasonwithin theMT weather type, where respiratory risk estimates are 14% abovethat of cardiovascular risk for CO, and 11% above for the threeremaining pollutants. When hot and dry air is present (DT), resultsconsistently demonstrate average increases of 5e10% incardiovascular-related RR for all pollutants in the spring andsummer seasons. Separating the RR estimates for each air pollutanton the DT days in the springtime, cardiovascular risk due to SO2,NO2, and O3 is 3.5, 4.4, and 3.3% greater, respectively, than theaverage of the remaining weather types (Table 4). The greatestnumber of significantly different RR estimates was found whencomparing the DT weather type to the others. For example, duringsummertime DT weather, cardiovascular-related mortality due to

WM, with 95% CI), due to CO within six weather types (10 Canadian cities: 1981e1999).ater than mortality from CVD for the given weather type and season.

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332328

SO2 was significantly greater than the RR estimates in theremaining weather types.

The remaining pollutants (CO, NO2 and O3), demonstrate lessdifference by weather type, yet have a higher average risk ofcardiovascular-related mortality throughout the year, at 5.8, 4.5,and 6.0% average increases, respectively. Fall, winter, andspringtime exposures to SO2 result in the lowest estimates ofcardiovascular-related RR. The highest cardiovascular-relatedrisk estimates occur in the spring, specifically within DT air,and are weakest and least variable in the fall and winter(Figs. 3e6).

4. Discussion

In the current study, we analyse the association between short-term exposure to individual air pollutants and mortality within sixdifferent weather types and transitional weather, and stratify byseason and the cause of mortality.

Across Canada on a yearly basis, we find the greatest differencein RR within the hottest weather types of DT and MT (þ11.0and þ9.0%). Studies relating high temperature days to respiratoryand cardiovascular mortality (D’Ippoliti et al., 2010; Hoffmann et al.,2008) report a more prominent effect of heat waves on respiratorycauses, with the study in Europe by D’Ippoliti et al. (2010) sug-gesting this is due to increased susceptibility among people with

Fig. 4. Pooled relative risks of cardiovascular e (CVD) and respiratory-related mortality (at1999). Asterisk (*) indicates the RR estimate of respiratory (Resp) mortality to be significan

pre-existing chronic respiratory diseases. However, our findingssuggest that the interaction between temperature and specific airpollutants during these extreme heat days work synergistically tonegatively affect respiratory mortality, and to a smaller extentcardiovascular mortality. Anderson and Bell (2009) also foundrespiratory mortality effects to be greater in both cold and heat.Alternatively, some studies have found cardiovascular ailments tobe more sensitive to certain air pollutants than weather, namelyPM. For example, Stafoggia et al. (2008) found no interactions be-tween two environmental exposures (PM10 and temperature) withrespiratory mortality, yet there was a significant linear interactionwith cardiovascular mortality. The mortality effects of PM10 werealso significantly higher onwarmer days. Further, chronic exposure,rather than acute, has been shown to be strongly associated withcardiovascular mortality and air pollutants, with PM exposure ofmost concern (Dockery et al., 1993; Pope et al., 2004).

The significant differences that are found between weathertypes, specifically in the spring and summer seasons (Table 4,Figs. 3e6), reveal weather-pollution interactions. Pollution chemi-cal interactions and reaction rates vary with temperature, sunlight,and moisture conditions, resulting in higher pollution levels. This ismost evident with O3 in the DT weather type in this study andothers (Davis and Kalkstein, 1990; Davis et al., 2010), with highestconcentrations during the summer season (Fig. 1). High levels ofNOx in urbanized areas (due to combustion emissions from

PWM, with 95% CI), due to NO2 within six weather types (10 Canadian cities: 1981etly greater than mortality from CVD for the given weather type and season.

Fig. 5. Pooled relative risks of cardiovascular e (CVD) and respiratory-related mortality (at PWM, with 95% CI), due to SO2 within six weather types (10 Canadian cities: 1981e1999).Asterisk (*) indicates the RR estimate of respiratory (Resp) mortality to be significantly greater than mortality from CVD for the given weather type and season.

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332 329

vehicles, industry, and natural processes), lead to even greater O3production via the photolytic cycle (Finlayson-Pitts and Pitts, 2000;Notario et al., 2012). Ozone has been continuously highlighted inthe literature as a harmful ground level pollutant and is projected toincrease in the 21st century (Bell et al., 2007; Holloway et al., 2008;Knowlton et al., 2004). Synoptic weather typing accounts for thevariables responsible for this ozone formation (clear, dry, highpressure, low winds), and thus we demonstrate its use as a holisticway to account for and understand how the entire ambient situa-tion acts as a modifying factor on the air pollutionemortalityrelationship.

Correlations between air pollutants, as well as between pollu-tion and air temperature, often result in conflicting evidence for aneffect modification on human health outcomes; hence it becomesmore difficult to separate the model signals for individual variables.For example, Samoli et al. (2007) addressed confounding by airpollutants, where adjusting CO estimates by NO2 resulted inlowering of the single CO estimate, yet remained marginally sta-tistically significant. Smoyer et al. (2000a) found the MT weathertype to have a greater negative health impact on mortality thanhigh concentrations of total suspended (TSP) particles or O3 inBirmingham, USA. Similarly, extreme heat days in Australia (basedon the ‘temporal synoptic index’), significantly increased mortalityrisk, with O3 confounding the results on humid, hot days, and PMdoing so on dry, hot days (Vaneckova et al., 2008). Alternatively

Keatinge and Donaldson (2001) found that excess deaths wereassociated with cold weather patterns more so than ambient SO2and CO concentrations in the Greater London area (1976e1995).

Many studies report that the temperature effect on mortality isgreater when pollutant levels are higher, commonly referring to O3,and emphasize the synergism between air pollution and heat orcold. For example, Ren et al. (2007) found a positive modifyingeffect of O3, with higher levels resulting in stronger temperature-related cardiovascular mortality across different regions of theUnited States in the summer. Cheng et al. (2009) also foundelevated mortality associated with heat and air pollution in fivemajor Canadian cities, with 80% attributable to air pollution, and20% to temperature. It is suggested that such temperature effectsare likely to persist even after controlling for multiple air pollutantsin modelling, particularly O3 and PM2.5 (O’Neill et al., 2005). Chenget al. (2009) found that three pollutants (O3, SO2, and NO2) wereassociated with w75% of total mortality due to air pollution, withO3 the most significant, accounting for 33% of the total.

An understanding of the air pollution effects on cardiovascularevents has proven much more difficult to achieve than for respi-ratory disease (Brunekreef and Holgate, 2002). Further, specificcardiovascular diseases have been closely associated with heat-related mortality, such as ischemic heart disease, congestive heartfailure, and myocardial infarction (MI) (Basu, 2009). Hanna et al.(2011) found MI to be affected by O3 exposure only under MTþ

Fig. 6. Pooled relative risks of cardiovascular e (CVD) and respiratory-related mortality (at PWM, with 95% CI), due to O3 within six weather types (10 Canadian cities: 1981e1999).Asterisk (*) indicates the RR estimate of respiratory (Resp) mortality to be significantly greater than mortality from CVD for the given weather type and season.

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332330

(an extreme subset of MT) conditions, and asthma affected by O3under both DT and MTþ. Hence, extreme weather conditions plusair pollution can elicit a greater overall effect of the full atmosphericenvironment.

Higher winter mortality rates (as compared to summer) inCanada were also found by Rainham and Smoyer-Tomic (2003) andMartin et al. (2012). The latter researchers emphasized the findingthat cold-relatedmortality (versus heat-related) was a greater issueacross Canada (excess mortality per 100,000: 58.5 versus 1.4). Ex-posures to both hot and cold temperatures give rise to direct car-diovascular stress (Keatinge and Donaldson, 1995; Huynen andMartens, 2001; McGeehin and Mirabelli, 2001). Hence, the re-lationships found between cardiovascular-related mortality and airpollution may be confounded to a greater extent by weather, andaids in explaining why understanding the air pollution effects oncardiovascular disease aremore difficult to achieve (Brunekreef andHolgate, 2002).

Carbon monoxide has significant associations with mortality inall seasons, agreeing with other research that found significantassociations with all-cause mortality (Burnett et al., 1998b) andtotal and cardiovascular mortality (Samoli et al., 2007). Burnettet al. (1998b) found the mortality risk attributed to CO to begreater than that of any pollutant examined, and thus CO inparticular should be considered a potential cause of increased

mortality in urban populations. These researchers, and the currentstudy, have shown that CO, as well as NO2, are also equallyimportant components of the urban pollution mixture as PM andO3, if not more.

The presence of aeroallergens in the spring is a possible expla-nation for the heightened effects of air pollution across weathertypes; however, these were not accounted for in the current study.Aeroallergens have been associated with significant increased riskof asthma exacerbations, hospitalizations, allergic disease, andmorbidity (Reid and Gamble, 2009; Cakmak et al., 2011; Dales et al.,2004). Further heightened risk in spring may also be due to largefluctuations in weather, as well as negative impacts of early seasonhigh temperatures (Sheridan and Kalkstein, 2010). Cardiovascular-related RR in tropical weather was significantly greater in springthan in the summer, highlighting how time of season is a significanttime variable, as also found in earlier synoptic studies (Kalksteinand Davis, 1989; Kalkstein and Smoyer, 1993; Smoyer et al.,2000a,b).

The significant and elevated respiratory-related RR estimateson moderate or cool days indicates that all air pollutant exposureis cause for concern. Similar results were found by Cheng et al.(2009), where the ‘other’ groups (associated with good air qual-ity and comfortable conditions) were found to have elevatedmortality due to air pollution, with significance in four of five

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332 331

tested cities for specific pollutants. Such pollutant specific results,including here and by others (Basrur et al., 2001; Lambert et al.,1998; Vedal et al., 2002; Rainham et al., 2005), suggest that evi-dence to propose clear pollution concentration thresholds basedon mortality does not exist. Any reduction in ambient pollutionlevels is a benefit to human health, and no threshold has beendefined below which no effects to human health are likely(Brunekreef and Holgate, 2002).

Although the pooled estimates of all ten cities compare wellwith similar time-series pollution and temperature studies (e.g.,Anderson et al., 2004; Baccini et al., 2008; Stieb et al., 2003;Thurston and Ito, 2001), the combined values cannot be appliedto each city, as they are overall values for the entire country/area.City-specific information is provided in the Supplementary Tables;however, specific examination of these results should be conductedso that policies can be implemented that are city-specific. Further,city characteristics, such as topography, weather patterns, andanthropogenic pollution sources, can be considered. This can betteradvise local health care institutions, governments, and weatherservices to set targeted guidelines for air-pollution-healthrelationships.

5. Conclusions

Air pollution has a significant influence on acute humanmortality, which is overall greater than influences from weather,and varies with season and the entire weather situation present.Stronger effect estimates are found for respiratory mortalitythan cardiovascular or all-cause. The risk of dying due toexposure to all air pollutants from respiratory-related causes issignificantly higher than that of cardiovascular causes in 61% ofthe cases, with the RR found to be 6e10% greater, on average.The combined effect of weather and air pollution is greatestwhen tropical-type weather is present in the spring or summer.Dry tropical days are found to be most harmful in spring forboth causes of mortality, and in the summer for cardiovascular-related mortality. The spring season presents the overall great-est risk of respiratory-related death compared to cardiovascular;in particular, the risks due to air pollution exposure are higheston DT and MT days, with CO and NO2 the most harmful airpollutants.

These results underscore the importance of addressing thecause of mortality, full weather type, and season, when estimatingthe mortality risk of air pollution exposure. This epidemiologicalstudy gives evidence to environmental-health policy makers toproactively implement prevention strategies to reduce all levelsambient air pollutants. Further, since all weather and air pollutiontypes present significant health risks, it is suggested that anyreduction in ambient pollution levels is a benefit to human health.The investment in early warning systems based on synopticmeteorological forecasts, such as an integrated weather-pollutionindex, is suggested in order to predict accurate spatiotemporalhealth outcomes for the public. This study demonstrates the po-tential usefulness in estimating precise short-term future health/pollution problems associated with specific incoming synopticweather patterns for each season.

Acknowledgements

The authors would like to thank Environment Canada forproviding the air pollution data and the reviewers for their helpfulcomments. Funding for completion of this project was generouslyprovided to Dr. Jennifer Vanos by a Natural Sciences and Engi-neering Research Council (NSERC) Postdoctoral Fellowship fromthe Government of Canada.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.envpol.2013.11.007.

References

Anderson, B.G., Bell, M.L., 2009. Weather-related mortality: how heat, cold, andheat waves affect mortality in the United States. Epidemiology (Cambridge,Mass.) 20 (2), 205.

Anderson, H.R., Bremner, S.A., Atkinson, R.W., Harrison, R.M., Walters, S., 2001.Particulate matter and daily mortality and hospital admissions in the westmidlands conurbation of the United Kingdom: associations with fine and coarseparticles, black smoke and sulphate. Occup. Environ. Med. 58, 504e510.

Anderson, H.R., Atkinson, R.W., Peacock, J.L., Marston, L., Konstantinou, K., 2004.Meta-analysis of Time-series Studies and Panel Studies of Particulate Matter(PM) and Ozone (O3). Report of a WHO Task Group, Report of a WHO TaskGroup. World Health Organization, Copenhagen.

Baccini, M., Biggeri, A., Accetta, G., Kosatsky, T., Katsouyanni, K., Analitis, A.,Anderson, H.R., Bisanti, L., D’Ippoliti, D., Danova, J., Forsberg, B., Medina, S.,Paldy, A., Rabczenko, D., Schindler, C., Michelozzi, P., 2008. Heat effects onmortality in 15 European cities. Epidemiology 19, 711e719.

Basrur, S., Pengelly, D., Campbell, M., Macfarlane, R., Li-Muller, A., 2001. Toronto AirQuality Index: Health Links Analysis. Report. Toronto Public Health, Toronto.

Basu, R., 2009. High ambient temperature and mortality: a review of epidemiologicstudies from 2001 to 2008. Environ. Health 8, 40.

Basu, R., Samet, J.M., 2002. Relation between elevated ambient temperature andmortality: a review of the epidemiologic evidence. Epidemiol. Rev. 24, 190e202.

Bell, M.L., Peng, R.D., Dominici, F., 2006. The exposure-response curve for ozone andrisk of mortality and the adequacy of current ozone regulations. Environ. HealthPerspect. 114, 532e536.

Bell, M.L., Goldberg, R., Hogrefe, C., Kinney, P.L., Knowlton, K., Lynn, B., Rosenthal, J.,Rosenzweig, C., Patz, J.A., 2007a. Climate change, ambient ozone, and health in50 US cities. Clim. Change 82, 61e76.

Bell, M.L., Kim, J.Y., Dominici, F., 2007b. Potential confounding of particulate matteron the short-term association between ozone and mortality in multisite time-series studies. Environ. Health Perspect. 115, 1591e1595.

Brunekreef, B., Holgate, S.T., 2002. Air pollution and health. Lancet 360, 1233e1242.Burnett, R.T., Cakmak, S., Brooke, J.R., 1998a. The effect of the urban ambient air

pollution mix on daily mortality rates in 11 Canadian cities. Can. J. Public Health89 (3), 152e156.

Burnett, R.T., Cakmak, S., Raizenne, M.E., Stieb, D., Vincent, R., Krewski, D.,Brook, J.R., Philips, O., Ozkaynak, H., 1998b. The association between ambientcarbon monoxide levels and daily mortality in Toronto, Canada. J. Air WasteManag. Assoc. 48, 689e700.

Cakmak, S., Dales, R.E., Judek, S., 2006. Respiratory health effects of air pollutiongases: modification by education and income. Arch. Environ. Occup. Health 61,5e10.

Cakmak, S., Dales, R.E., Rubio, M.A., Vidal, C.B., 2011. The risk of dying on days ofhigher air pollution among the socially disadvantaged elderly. Environ. Res. 111.

Cheng, C., Campbell, M., Li, Q., Li, G., Auld, H., Day, N., Pengelly, D., Gingrich, S.,Klaassen, J., MacIver, D., Comer, N., Mao, Y., Thompson, W., Lin, H., 2009. Dif-ferential and combined impacts of extreme temperatures and air pollution onhuman mortality in south-central Canada. Part I: historical analysis. Air Qual.Atmos. Health 1, 209e222.

Choi, K.S., Inoue, S., Shinozaki, R., 1997. Air pollution, temperature, and regionaldifferences in lung cancer mortality in Japan. Arch. Environ. Health 52,160e168.

Confalonieri, U., Menne, B., Akhtar, R., Ebi, K.L., Hauenque, M., Kovats, R.S., Revich, B.,Woodward, A., 2007. Human health. Climate change 2007: impacts, adaptationand vulnerability. In: Contribution of Working Group II to the Fourth Assess-ment Report of the Intergovernmental Panel on Climate Change. CambridgeUniversity Press, Cambridge, UK.

Dales, R.E., Cakmak, S., Judek, S., Dann, T., Coates, F., Brook, J.R., Burnett, R.T., 2004.Influence of outdoor aeroallergens on hospitalization for asthma in Canada.J. Allergy Clinical Immunol. 113, 303e306.

Dales, R.E., Cakmak, S., Doiron, M.S., 2006. Gaseous air pollutants and hospitaliza-tion for respiratory disease in the neonatal period. Environ. Health Perspect.114, 1751e1754.

Davis, R.E., Kalkstein, L.S., 1990. Using a spatial synoptic climatological classificationto assess changes in atmospheric pollution concentrations. Phys. Geogr. 11,320e342.

Davis, R.E., Normile, C.P., Stika, L., Hondula, D.M., Knight, S.D., Gawtry, S.D.P., ,et al.Sitka, L., Knight, D.B., Stenger, P.J., 2010. A comparison of trajectory and airmass approaches to examine ozone variability. Atmos. Environ. 44, 64e74.

Dockery, D.W., Pope III, C.A., 1994. Acute respiratory effects of particulate airpollution. Annu. Rev. Public Health 15, 107e132.

Dockery, D.W., Pope, C.A., Xu, X., Spengler, J.D., Ware, J.H., Fay, M.E., Speizer, F.E.,1993. An association between air pollution and mortality in six US cities. NewEngl. J. Med. 329 (24), 1753e1759.

D’Ippoliti, D., Michelozzi, P., Marino, C., De’Donato, F., Menne, B., Katsouyanni, K.,Kirchmayer, U., Analitis, A., Medina-Ramón, M., Paldy, A., 2010. The impact of

J.K. Vanos et al. / Environmental Pollution 185 (2014) 322e332332

heat waves on mortality in 9 European cities: results from the EuroHEATproject. Environ. Health 9, 1e9.

Filleul, L., Cassadou, S., Medina, S., Fabres, P., Lefranc, A., Eilstein, D., Le, T.A.,Pascal, L., Chardon, B., Blanchard, M., Declercq, C., Jusot, J.F., Prouvost, H.,Ledrans, M., 2006. The relation between temperature, ozone, and mortality innine French cities during the heat wave of 2003. Environ. Health Perspect. 114,1344e1347.

Finlayson-Pitts, D.J., Pitts, J.N.J., 2000. Chemistry of the Upper and Lower Atmo-sphere: Theory, Experiments and Application. Wiley, New York.

Hajat, S., Kosatky, T., 2010. Heat-related mortality: a review and exploration ofheterogeneity. J. Epidemiol. Community Health 64 (9), 753e760. http://dx.doi.org/10.1136/jech.2009.087999.

Hales, S., Salmond, C., 2000. Daily mortality in relation to weather and air pollutionin Christchurch, New Zealand. Aust. N. Z. J. Public Health 24, 89e91.

Hanna, A.F., Yeatts, K.B., Xiu, A., Zhu, Z., Smith, R.L., Davis, N.N., Talgo, K.D.,Arora, G., Robinson, P.J., Meng, Q., Pinto, J.P., et al., 2011. Associations be-tween ozone and morbidity using the spatial synoptic classification system.Environ. Health 10, 49.

Hoffmann, B., Hertel, S., Boes, T., Weiland, D., Jöckel, K., 2008. Increased cause-specific mortality associated with 2003 heat wave in Essen, Germany.J. Toxicol. Environ. Health A 71, 759e765.

Holloway, T., Spak, S.N., Barker, D., Bretl, M., Moberg, C., Hayhoe, K., Van Dorn, J.,Wuebbles, D., 2008. Change in ozone air pollution over Chicago associated withglobal climate change. J. Geophys. Res. Atmos. 113.

Huynen, M., Martens, P., 2001. The impact of heat waves and cold spells on mor-tality rates in the Dutch population. Environ. Health Perspect. 109, 463e470.

Kalkstein, L.S., Davis, R.E., 1989. Weather and human mortality: an evaluation ofdemographic and inter-regional responses in the United States. Ann. Assoc. Am.Geogr. 79, 44e64.

Kalkstein, L.S., Smoyer, K.E., 1993. The Impact of Climate Change on Human Health:Some International Implications. Experientia, Downsview, Ontario.

Keatinge, W.R., Donaldson, G.C., 1995. Cardiovascular mortality in winter. Arct. Med.Res. 54 (Suppl. 2).

Keatinge, W., Donaldson, G., 2001. Winter deaths: warm housing is not enough. Br.Med. J. 323, 166e167.

Knowlton, K., Rosenthal, J.E., Hogrefe, C., Lynn, B., Gaffin, S., Goldberg, R.,Rosenzweig, C., Civerolo, K., Ku, J.Y., Kinney, P.L., 2004. Assessing ozone-relatedhealth impacts under a changing climate. Environ. Health Perspect. 112, 1557e1563.

Lambert, W.E., Samet, J.M., Dockery, D.W., 1998. Community air pollution. In:Rom, W.N. (Ed.), Environmental and Occupational Medicine. Lippincott-Raven,Philadelphia, pp. 1501e1522.

Levy, J.I., Chemerynski, S.M., Sarnat, J.A., 2005. Ozone exposure and mortality: anempiric bayes metaregression analysis. Epidemiology 16, 458e468.

Martin, S., Cakmak, S., Hebbern, C., Avramescu, M.-L., Tremblay, N., 2012. Climatechange and future temperature-related mortality in 15 Canadian cities. Int. J.Biometeorol. 56, 605e619.

McGeehin, M.A., Mirabelli, M., 2001. The potential impacts of climate variability andchange on temperature-related morbidity and mortality in the United States.Environ. Health Perspect. 109 (Suppl.), 185e189.

Notario, A., Bravo, I., Adame, J.A., Diaz-de-Mera, Y., Aranda, A., Rodriguez, A., et al.,2012. Analysis of NO, NO2, NOx, O3 and oxidant (OX ¼ O3 þ NO2) levelsmeasured in a metropolitan area in the southwest of Iberian Peninsula. Atmos.Res. 104e105, 217e226.

O’Neill, M.S., Hajat, S., Zanobetti, A., Ramirez-Aguilar, M., Schwartz, J., 2005. Impactof control for air pollution and respiratory epidemics on the estimated associ-ations of temperature and daily mortality. Int. J. Biometeorol. 50, 121e129.

Pope III, C.A., Kalkstein, L.S.S., 1996. Synoptic weather modeling and estimates ofthe exposure-response relationship between daily mortality and particulate airpollution. Environ. Health Perspect. 104.

Pope, C.A., Burnett, R.T., Thurston, G.D., Thun, M.J., Calle, E.E., Krewski, D.,Godleski, J.J., 2004. Cardiovascular mortality and long-term exposure to par-ticulate air pollution epidemiological evidence of general pathophysiologicalpathways of disease. Circulation 109 (1), 71e77.

Rainham, D., Smoyer-Tomic, K., 2003. The role of air pollution in the relationship be-tween a heat stress index and humanmortality in Toronto. Environ. Res. 93, 9e19.

Rainham, D.D.D.G.C., Smoyer-Tomic, K.K.E., Sheridan, S.C.S., Burnett, R.T.R., 2005.Synoptic weather patterns and modification of the association between airpollution and human mortality. Int. J. Environ. Health Res. 15, 347e360.

Reid, C.E., Gamble, J.L., 2009. Aeroallergens, allergic disease, and climate change:impacts and adaptation. Ecohealth 6, 458e470.

Ren, C., Tong, S., 2006. Temperature modifies the health effects of particulate matterin Brisbane, Australia. Int. J. Biometeorol. 51, 87e96.

Ren, C., Williams, G.M., Tong, S., 2006. Does particulate matter modify the associ-ation between temperature and cardiorespiratory diseases? Environ. HealthPerspect. 1690, 1690e1696.

Ren, C., Williams, G.M., Morawska, L., Mengersen, K., Tong, S., 2007. Ozone modifiesassociations between temperature and cardiovascular mortality analysis of theNMMAPS data. Occup. Environ. Med. 65, 255e260.

Roberts, S., 2004. Interactions between particulate air pollution and temperature inair pollution mortality time series studies. Environ. Res. 96, 328e337.

Samet, J., Zeger, S., Kelsall, J., Xu, J., Kalkstein, L., 1998. Does weather confound ormodify the association of particulate air pollution with mortality? An analysisof the Philadelphia data, 1973e1980. Environ. Res. 77, 9e19.

Samoli, E., Touloumi, G., Schwartz, J., Anderson, H.R., Schindler, C., Forsberg, B.,Vigotti, M.A., Vonk, J., Kosnik, M., Skorkovsky, J., Katsouyanni, K., 2007. Short-term effects of carbon monoxide on mortality: an analysis within the APHEAproject. Environ. Health Perspect. 115, 1578e1583.

Schwartz, J., 2000. Harvesting and long term exposure effects in the relation be-tween air pollution and mortality. Am. J. Epidemiol. 151, 440e448.

Sheridan, S.C., 2002. The redevelopment of a weather-type classification scheme forNorth America. Int. J. Climatol. 22, 51e68.

Sheridan, S.C., Kalkstein, A.J., Kalkstein, L.S., 2009. Trends in heat-related mortalityin the United States, 1975-2004. Nat. Hazards 50 (1), 145e160.

Sheridan, S.C., Kalkstein, A.J., 2010. Seasonal variability in heat-related mortalityacross the United States. Nat. Hazards 55, 291e305.

Smoyer, K.E., Kalkstein, L.S., Greene, J.S., Ye, H., 2000a. The impacts of weather andpollution on human mortality in Birmingham, Alabama and Philadelphia,Pennsylvania. Int. J. Climatol. 20, 881e897.

Smoyer, K.E., Rainham, D.G.C., Hewko, J.N., 2000b. Heat-stress-related mortality infive cities in Southern Ontario: 1980e1996. Int. J. Biometeorol. 44, 190e197.

Smoyer-Tomic, K.E., Kuhn, R., Rainham, D.G.C., 2003. Heat Wave Hazards: anoverview of heat wave impacts in Canada. Nat. Hazards 28.

Stafoggia, M., Schwartz, J., Forastiere, F., Perucci, C.A., 2008. Does temperaturemodify the association between air pollution and mortality? A multicity case-crossover analysis in Italy. Am. J. Epidemiol. 167 (12), 1476e1485.

Stieb, D.M., Judek, S., Burnett, R.T., 2003. Meta-analysis of time-series studies of airpollution and mortality: update in relation to the use of generalized additivemodels. J. Air Waste Manag. Assoc. 53, 258e261.

Sujaritpong, S., Dear, K., Cope, M., Walsh, S., Kjellstrom, T., 2013. Quantifying thehealth impacts of air pollution under a changing climateda review of ap-proaches and methodology. Int. J. Biometeorol., 1e12.

Thurston, G., Ito, K., 2001. Epidemiological studies of acute ozone exposures andmortality. J. Expo. Anal. Environ. Epidemiol. 11, 286e294.

USEPA, 2004. The Ozone Report: Measuring Progress Through 2003. EnvironmentalResearch, Research Triangle Park, NC.

Vaneckova, P., Beggs, P.J., De Dear, R.J., McCracken, K.W.J., 2008. Effect of tempera-ture on mortality during the six warmer months in Sydney, Australia, between1993 and 2004. Environ. Res. 108, 361e369.

Vanos, J.K., Cakmak, S., 2013. Changing air mass frequencies in Canada: potentiallinks and implications for human health. Int. J. Biometeorol. http://dx.doi.org/10.1007/s00484-013-0634-2.

Vedal, S., Brauer, M., White, R., Petkau, J., 2002. Air pollution and daily mortality in acity with low levels of pollution. Environ. Health Perspect. 111, 45e51.

Zanobetti, A., Schwartz, J., 2005. The effect of particulate air pollution on emergencyadmissions for myocardial infarction: a multicity case-crossover analysis. En-viron. Health Perspect. 113, 978e982.