A national fine spatial scale land-use regression model for ozone

- Few land use regression (LUR) models of fine scale spatial variation of ozone exist.
- We tested the feasibility of developing a national fine-scale LUR model for ozone.
- Ozone correlated highly negatively with nitrogen dioxide and fine particles.
- A LUR model explained 71% of ozone spatial variation.
- The high correlation with NO2 limits application in epidemiology.

Uncertainty about health effects of long-term ozone exposure remains. Land use regression (LUR) models have been used successfully for modeling fine scale spatial variation of primary pollutants but very limited for ozone.Our objective was to assess the feasibility of developing a national LUR model for ozone at a fine spatial scale. Ozone concentrations were measured with passive samplers at 90 locations across the Netherlands (19 regional background, 36 urban background, 35 traffic). All sites were measured simultaneously during four 2-weekly campaigns spread over the seasons. LUR models were developed for the summer average as the primary exposure and annual average using predictor variables obtained with Geographic Information Systems.Summer average ozone concentrations varied between 32 and 61 µg/m3. Ozone concentrations at traffic sites were on average 9 µg/m3 lower compared to regional background sites. Ozone correlated highly negatively with nitrogen dioxide and moderately with fine particles. A LUR model including small-scale traffic, large-scale address density, urban green and a region indicator explained 71% of the spatial variation in summer average ozone concentrations.Land use regression modeling is a promising method to assess ozone spatial variation, but the high correlation with NO2 limits application in epidemiology.