Evaluation of the CALIOPE air quality forecasting system for epidemiological research: the example of NO2 in the province of Girona (Spain)

Abstract

Background Air quality models are being increasingly used to estimate long-term individual exposures to air pollution in epidemiological studies. Most of them have been evaluated against measurements from a limited number of monitoring stations, which may not properly reflect the exposure characteristics of the study population. Methods We evaluated the performance of the high-resolution CALIOPE air quality forecasting system over a large sample of passive measurements of NO2 conducted at 635 home outdoor locations of the Girona province (Spain) during several 4-week sampling campaigns over one year (July 2007–June 2008). Sampling sites were superposed over the 4 km × 4 km CALIOPE grid, and average NO2 modeled concentrations were derived for all measurements conducted during the same sampling campaign at all the sampling sites located within the same grid cell. In addition, the ratio between measured and modeled concentrations for the whole study period at one fixed monitoring station was used to post-process the modeled values at the home outdoor locations. Results The correlation between measured and modeled concentrations for the entire study area (which includes urban settings, middle-size towns, and rural areas) was 0.78. Modeled concentrations were underestimated in the whole study area. After correcting the modeled concentrations by the measured to modeled ratio at the fixed station (r = 0.25), they were very similar to the measured concentrations (27.7 μg m−3 and 29.3 μg m−3, respectively). However, the performance of the modeling system depends on the type of subarea and is affected by the sub-grid emission sources. Conclusions The evaluation over the heterogenous Girona province showed that CALIOPE is able to reproduce the spatial variability of 4-week NO2 concentrations at the small regional level. CALIOPE output data is a valuable tool to complement study-specific air pollution measurements by incorporating regional spatial variability as well as short- and long-term temporal variability of background pollution in epidemiological research.