UBC Faculty Research and Publications

Evaluation of ambient air pollution in the Lower Mainland of British Columbia: Public health impacts,.. Brauer, Michael 2008

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
LMAirQualityReport.pdf
LMAirQualityReport.pdf [ 5.94MB ]
LMAirQualityReport.pdf
Metadata
JSON: 1.0048194.json
JSON-LD: 1.0048194+ld.json
RDF/XML (Pretty): 1.0048194.xml
RDF/JSON: 1.0048194+rdf.json
Turtle: 1.0048194+rdf-turtle.txt
N-Triples: 1.0048194+rdf-ntriples.txt
Citation
1.0048194.ris

Full Text

 Evaluation of ambient air pollution in the Lower Mainland of British Columbia: Public health impacts, spatial variability, and temporal patterns   Michael Brauer, ScD1  Jochen Brumm, MSc2  Stefanie Ebelt, MSc1   The University of British Columbia 1School of Occupational and Environmental Hygiene 2Department of Statistics   July 17, 2000  Final Report to:  Dr. John Blatherwick Chair, Administrative Council of Lower Mainland Medical Health Officers c/o Vancouver- Richmond Regional Health Board 1060 W 8th Avenue Vancouver, BC V6H 1C4   DO NOT QUOTE OR CITE i Executive Summary  British Columbia Lower Mainland air quality data for the period 1994-1998 inclusive were obtained and used to assess the public health impacts associated with ambient air pollution. This assessment included 5 components:  1) an estimation of deaths attributable to air pollution using data from a local epidemiological study  2) comparison of the estimated attributable deaths to other selected attributable causes of death  3) comparison of air pollutant concentrations in the Lower Mainland to those measured in selected western U.S. metropolitan areas  4) analysis of the spatial distribution of major air pollutants  5) assessment of temporal trends in air pollutant concentrations and their spatial patterns    Attributable death estimates  Since the estimation of the number of deaths which may be attributable to air pollution is strongly dependent upon assumptions for no-effect thresholds, we evaluated several scenarios based upon different "low pollution" levels at which we assumed no effect of air pollution on mortality. The numbers of attributable deaths were then estimated on the basis of a factor which reflects the difference between actual levels of pollution on a given day and the selected "low pollution" level. This factor was then multiplied by the observed baseline mortality rate for the Lower Mainland region.  If the actual pollution level was below the defined “low pollution” level, then the predicted number of deaths was 0. Depending upon the specified "low pollution" level (which ranged from the 10th percentile of measured values to selected Health Guideline values from Environment Canada, WHO or U.S. EPA) mean estimated of attributable deaths ranged from 0 to approximately 600 per year. Essentially all of these estimated deaths were for individuals greater than 65 years old and were primarily associated with cardiovascular causes. These mortality estimates indicated the potential for air pollution to be associated with numbers of deaths which were comparable to those attributable to causes of death such as motor vehicle traffic accidents, suicides and HIV, but much lower than the numbers of deaths attributable to smoking.    Comparison of air pollutant concentrations  In comparison with Western US cities of comparable population, average concentrations of major air pollutants measured in the Vancouver region were quite low, although occasionally short-term peak concentrations (especially for PM10) are reached which are as high or higher than peak concentrations reached in other cities.  Specifically, annual average concentrations of PM10 were lower than those measured in each of the metropolitan areas used for comparison and well below the GVRD objective and U.S. NAAQS for annual average. Measures of peak 24-hour PM10 concentrations indicate, however, that this region does experience occasional maximum concentrations that are higher than several of the comparison locations and that approach and even exceed the GVRD objective. For NO2, the mean annual average concentration in Vancouver was ii higher than in some locations and well below the Canadian Level B objective and the U.S. NAAQS, but only slightly (14%) below the WHO guideline value. The mean SO2 annual average was similar to those from the comparison locations and well below any of the guideline values. For CO, concentrations were well below any of the guideline values and generally below those measured in the majority of the comparison locations. Regardless of the metric and averaging method used, O3 concentrations in the Lower Mainland region were below those measured in all of the comparison areas, although still occasionally exceeding the Canadian Level B objective.    Spatial and temporal patterns  The assessment of spatial and temporal patterns in air pollutant concentrations used statistical interpolation software to estimate air pollutant concentrations at locations without monitoring data. These estimates are based upon distance-weighted correlations between measured concentrations and do not explicitly incorporate physical or chemical factors. This assessment revealed no strong temporal trends between 1994-1998 with the exception of CO which experienced a slight decrease. Strong seasonal patterns were observed for CO (winter > summer), O3 (summer > winter) and PM10 (summer > winter). No seasonal pattern was evident for NO2.  Interpolation was not possible for SO2 due to the poor predictive ability of interpolated values for this pollutant. CO and NO2 concentrations were higher in Vancouver relative to locations in the eastern part of the region. In contrast, O3 concentrations were significantly higher in the Fraser Valley relative to Vancouver/Burnaby. PM10 did not exhibit much spatial variability, with concentrations being relatively homogeneously distributed within the region.  1 Introduction  This report described work that has been conducted in response to a request from the Administrative Council of Lower Mainland Medical Health Officers to pursue an evaluation of ambient air quality in the British Columbia Lower Mainland in order to better understand the potential public health impacts. With this objective, the was analysis had the following specific aims  1. To apply data from a local epidemiological analysis to estimate the numbers of annual deaths that can be attributed to air pollution within the Lower Mainland (Vancouver – Hope) airshed.  2. To compare these estimates to selected BC attributable causes of death statistics.  3. To compare summary statistics of Lower Mainland air quality (1994-1998) to those of other western North American metropolitan areas with similar populations.  4. To conduct an analysis of the spatial variability of summary pollutant measures for:  • Ozone (O3) • Sulfur dioxide (SO2) • Carbon monoxide (CO) • Nitrogen dioxide (NO2) • Inhalable Particles (PM10)  using various summary measures (annual averages, daily maximum values, seasonal averages, etc.) and resulting in graphical displays of air pollution contours.  5. To conduct analysis of temporal trends for the same pollutants and period indicated above.  It should be noted that these analyses were conducted based upon available air quality data routinely collected by the Greater Vancouver Regional District.  As such, analyses are restricted to priority air pollutants and do not include other measurements conducted in the context of research or temporary monitoring programs. Further, the analyses described in this report do not consider measurements of visibility per se, although visibility measurements may be as or more important, in terms of public perception of air quality, than the measurement of specific pollutants. The assessment of health impacts described here is limited to acute mortality; numerous other health outcomes (chronic effects, morbidity) have also been associated with air pollution although these are not addressed in this analysis.  2 I. Estimating the Number of Deaths Attributable to Air Pollution in the BC Lower Mainland We want to estimate the number of deaths per year that can be attributed to air pollution. We base this inference on the statistical model developed in an epidemiological study of air pollution and daily mortality in the Lower Mainland1. This model relates the number of deaths d(t) on day t in the Lower Mainland and air pollution for the concentration of a particular pollutant p(t) as  log d(t) = f1(t) + f2(t) + β p(t)                             (1)  where the function f1(t) adjusts for long-term cycles and the function f2(t) adjusts for the meteorological variables temperature and relative humidity. A more detailed description of this model is given in Appendix 1.  We stratified the number of deaths by cause of death (respiratory= ICD9 codes 460-519, circulatory = ICD9 codes 390-459, other = all ICD9 codes excluding those above 800 (trauma) and excluding those in circulatory and respiratory categories, total = sum of circulatory, respiratory and other) and age (younger than 65, older than 65 and all ages); we have therefore considered 12 strata in total. To use model (1) for prediction, for each pollutant and each stratum we first looked at lags2 of 0-2 days individually and then chose the lag period that had the largest positive association between the air pollutant and the mortality count. We introduced these chosen lags for each of the pollutants simultaneously in model (1) and estimated the regression coefficients using the data from the years 1994-1996 inclusive. Table 1 shows the lag periods that were selected for each pollutant and each stratum, since we allow these lags to be different for different strata.  Table 1.  Lags for each strata and pollutant.    PM10   SO2 CO  O3 NO2  Respiratory, all ages   1 1 2 2 2 Respiratory, >65 years old 1 1 2 2 2 Respiratory, < 65 years old 2 1 0 1 0 Circulatory, all ages 1 1 2 0 0 Circulatory, > 65 years old 0 2 2 0 0 Circulatory, < 65 years old Total, all ages 2 1 2 0 0 Total, > 65 years old 1 1 2 0 0 Total < 65 years old Other, all ages 2 1 0 0 0 Other, > 65 years old 2 1 1 0 1 Other, < 65 years old  Note that no lags are given for circulatory, total, and other deaths for those under age 65, since mortality coefficients were negative and therefore were not included in model. This is likely due to low numbers of deaths in these strata.  1 Vedal S, Brauer M, White R, Petkau J.  Very low concentrations of PM10 and daily mortality. American Journal of Respiratory and Critical Care Medicine. 1999; 159(3): A322  2 a lag refers to the period between the exposure and effect which is considered in the analysis. A lag of 1 means that the association is between air pollution recorded one day before the death count; a lag of 0 means the association is between air pollution and death count on the same day. 3   Using these estimated regression coefficients, we predict for the years 1994-1998 inclusive the number of deaths expected to occur at a “low pollution” level. Since the definition of “low pollution” is relatively arbitrary, we considered five scenarios: the 10th, 25th, 50th, 75th and 90th quantiles from the respective air pollution data, and several health guideline values for the different pollutants (indicated in Bold in Table 2).  Table 2.  Ambient Air Quality Standards/Guidelines for Canada, United States, World Health Organization. Pollutant Averaging time Canadian Level B Objective (Proposed) Canada-Wide Standards3 WHO Guidelines4 EPA NAAQS5 SO2 1 hr 24 hr annual 0.34 0.11 0.02 NA - 0.048 0.019 - 0.14 0.03 NO2 1 hr 24 hr annual 0.210 0.110 0.050 NA 0.106 - 0.021 - - 0.053 CO 1 hr 8 hr 30 13 NA 26 9 35 9 O3 1 hr 8 hr 24 hr 0.082 - 0.025  0.065 - 0.06 - 0.120 0.08 - Particulate < 10 µm (PM10) 24 hr annual 506 306 30 No guideline value (impact relationship) 150 50 Particulate < 2.5 µm (PM2.5) 24 hr annual  30 No guideline value (impact relationship) 65 15 All concentrations in ppm except PM in µg/m3 Guideline values used in assessment are indicated in bold.   To predict the daily number of deaths attributable to pollution, we multiply the observed number of deaths with a factor that reflects the difference in actual pollution and the “low pollution” level on this day. If the actual pollution is below the defined “low pollution” level, then the predicted number of deaths is set to 0. The annual number of deaths is then obtained by summing these values for the entire time-period 1994-1998 and dividing it by the number of years. With this approach we estimate the number of actual deaths that could be attributable due to air pollution.  3 A Canada-wide standard for PM10 was not put forward. Standards are: PM2.5 = 30 ug/m3 (24 hr, 98th percentile, averaged over 3 years), Ozone = 0.065 (8hr, 4th highest reading, averaged over 3 years). http://www.ccme.ca/3e_priorities/3ea_harmonization/3ea2_cws/3ea2.html  4 http://www.who.int/peh/air/airqualitygd.htm  5 http://www.epa.gov/airprogm/oar/oaqps/greenbk/criteria.html  6 GVRD Objective (No Canadian Objective) 4 The mean number of deaths from all causes in the Lower Mainland is 12939 per year (35 per day). Appendix 1 gives the technical details for this approach.  All of the measured criteria pollutants were entered simultaneously into the model.  Although we entered into the model the lag periods corresponding to the largest positive coefficient between an individual air pollutants and the mortality count, in the final model we considered all pollutants simultaneously as we did not wish to differentiate between individual pollutants. Therefore, individual pollutants may have negative coefficients in this model. We also only counted days for which observed pollution was above the respective cut-off levels. In this way we would avoid counting any “negative” deaths.  The time-series plots of the individual air quality metrics (Figures 1-8) show that the guideline values are much greater than the actual air pollution levels on almost every day; therefore the predicted number of deaths is 0 for this “low pollution” scenario. A dashed line on the x-axis of Figures 1-8 indicates the period covered by the epidemiological study, from which the coefficients were generated.  The concentrations indicated in Figures 1-8 are averages of measurements collected at all available GVRD monitoring locations and are therefore not representative of peak values which may be experienced at individual locations. Spatially averaged values were used in the estimation of attributable deaths, as this was the same approach used in the epidemiological analysis upon which these estimates are based.  Figures 9-12 show the (mean ± standard error) estimated number of deaths for the different strata and for the different “low pollution” cut-of levels. Note that only the estimated numbers of deaths for “all ages” and ages >65 years are shown. Estimated deaths for ages <65 years were either zero or very low.  The majority of attributable deaths are found in the circulatory category, followed by the 'other' (non-circulatory, non-respiratory, non-trauma) category. Respiratory deaths accounted for a small proportion of the estimated attributable deaths.  Essentially all of the estimated attributable deaths are for individuals older than 65. The estimates vary widely due to the large standard errors and the assumption of "low pollution" level. The mean worst-case estimate indicates approximately 600 deaths per year attributable to air pollution, or approximately 4.6% of total non-trauma deaths. The upper limit worst case estimate indicates approximately 900 deaths per year or 7% of total non- trauma deaths. Estimating attributable deaths in this manner assumes that the vast majority of such attributable deaths occur when air pollution concentrations are below health guideline/standard levels. Use of these health guideline values as indicators of no-effect level, however, would indicate that there are no deaths in the Lower Mainland that are attributable to air pollution. It should be recognized that studies have indicated evidence for health impacts at levels below those of health guidelines / standards.  It is for this reason that a range of “low pollution” cut-off levels are used in this analysis. The actual threshold level, if any, below which no effects occur in the population is uncertain.   5  6 7  8  9 10 11 II. Comparison of estimated air pollution attributable deaths to with selected attributable causes of death The estimated numbers of deaths, derived in Section I above, can then be compared to selected attributable causes of deaths, obtained from the BC Vital Statistics Agency (Table 3). BC Vital Statistics Agency Data for “Lower Mainland” includes the following Health Regions: Vancouver/Richmond, Simon Fraser/Burnaby, North Shore, Fraser Valley, South Fraser Valley7.  Table 3.  Annual attributable deaths for selected causes in Lower Mainland (1994-1998 estimates) Cause Number (attributable deaths/year) Alcohol-related 869 Drug-induced 311 Suicide 239 HIV 177 Motor Vehicle Traffic Accidents 169 Accidental Falls 167 Smoking8 4446  This approach has many limitations. For example, it can be easily seen that the estimates are highly sensitive to the “low pollution” cut-off point that is chosen. Further, although we evaluate all pollutants simultaneously, only those pollutants with positive coefficients (i.e. those associating increased air pollution with increased daily mortality) have been included. Any pollutants with negative coefficients have not been included, as we do not want to count any “negative” deaths. It should also be considered that an epidemiological association between air pollution and daily mortality does not in fact mean that all such attributable deaths are in fact caused by air pollution. There are no clinical features which would allow for a diagnosis of an air pollution-related death. While the estimated attributable deaths associated with air pollution are of similar magnitude to drug induced deaths or those from suicide, motor vehicle accidents and accidental falls, they are lower than alcohol-related deaths and much lower than deaths attributable to smoking.  Estimated deaths attributable to air pollution are limited to those greater than 65 year of age, whereas the other causes of death indicated in Table 3 may be more evenly distributed across ages or in some cases restricted to those of much younger age and therefore may be more significant in terms of public health significance.  7 Selected vital statistics and health status indicators. 1998 Annual Report. The British Columbia Vital Statistics Agency.  8 Crude estimate based on 1998 provincial total and ratio of Lower Mainland population to provincial population.  12 III.  Comparison of US cities and Vancouver We compared the levels of selected criteria air pollutants for selected Vancouver and US cities. Table 4 lists the population of the Lower Mainland region relative to the selected US metropolitan areas. U.S. locations with similar populations located in the western half of the country were included for comparison. Although it has a substantially greater population, Los Angeles was also included due to specific interest from Lower Mainland residents regarding local air quality as it related to air quality in Los Angeles.  U.S. Air quality data were obtained from the U.S. EPA AIRS database summary data9.  All industrial area sites were excluded. For parameters with daily (24 hour) data, only sites with greater than 50 measurement days per year were included. For hourly parameters (1 hour and 8 hour data) only sites with greater than 1500 measurement hours per year were included.  Table 5 lists the counties that were included in each metropolitan area.  Lower Mainland Air Quality data obtained from the GVRD for the period January 1994- December 1998 from 21 stations.  This includes all operating stations from the Vancouver Airport (T31) east to Hope (T29) with the exception of the Burmount, Capitol Hill, and Burnaby North (T24) sites that are specifically located to monitor industrial emissions and are therefore not appropriate as indicators of ambient air quality. The data file for the US cities contained, for each exposure metric (e.g. 2nd highest 8-hour max), one number per year per station. We calculated the respective numbers for Vancouver and then compared the cities in 3 ways. This was done so that the Vancouver data would be comparable to the limited reporting format of the U.S. data.  (i) Annual averages (Tables 6-8). For the annual average metric the two approaches described below result in the same values.  (ii) Averages across stations and years for each metric (Tables 9-16). In this approach we average all of the summary measures from all monitoring stations and average them over all five years. For example, we identify the 2nd highest 1-hr ozone concentration from each location, and take the average of these values from all stations and all years. This approach would tend to reduce the impact of one or several sites with particularly high measurements and is probably most appropriate for assessing the relative concentrations between the different metropolitan areas. However, the actual concentrations listed are averages.  (iii) Maximum of stations and years for each metric (Tables 17-24). In this approach we identify the highest values for each metric for each site and for all years. For the example above, we would identify the 2nd highest 1-hr ozone concentration from each location for each year and then select the single highest value of these from all sites and years. This approach will give concentration values that better represent the highest concentrations within the entire region during the 5-year period of interest (with the restriction that the values are summarized by site and year before they are selected; this is done to correspond to the U.S. data format). However, isolated extreme measurements will be highly influential in the rankings.  As indicated in the following tables, annual average concentrations of PM10 were lower than those measured in each of the metropolitan areas used for comparison and well below the GVRD objective and U.S. NAAQS for annual average. Measures of peak 24-hour PM10 concentrations (Tables 14-16, 22-24) indicate, however that this region does experience occasional maximum concentrations that are higher than several of the comparison locations and that approach and even exceed the GVRD objective. Rankings shown in Tables 22-24 indicate individual maximum measured concentrations and therefore are strongly influenced by individual high readings from  9 (http://www.epa.gov/airsweb/monreps.htm Accessed December 3, 1999). 13 individual sites, while those indicated in Tables 14-16 were more representative of regional maximum concentrations. For NO2, the annual average concentration in Vancouver was higher than in some locations, there was little variability in the concentrations for the lower half of the distribution. The annual average for the Vancouver area was well below the Canadian Level B objective and the U.S. NAAQS, but only slightly (14%) below the WHO guideline value. The SO2 annual average was similar to those from the comparison locations and well below any of the guideline values. For CO, concentrations were well below any of the guideline values and generally below those measured in the majority of the comparison locations. Regardless of the metric and averaging method used, O3 concentrations in the Lower Mainland region were below those measured in all of the comparison areas, although still occasionally exceeding the Canadian Level B objective. In summary, in comparison with Western US cities of comparable population, average concentrations of major air pollutants measured in the Vancouver region were quite low, although occasionally short-term peak concentrations are reached which are as high or higher than peak concentrations reached in other cities and above health guideline values.   14 Table 4.  Population of Lower Mainland and selected comparison metropolitan areas Metropolitan Area Population10 Loa Angeles 15549614 San Francisco - Oakland, CA PMSA (combined)             3865083 San Diego, CA MSA                                                    2655463 Minneapolis-St. Paul, MN-WI MSA                       2765116 Phoenix-Mesa, AZ MSA                                          2746703 Seattle-Bellevue-Everett2, WA PMSA                   2234707 Vancouver – Lower Mainland3   2215391 Denver, CO PMSA                                                1866978 Portland-Vancouver, OR-WA PMSA                  1758937 San Jose, CA PMSA                                             1599604 San Antonio, TX MSA                                                 1490111 Sacramento, CA PMSA                                         1482208   Table 5.  Sources of air quality data for Lower Mainland and selected comparison metropolitan areas Metropolitan Area Counties included Los Angeles Los Angeles, Ventura, San Bernadino, Orange, Riverside San Francisco – Oakland, CA PMSA (combined)           San Francisco, Contra Costa, Alameda, San Mateo, Marin San Diego, CA MSA                                                  San Diego Minneapolis-St. Paul, MN-WI MSA                     Ramsey, Hennepin Phoenix-Mesa, AZ MSA                                        Maricopa Seattle-Bellevue-Everett11, WA PMSA                 King, Pierce Vancouver – Lower Mainland12 26 GVRD/MoE stations: Vancouver Airport east to Hope Denver, CO PMSA                                              Denver Portland-Vancouver, OR-WA PMSA                Multnomah (OR), Clark (WA) San Jose, CA PMSA                                           Santa Clara San Antonio, TX MSA                                               Bexar Sacramento, CA PMSA                                       Sacramento        10 U.S. Census 7/1/96 population estimates http://www.census.gov/population.  Los Angeles estimate is for 1997  11 Air Quality data do not included Everett due to high number of industrial sources  12 1998 Lower Mainland population includes: Vancouver/Richmond, Simon Fraser/Burnaby, North Shore, Fraser Valley, South Fraser Valley Health Regions. Source: The British Columbia Vital Statistics Agency  15 (i) ANNUAL AVERAGES  Table 6. Inhalable Particulate Matter (PM10), Annual average Metropolitan Area Concentration (µg/m3) Phoenix                                        41.2 Los Angeles 35.3 San Diego                                                  30.5 Denver                                              26.2 Sacramento    23.8 San Jose                                           22.9 Minneapolis-St. Paul                     21.4 San Francisco – Oakland 21.0 San Antonio                                               20.3 Portland                19.1 Seattle                 18.6 Vancouver-Lower Mainland 14.0     Table 7.  Nitrogen Dioxide (NO2), Annual average Metropolitan Area Concentration (ppm) Denver                                              0.034 Phoenix                                        0.029 Los Angeles 0.026 San Jose                                           0.025 Minneapolis-St. Paul                     0.019 San Diego                                                  0.019 Vancouver-Lower Mainland 0.018 San Francisco – Oakland 0.016 Portland                0.014 San Antonio                                               0.013 Sacramento    0.013 Seattle                 0.013     Table 8.  Sulfur Dioxide (SO2), Annual average Metropolitan Area Concentration (ppm)  Seattle                 0.005 Denver                                              0.005 Vancouver-Lower Mainland 0.003 Phoenix                                        0.003 San Diego                                                  0.003 Minneapolis-St. Paul                     0.002 Los Angeles 0.002 San Francisco – Oakland 0.001 San Jose                                           NA Portland                NA San Antonio                                               NA  16 (ii) AVERAGES ACROSS STATIONS AND YEARS FOR EACH METRIC  Table 9.  Carbon Monoxide (CO), 2nd highest 1 hour maximum Metropolitan Area Concentration (ppm) Denver                                              10.5 San Antonio                                               8.9 Phoenix                                        8.3 Portland                8.3 Seattle                 8.1 Minneapolis-St. Paul                     7.7 Los Angeles 7.4 San Diego                                                  6.5 Sacramento                                       6.5 Vancouver-Lower Mainland 5.2 San Francisco – Oakland 4.9 San Jose                                           NA  Table 10.  Carbon Monoxide (CO), 2nd highest 8 hour average Metropolitan Area Concentration (ppm) Denver                                              5.8 Phoenix                                        5.4 Portland                5.3 Seattle                 5.1 Los Angeles 5.0 Sacramento                                       4.8 Minneapolis-St. Paul                     4.5 San Antonio                                               4.0 San Diego                                                  4.0 Vancouver-Lower Mainland 3.1 San Francisco – Oakland 2.9 San Jose                                           NA  Table 11.  Ozone (O3), 2nd highest 1 hour maximum Metropolitan Area Concentration (ppm) Los Angeles 0.139 Sacramento    0.117 San Antonio                                               0.109 Phoenix                                        0.108 San Diego                                                  0.107 San Jose                                           0.100 San Francisco – Oakland 0.097 Seattle                 0.093 Portland                0.092 Denver                                              0.090 Vancouver-Lower Mainland 0.072 Minneapolis-St. Paul                     NA         17    Table 12.  Ozone (O3), 3rd highest 1 hour maximum Metropolitan Area Concentration (ppm) Los Angeles 0.133 Sacramento    0.114 San Antonio                                               0.105 Phoenix                                        0.105 San Diego                                                  0.102 San Jose                                           0.095 San Francisco – Oakland 0.090 Seattle                 0.086 Denver                                              0.086 Portland                0.084 Vancouver-Lower Mainland 0.070 Minneapolis-St. Paul                     NA   Table 13.  Ozone (O3), 4th highest 1 hour maximum Metropolitan Area Concentration (ppm) Los Angeles 0.129 Sacramento    0.109 San Antonio                                               0.102 Phoenix                                        0.102 San Diego                                                  0.100 San Jose                                           0.090 San Francisco – Oakland 0.086 Seattle                 0.081 Portland                0.078 Denver                                              0.084 Vancouver-Lower Mainland 0.068 Minneapolis-St. Paul                     NA  Table 14.  Inhalable Particulate Matter (PM10), 2nd highest 24 hour average Metropolitan Area Concentration (µg/m3) Phoenix                                        87.3 Los Angeles 76.3 Sacramento    65.7 Denver                                              62.4 San Diego                                                  58.8 San Jose                                           51.9 Seattle                 49.7 Minneapolis-St. Paul                     48.3 San Francisco – Oakland 48.3 Portland                43.7 San Antonio                                               41.8 Vancouver-Lower Mainland 41.5       18 Table 15.  Inhalable Particulate Matter (PM10), 3rd highest 24 hour average Metropolitan Area Concentration (µg/m3) Phoenix                                        80.2 Los Angeles 68.8 Denver                                              57.6 Sacramento    56.9 San Diego                                                  53.4 San Jose                                           45.8 Seattle                 44.4 Minneapolis-St. Paul                     44.0 San Francisco – Oakland 42.8 Portland                39.9 San Antonio                                               38.5 Vancouver-Lower Mainland 38.3   Table 16.  Inhalable Particulate Matter (PM10), 4th highest 24 hour average Metropolitan Area Concentration (µg/m3) Phoenix                                        72.8 Los Angeles 63.6 Denver                                              53.5 Sacramento    50.8 San Diego                                                  50.3 Minneapolis-St. Paul                     41.9 Seattle                 41.8 San Jose                                           41.5 San Francisco – Oakland 39.2 Portland                36.8 Vancouver-Lower Mainland 35.8 San Antonio                                               34.5  19 (iii) MAXIMUM OF STATIONS AND YEARS FOR EACH METRIC.  Table 17.  Carbon Monoxide (CO), 2nd highest 1 hour maximum Metropolitan Area Concentration (ppm) Los Angeles 21.3 Minneapolis-St. Paul                     17.1 Denver                                              17.1 San Antonio                                               14.0 Phoenix                                        13.3 Portland                12.9 Vancouver-Lower Mainland 12.6 Seattle                 11.6 San Diego                                                  11.1 San Jose                                           10.5 Sacramento                                       10.0 San Francisco – Oakland 7.9    Table 18.  Carbon Monoxide (CO), 2nd highest 8 hour average Metropolitan Area Concentration (ppm) Los Angeles 15.3 Phoenix                                        9.6 Denver                                              9.5 Sacramento                                       8.0 Portland                7.8 Minneapolis-St. Paul                     7.6 Seattle                 7.5 San Jose                                           7.5 San Diego                                                  7.0 Vancouver-Lower Mainland 5.4 San Francisco – Oakland 5.1 San Antonio                                               5.0   Table 19.  Ozone (O3), 2nd highest 1 hour maximum Metropolitan Area Concentration (ppm) Los Angeles .241 Sacramento    .154 San Francisco – Oakland .149 San Diego                                                  .144 San Jose                                           .142 Seattle                 .135 Phoenix                                        .130 San Antonio                                               .126 Portland                .108 Denver                                              .107 Vancouver-Lower Mainland .106 Minneapolis-St. Paul                     N/A       20 Table 20.  Ozone (O3), 3rd highest 1 hour maximum Metropolitan Area Concentration (ppm) Los Angeles .228 Sacramento    .148 San Francisco – Oakland .142 San Diego                                                  .139 San Jose                                           .135 Phoenix                                        .129 Seattle                 .123 San Antonio                                               .119 Denver                                              .105 Portland                .102 Vancouver-Lower Mainland .098 Minneapolis-St. Paul                     N/A   Table 21.  Ozone (O3), 4th highest 1 hour maximum Metropolitan Area Concentration (ppm) Los Angeles .223 Sacramento    .148 San Francisco – Oakland .138 San Diego                                                  .137 San Jose                                           .128 Phoenix                                        .122 San Antonio                                               .119 Seattle                 .112 Denver                                              .103 Portland                .095 Vancouver-Lower Mainland .094 Minneapolis-St. Paul                     N/A   Table 22.  Inhalable Particulate Matter (PM10), 2nd highest 24 hour average Metropolitan Area Concentration (µg/m3) Phoenix                                        308 Los Angeles 236 Sacramento    156 Denver                                              104 San Diego                                                  121 Seattle                 93 Minneapolis-St. Paul                     91 San Jose                                           86 Vancouver-Lower Mainland     8213 San Francisco – Oakland 78 Portland                70 San Antonio                                               53       13 All three maximum values for Vancouver are due to high readings at the Chilliwack (T12) site during December 30, 1994 - January 7,1995. Excluding this site from the analysis gives maximum measurements of 65, 64 and 53 µg/m3 for the 2nd, 3rd and 4th highest 24-hour average measurements. 21     Table 23.  Inhalable Particulate Matter (PM10), 3rd highest 24 hour average Metropolitan Area Concentration (µg/m3) Phoenix                                        302 Los Angeles 187 Sacramento    128 San Diego                                                  119 Denver                                              99 Minneapolis-St. Paul                     77 Vancouver-Lower Mainland 76 Seattle                 74 San Jose                                           72 San Francisco – Oakland 65 Portland                62 San Antonio                                               50   Table 24.  Inhalable Particulate Matter (PM10), 4th highest 24 hour average Metropolitan Area Concentration (µg/m3) Phoenix                                        205 Los Angeles 177 San Diego                                                  104 Sacramento    100 Denver                                              88 Minneapolis-St. Paul                     74 Seattle                 73 San Jose                                           69 Vancouver-Lower Mainland 64 San Francisco – Oakland 62 Portland                59 San Antonio                                               48 22   IV.  Spatial variability and temporal patterns   Interpolation and GIS Methods Hourly ambient air pollutant data, measured at 21 sites within the Greater Vancouver Regional District and Fraser Valley between 1994-1998 were obtained.  The pollutants of interest included carbon monoxide (CO), nitrogen dioxide (NO2), ozone (O3), PM10 and sulfur dioxide (SO2).  From hourly data, we evaluated the daily 1-hour maximums for CO, NO2, O3 and SO2.  CO and O3 were also summarized into daily maximum 8-hour averages.  PM10 and SO2 data was analyzed using 24- hour averaged values.  To help evaluate the spatial pattern of pollutants, an interpolation program14,15 was used to generate estimates of pollutant concentrations at unmeasured locations within the region.  A rectangular grid, containing 324 square cells (cell length = 4947 meters), was constructed to cover the geographical region in which the 21 sampling stations were located.  The center coordinates of each cell were used as interpolation points for the statistical interpolation program.  The daily (1-hour) maximum, daily maximum 8-hour average, and 24-hour average data from each site for each pollutant were entered into the interpolation software to create daily pollutant coverage grids including each of the 324 cells. The interpolation is purely statistical, not physical or chemical, and is based upon computing the distance -weighted correlations between all monitoring sites, while at the same time accounting for temporal patterns within the data. Interpolations were conducted for actual pollutant concentrations (monthly averages) as well as for the # of days above 75th and 95th percentile values.  Output values from the interpolation software were imported into a geographical information system (GIS) to visualize the spatial pollutant distribution (ArcView GIS Version 3.2 and ArcView Spatial Analyst Version 1.1).  For ease of interpretation and presentation, monthly grids were averaged by year as well as by season (Winter = October – March; Summer = April – September).  The concentration data from each interpolation point were displayed in the cells of the grid using graduated colours.  The areas for which we were confident to report interpolated results were determined.  It was assumed that the ability of the interpolation program to accurately predict concentrations would decrease with distance from sampling stations.  Therefore, on the map, “circles of influence” were drawn around each of the 21 sites, which extended from each individual site to the next nearest site. Any cells lying outside of these circles were excluded from the pollutant-coverage maps.  Any cells with main coverage of the ocean were also excluded.  This provided a base coverage map.  From this base map, filters were created on a pollutant-specific basis to further exclude: a) regions covered by sites that did not sample that pollutant, and b) regions surrounding sites that could not be accurately predicted by the interpolator.  “Exclusion” in these cases meant excluding the cells falling inside the site’s circle of influence that were not included in any other site’s circle.    14 Brown PJ, Le ND, Zidek JV.  Multivariate spatial interpolation and exposure to air-pollutants. Canadian J Statistics. 1994; 22: (4) 489-509  15 Li KH, Le ND, Sun L, Zidek JV.  Spatial-temporal models for ambient hourly PM10 in Vancouver. Environmetrics 1999; 10: (3) 321-338. 23  For a), the number of sites covering each pollutant varied between pollutants and by year as indicated in Table 25.  Table 25.  Number of monitoring sites for each air pollutant, 1994-1998. # of sites: CO NO2 O3 PM10 SO2 1994 14 16 18 8 7 1995 14 16 18 10 7 1996 14 16 18 10 7 1997 15 17 18 11 7 1998 17 19 20 13 7  For b), a cross validation was performed in order to determine the ability of the interpolation program to predict the concentration for the cells containing each site (when not including the actual site in the interpolation).  Criteria were developed to exclude sites that could not be accurately interpolated.  Sites with correlations above 0.7 and where the absolute differences (|observed- predicted|)/mean observed concentration were less than 0.25 were excluded.  The cell in which the site was located remained included in all analyses since the difference between observed and predicted were good when the site was included in the interpolation. At this point the decision was made to exclude SO2 entirely from this analysis, as SO2 the interpolation software could not estimate concentrations within the defined criteria.  Pollutant and year-specific filters were created for maps to include only regions where sampling actually took place and for the cells that the interpolation program could reasonably predict.  These filters were applied to the yearly and seasonal grids.  For presentation purposes, standard legends for each pollutant were created to standardize the gradients representing concentration change.  The gradients were determined from quantiles of monthly values from each sampling site between 1994 and 1998.  Each change in color represents a 5 percentile difference between 0% to 100% (Table 26).  The actual concentrations from sampling sites were also presented on each map according to the same legends.  Table 26.  Minimum and maximum concentrations for each pollutant metric used to generate monthly average concentration maps. Percentile CO 8hr (ppm) CO max (ppm) NO2 max (ppm) O3 8hr (ppm) O3 max (ppm) PM10 24hr (µg/m3) 0 0.295 0.386 0.014 0.004 0.006 5.751 100 3.404 5.147 0.050 0.047 0.056 25.181   Figures are presented for each of the pollutants and for several different metrics. In each figure, different pollutants concentrations are displayed by different color intensities on a background map including census subdivisions, major roads (line thickness corresponds to traffic volumes) and geographical features.  Measured concentrations at individual monitoring sites can be visualized by the color within each circle denoting the location of an individual monitoring site.  Although maps are available16 and can be generated for any of the metrics for any monthly time period (or  16 Requests for individual maps or for a copy of the ArcView project (from which any map can be constructed) should be directed to Dr. Michael Brauer - brauer@interchange.ubc.ca   24 combination of months) during 1994-1998, only selected maps are displayed here. Specifically, only seasonal (summer and winter) and annual summary maps for the years 1995 and 1998 are presented unless otherwise specified.  For O3 and PM10 only, maps are also shown which present the number of days above specified concentrations, the 75th and 95th percentiles of all measurements made across all years and locations. From these figures it is possible to evaluate the spatial and temporal distribution of peak concentrations for each pollutant while the maps which indicate interpolated concentration values present an estimate of average spatial patterns. 25 26  27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47  48 49 Interpretation of maps  a) Carbon Monoxide  According to the results of the cross validation analyses, CO concentrations were predictable for a continuous region over the Vancouver and Richmond areas.  Only single grid cells could be shown further east.  Daily 1-hour maximum and maximum 8-hour averaged data showed similar patterns, although as expected concentrations were higher for the daily maximum metric.  In general, the CO concentration throughout the Vancouver and Richmond area was higher than the single grid points in the Fraser Valley.  Viewing the Vancouver/Richmond area more closely, between 1994 and 1997, concentrations were highest in the downtown core and towards the ocean.  Concentrations were slightly lower in North and South Burnaby and the North Shore.  Two inner cells seemed to be influenced by the Rocky Point Park and Eagle Ridge sites, as they were consistently higher than surrounding cells. Concentrations were very low for the rural valley sites with the exception of Abbotsford, where concentrations were consistently higher than surrounding areas (as high as the downtown core concentration).  This is likely due to the Abbotsford sampling site being in an urban location that is influenced by traffic whereas neighboring sites (Surrey East, Langley) were rural in comparison.  Concentrations in the Vancouver/Richmond area decreased slightly over time during 1994-1998. This observation was largely attributed to 1998, where concentrations were lower than earlier years. The decreasing trend was slightly more prominent for daily maximum CO data.  CO demonstrated high concentrations in the winter and low concentrations in the summer, without alterations to the spatial patterns described above.  The summer pattern did not change much over time with the grid cell encompassing the Robson Square (downtown Vancouver) site usually showing the highest concentration.  Winter data also indicated high concentrations over the entire Vancouver area.   b) Nitrogen Dioxide  The daily 1-hour maximum value of NO2 was assessed.  The region remaining predictable after the cross validation was larger than for CO, with the Hope site being the only site completely separated from the continuous region.  The highest concentrations were found in the downtown core and decreased towards the south and east (towards Richmond and Burnaby).  This pattern was different from that of CO, which was high over the entire Vancouver/Richmond area.  Concentrations over Surrey and extending east towards Hope were very low.  Abbotsford again displayed slightly higher values than surrounding areas.  This spatial pattern was well pronounced for 1995-1997 yearly summaries.  There were no distinguishable temporal or seasonal trends.  Spatially, however, the high concentrations in the downtown core during winters 95/96 and 96/97 were spread out to Richmond.  Such spread was not seen for the corresponding summer seasons.  Nevertheless, the summer–winter differences were slight when compared to other pollutants.    50 c) Ozone  Ozone was measured at 18 of the monitoring sites and the interpolation program was able to predict the spatial distribution of pollutant quite well.  Thus, predicted ozone concentrations had the greatest continuous coverage within the Lower Mainland.  Daily 1-hour maximum and daily maximum 8-hour averaged data showed similar spatial and temporal patterns.  Spatial patterns varied between each year.  In general, concentrations in the downtown core were low and increased within the east of the Fraser Valley.  The 1998 summer map demonstrated this pattern well.  A band of increased values in the middle of the grid, covering the area northeast of the Langley site was also demonstrated in most maps.  There were no prominent trends over time, however, 1998 values were slightly higher than in previous years.  Seasonally, ozone concentrations showed summer highs and winter lows with the same spatial patterns as described above.  Pollutant distributions varied within the Fraser Valley, but a band of increased concentrations was observed between Langley and Hope in the northeast direction.  This band was observable for most seasons between 1994 and 1998.  Maps indicating the number of days above the 75th or 95th percentiles indicated similar patterns, indicating that the longer-term spatial average concentrations also reflect the spatial patterns in which peak concentrations are distributed.   d) PM10  Daily 24-hour average data was analyzed.  There was little spatial variability compared to the other pollutants with no prominent patterns or differences between downtown and valley. Due to a lack of sites sampling PM10 in 1994, there was not a large area to display concentrations for this period. For most annual and seasonal maps, slightly higher concentrations were observed in a horizontal band along the southern part of the grid - this is likely due to the influence of the Abbotsford and South Richmond monitoring sites which are known to experience short-term high peak concentrations during summer periods of stagnation due to their proximity to major traffic routes17. The Abbotsford site typically has the highest PM10 concentrations in the region.  Overall, the highest concentrations occurred during 1995 and the lowest were seen in 1997. Otherwise, the yearly summaries did not show any prominent temporal patterns for PM10.  As was the case for ozone, there were clear summer high and winter low concentrations.  1995 and 1998 summers were especially high in concentration and winter 1997/98 was especially low in concentration over entire region.  Maps indicating the number of days above the 75th or 95th percentiles indicated similar patterns to the maps of annual or seasonal-average concentrations. However, for the plots indicating number of days above certain values, the peak concentrations measured at the Abbotsford location are more easily observed, as is the general pattern for slightly more of the highest concentration measurements to be experienced in the Fraser Valley relative to the more western areas of the region.  17 McKendry I.  PM 10 Levels in the Lower Fraser Valley, British Columbia,Canada: An Overview of Spatiotemporal Variations and Meteorological Controls.  J. Air & Waste Manage. Assoc. 2000; 50:443-452  51  Summary of spatial variability and temporal patterns  Spatially, the main differences in pollutant concentrations were distinguished between the Downtown Vancouver/Richmond areas compared to the eastern areas of the Fraser Valley, such as Langley, Abbotsford, Chilliwack and Hope.  CO and NO2 concentrations were higher in the downtown area than in the valley.  The pattern of CO however was difficult to distinguish since this pollutant was not interpolated well, probably due to the localised nature of CO emissions.  Ozone, a regional pollutant, displayed the opposite trend with higher concentrations in the Fraser Valley relative to Vancouver/Richmond/Burnaby.  PM10 showed little spatial variability.  No prominent increasing or decreasing temporal trends were observed, although the 5-year time frame for analysis limited the sensitivity to observe temporal trends.  CO showed a slight decrease in 1998, although it was difficult to distinguish whether this was a real trend over time.  Other pollutants (NO2 and PM10) showed slightly higher concentrations in 1998.  Seasonal patterns were consistent with expectations as ozone and PM10 concentrations were high in the summer and low in the winter as opposed to CO, which demonstrated higher concentrations in the winter than in the summer.  NO2 did not demonstrate prominent seasonal differences.  Thus, according to these results, CO concentrations are higher in the Vancouver/Richmond area during the winter and ozone concentrations peak in the Fraser Valley during the summer.  These trends have not changed over time since 1994.     Acknowledgments Air quality data were kindly provided by the Greater Vancouver Regional District Air Quality Department.  Jochen Brumm’s work on this project was supported in part by a grant from the U.S. Environmental Protection Agency (subcontract to the University of British Columbia) to the University of Washington, National Research Center for Statistics and the Environment, entitled “Statistical methods for particulate matter air pollution research.”  M. Brauer also acknowledges the support of the Medical Research Council of Canada and the British Columbia Lung Association (Scientist Award), and the American Lung Association (Career Investigator Award). 52 Appendix 1.  Technical details for air pollution attributable death estimation.  Model (1) adjusts for meteorology and cycles. More precisely, the model included yearly, half-yearly, 3 months and 4 months cycles (both sine and cosine waves for each frequency). We denote the vector of the values of these cycles on day t by c1(t),…,c8(t). The adjustment for meteorology was done through joint loess-smoothing of the temperature and relative humidity data with a span of 0.1 (as implemented in Splus 3.4 in the function for generalized additive models gam); denote the value of this function on day t by f(temp(t),rh(t)) where temp(t) denotes the temperature  and rh(t) is the relative humidity on day t. If we denote the vector of regression coefficients for the pollutants from this model over the time-period 1994-1996 by β and the corresponding levels of pollution on day t by x(t), then the model becomes  log d(t) ~  c1(t) +…+c8(t) + f(temp(t),rh(t)) + p(t)'β   We considered lags 0-2 for each pollutant separately as outlined earlier, hence the vector of pollutant levels consists of different pollutants at different lags (Table 1).  We denote the estimated regression coefficients for the model fit to the data from 1994-1996 by β^ where we have a different regression coefficient and different lags for the pollutants for each of the outcomes (respiratory, circulatory and so on).  Now we derive the formula for the estimated number of deaths for a specified outcome d (respiratory, for example). Let plow denote the vector of cut-off values for low pollution, pobs(t) the vector of observed pollution levels on day t (corresponding to the appropriate lags). We use the difference ∆p(t) =pobs(t) - plow to estimate the number of people dying due to air-pollution. More precisely, the sum of deaths S, over 5 years (1826 days, t represents days in the formula) attributed to air-pollution is calculated as   where dobs(t) is the actual observed number of deaths (in the according stratum), d^low(t) is the predicted number of deaths at low pollution and d^obs(t) is the predicted number of deaths at the actually observed pollution levels. To get the estimated number of deaths per year, we divide this expression by the number of years (5).  The variance of this expression is calculated using the Delta-method. The expression for the approximate variance is   = ∧ ∧           −−= 1826 1 1 )( )()( t obs low obs td tdtdS  = ∧∧         −++−= 1826 1 2211 1...))(p)(pexp()( t obs tttdS ββ 53                 ∆∆         ∆∆≈  ∧∧∧ t obs t obs tptptdtptptd )())(exp()(var()())(exp()( ' βββ

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
United States 10 1
China 7 1
Japan 2 0
Canada 1 0
City Views Downloads
Beijing 5 0
Unknown 4 0
San Francisco 3 0
Shenzhen 2 1
Ft. Pierce 2 0
Tokyo 2 0
Saskatoon 1 0
Ashburn 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}

Share

Share to:

Comment

Related Items