| Abstrakt: |
The standard approach to government-mandated aerometric monitoring of airborne particulates across North America is to sample every sixth day year round. However, such data are inadequate for epidemiological studies which aim to examine daily time series relationships of particulate air pollution to respiratory health responses. The aim of the present study was to estimate missing daily particulate matter ≤ 2.5 and ≤ 10 μm in aerometric diameter (PM2.5, and PM10) and sulfate (SO42−) to a degree sufficiently accurate and reliable to allow the use of these estimates, along with the measured data, in an investigation of the relationship of air pollution to respiratory hospital admissions in Montreal during the 1980s. Prediction equations were developed for May through October periods using available daily levels of predictor variables which included: relative humidity-corrected light extinction coefficient (bext) derived from airport visual range sightings, coefficient of haze (COH), SO2, NOx, CO, O3, wind speed, wind direction, barometric pressure (BP), temperature, relative humidity, and total precipitation. Three fourths of the available gravimetric particulate data were used to develop prediction models, while the remaining fourth was used to test the reliability of the model (holdout data). All final models explained over 70% of the variability in the particulate air pollutants and were reliable when tested against the holdout data. The strongest (P< 0.001) and most consistent predictors were bext, COH, and O3measured on the same day as the particulate, and BP lagged 1 day in the past. Other selected variables were same day NOx, BP, and minimum temperature. Although the present approach to the estimation of missing particulate air pollution may increase the level of exposure misclassification, it does allow for the use of existing network databases in epidemiological studies of daily air pollution health effects even though particulate data is only measured on certain days. |
