Abstracts: ISEE 21st Annual Conference, Dublin, Ireland, August 25–29, 2009: Poster Presentations
Background and Objective:
In time-series studies of ambient air pollution and health in large urban areas, measurement errors associated with instrument precision and spatial heterogeneity can affect risk estimates. Here, these errors are characterized and modeled.
Daily measures of 12 air pollutants were obtained from three ambient monitoring networks operating in Atlanta, USA during 1999–2004. Instrument precision was characterized using observations from collocated monitors. Spatial heterogeneity was assessed using data from multiple monitors in the study area to estimate population-weighted semivariograms that represent ratios of spatial variance to temporal variance. Monte Carlo simulations of both error types were generated for each pollutant, modeled as a function of the measured concentrations and including temporal autocorrelation.
Instrument error was small; relative to the total temporal variability, variance due to error ranged from 1% to 2% for continuously monitored pollutant gases (NO2, NOx, CO, SO2, O3) and 2% to 9% for particulate matter measures requiring laboratory analyses (PM10, PM2.5, and PM2.5 components sulfate, nitrate, ammonium, elemental carbon, organic carbon). In contrast, error due to spatial heterogeneity ranged from 4% to 12% for secondary air pollutants (O3, nitrate, sulfate, ammonium) and 28% to 59% for primary pollutants (NO2, NOx, SO2, CO, elemental carbon). Pollutants with substantial contributions from both primary and secondary sources (PM10, PM2.5, organic carbon) had intermediate levels of error. Time-series simulations with modeled error added to central monitor data were generated maintaining similar lognormal distribution features. For instrument error only, correlations between simulations closely matched the observed correlation between collocated monitor data. For spatial error, the semivariograms between simulations closely matched the population-weighted semivariogram.
Measurement error associated with instrument precision and spatial heterogeneity can be modeled and simulations can be produced for the assessment of the impact of ambient pollutant measurement error on health risk estimates.