GIS Approaches for the Estimation of Residential-Level Ambient PM Concentrations Duanping Liao,1 Donna J. Peuquet,2 Yinkang Duan,1 Eric A. Whitsel,3,4 Jianwei Dou,2 Richard L. Smith,5 Hung-Mo Lin,1 Jiu-Chiuan Chen,3 and Gerardo Heiss3 1Department of Health Evaluation Sciences, Pennsylvania State University College of Medicine, Hershey, Pennsylvania, USA; 2Department of Geography, Pennsylvania State University, College Park, Pennsylvania, USA; 3Department of Epidemiology, 4Department of Medicine, and 5Department of Statistics, University of North Carolina–Chapel Hill, Chapel Hill, North Carolina, USA
Abstract Spatial estimations are increasingly used to estimate geocoded ambient particulate matter (PM) concentrations in epidemiologic studies because measures of daily PM concentrations are unavailable in most U.S. locations. This study was conducted to a) assess the feasibility of large-scale kriging estimations of daily residential-level ambient PM concentrations, b) perform and compare cross-validations of different kriging models, c) contrast three popular kriging approaches, and d) calculate SE of the kriging estimations. We used PM data for PM with aerodynamic diameter ≤ 10 µm (PM10) and aerodynamic diameter ≤ 2.5 µm (PM2.5) from the U.S. Environmental Protection Agency for the year 2000. Kriging estimations were performed at 94,135 geocoded addresses of Women's Health Initiative study participants using the ArcView geographic information system. We developed a semiautomated program to enable large-scale daily kriging estimation and assessed validity of semivariogram models using prediction error (PE) , standardized prediction error (SPE) , root mean square standardized (RMSS) , and SE of the estimated PM. National- and regional-scale kriging performed satisfactorily, with the former slightly better. The average PE, SPE, and RMSS of daily PM10 semivariograms using regular ordinary kriging with a spherical model were 0.0629, –0.0011, and 1.255 µg/m3, respectively ; the average SE of the estimated residential-level PM10 was 27.36 µg/m3. The values for PM2.5 were 0.049, 0.0085, 1.389, and 4.13 µg/m3, respectively. Lognormal ordinary kriging yielded a smaller average SE and effectively eliminated out-of-range predicted values compared to regular ordinary kriging. Semiautomated daily kriging estimations and semivariogram cross-validations are feasible on a national scale. Lognormal ordinary kriging with a spherical model is valid for estimating daily ambient PM at geocoded residential addresses. Key words: cross-validation, geographic information systems, kriging, particulate air pollution, population-based studies. Environ Health Perspect 114:1374–1380 (2006) . doi:10.1289/ehp.9169 available via http://dx.doi.org/ [Online 8 June 2006] Address correspondence to D. Liao, Department of Health Evaluation Sciences, Pennsylvania State University College of Medicine, 600 Centerview Dr., A210, Hershey, PA 17033 USA. Telephone (717) 531-4149. Fax: (717) 531-5779. E-mail: dliao@psu.edu We acknowledge the contributions of WHI investigators and institutions (Appendix) . The National Institute of Environmental Health Sciences funded this ancillary study (5-R01-ES012238) . The National Heart, Lung, and Blood Institute, U.S. Department of Health and Human Services, funded the Women's Health Initiative (WHI) program. The authors declare they have no competing financial interests. Received 15 March 2006 ; accepted 8 June 2006. The full version of this article is available for free in HTML or PDF formats. |