Academic journal article Environmental Health Perspectives

Comparing the Health Effects of Ambient Particulate Matter Estimated Using Ground-Based versus Remote Sensing Exposure Estimates

Academic journal article Environmental Health Perspectives

Comparing the Health Effects of Ambient Particulate Matter Estimated Using Ground-Based versus Remote Sensing Exposure Estimates

Article excerpt

Introduction

Remote sensing (RS) and atmospheric chemistry models play an increasingly important role in exposure assessment for epidemiological and burden-of-disease studies. A wide array of products produced by several U.S. federal agencies, such as the National Aeronautics and Space Administration (NASA) and the U.S. Environmental Protection Agency (EPA), are now available. Sometimes, these models form the basis for more complex estimates combining ground-based data or several remote-sensing products.

Several recent epidemiological investigations have used remote sensing for the exposure assessment or as input into other health impact assessment or variable- imputation models. By combining retrievals of aerosol optical depth (AOD) from the Moderate Resolution Imaging SpectroRadiometer (MODIS) and Multiangle Imaging SpectroRadiometer (MISR) instruments onboard the Terra satellite with the GEOS-Chem model, van Donkelaar et al. (2010) developed 6-year mean global estimates of [PM.sub.2.5] at ~10 km resolution (van Donkelaar et al. 2010). These RS products were designed to avoid reliance on [PM.sub.2.5] monitors because these RS products can offer information about [PM.sub.2.5] in regions where [PM.sub.2.5] monitors are not generally available or where there are concerns about [PM.sub.2.5] data quality, as, for example, with Tapered Element Oscillating Microbalances (TEOMs). Researchers in Canada have used the van Donkelaar et al. (2010) estimates to assess the health effects of air pollution. Specifically, these [PM.sub.2.5] estimates were significantly associated with incidence of diabetes (Chen et al. 2013) and diabetes mortality (Brook et al. 2013) and cardiovascular mortality (Crouse et al. 2012, 2015). These RS estimates have also been used to estimate the global mortality associated with [PM.sub.2.5] (Evans et al. 2013; Lim et al. 2012).

A few studies have attempted to systematically compare the exposure estimates from ground-based versus RS models. Lee et al. (2012) developed national-level models using data from more than 1,300 ground monitors for [PM.sub.2.5] (Lee et al. 2012). Their results indicated that within ~98 km of a monitor, the ground-based estimates predicted [PM.sub.2.5] concentrations more accurately than the RS estimates discussed above (van Donkelaar et al. 2010). Beyond 98 km, however, the RS estimates were better predictors of ground-level [PM.sub.2.5]. For the most part, the estimates were highly correlated with each other, and the authors concluded that the differences in prediction capacity were fairly small. Another study compared NASA AOD retrievals to ground-based estimates derived from a generalized linear model that included ground information on land-use predictors and several statistical smoothing functions. The study concluded that the RS estimates were not generally better predictors than the ground-based models, and after applying smoothing functions in the models, there was little marginal benefit to the RS information on predicting ground-level [PM.sub.2.5] (Paciorek and Liu 2009). Subsequent studies have found that ground-based observations can be better predicted using exposure models with RS estimates than without (Beckerman et al. 2013a, 2013b; Kloog et al. 2012b; Ma et al. 2014; Vienneau et al. 2013).

RS estimates of air pollution generally lack the fine-scale resolution (< 1 km) needed for use in environmental epidemiological studies that aim to understand small-area variations in exposure. To achieve horizontal downscaling of the RS estimates, hybrid approaches that combine variants of land-use regression models, which predict pollutant concentrations from land use such as road length, traffic density, or open space with RS measurements are being employed (Beckerman et al. 2013a, 2013b; Kloog et al. 2012b; Ma et al. 2014; Vienneau et al. 2013). Through statistical modeling, proxy information about likely locations of pollution at smaller spatial resolution than AOD pixels can essentially distribute the [PM. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.