OP III – 2 Air pollution exposure assessment for the elapse project using hybrid lur models. (18th March 2018)

• Record Type: Journal Article
• Authors: Hoogh, Kees de; Strak, Maciej; Chen, Jie; Gulliver, John; Hertel, Ole; Brunekreef, Bert; Hoek, Gerard
• Is Part Of: Occupational and environmental medicine. Volume 75(2018)Supplement 1
• Page Start: A5
• Page End: A5
• Publication Date: 2018-03-18
• Subjects: Medicine, Industrial -- Periodicals; Environmental health -- Periodicals; 616.980305
• DOI: 10.1136/oemed-2018-ISEEabstracts.12
• Languages: English
• ISSNs: 1351-0711

Abstract

Abstract : Background/aim: In order to investigate associations between air pollution and adverse health effects fine spatial air pollution surfaces are needed to provide cohorts with exposures. In the ELAPSE project we developed hybrid land use regression models for multiple pollutants and linked these to 11 individual and 7 administrative cohorts in 10 countries for a total of 35 million participants. Methods: Europe-wide hybrid LUR models were developed for 2010 estimating annual mean PM2.5, NO2, O3 and BC (including cold and warm season O3 estimates). Models were developed based on AirBase routine monitoring data (PM2.5, NO2, O3 ) and ESCAPE monitoring data (BC), incorporating land use and traffic data plus satellite observations and dispersion model estimates as additional predictor variables. Universal kriging was performed on residual spatial variation. Main model was developed using all sites. To evaluate robustness, five more models were developed, each built on 80% of monitoring sites with remaining 20% used for validation. Models were applied to 100 × 100 m grids across Europe to allow for exposure assignment for all ELAPSE cohorts. Results: Main models explained: NO2 58%, PM2.5 71% (59% LUR +12% kriging), O3 62% and BC 51% of spatial variation in measured concentrations. Validation R2 ranged 0.55–0.60 for NO2, 0.63–0.77 for PM2.5, 0.51–0.69 for O3 and 0.43–0.57 for BC. Dispersion model estimates, road density, nature and residential area were predictor variables in NO2 model. PM2.5 model consisted of satellite derived and dispersion model estimates, altitude, road density, nature, ports and residential area. Satellite derived and dispersion model estimates, road density, residential area, urban green and Y-coordinate were predictors in BC model. O3 model included dispersion model estimates, road density, ports, residential area and altitude. Kriging proved an efficient technique to explain part of residual spatial variation. Conclusion: We were able to develop robust NO2, PM2.5, O3 and BC hybrid LUR models to provide exposure estimates for all cohort participants in the ELAPSE project.