Long-term ultrafine particle (UFP) exposure estimates at a fine spatial scale are needed for epidemiological studies. Land use regression (LUR) models were developed and evaluated for six European areas based on repeated 30 min monitoring following standardized protocols. In each area; Basel (Switzerland), Heraklion (Greece), Amsterdam, Maastricht, and Utrecht (“The Netherlands”), Norwich (United Kingdom), Sabadell (Spain), and Turin (Italy), 160–240 sites were monitored to develop LUR models by supervised stepwise selection of GIS predictors. For each area and all areas combined, 10 models were developed in stratified random selections of 90% of sites. UFP prediction robustness was evaluated with the intraclass correlation coefficient (ICC) at 31–50 external sites per area. Models from Basel and The Netherlands were validated against repeated 24 h outdoor measurements. Structure and model R2 of local models were similar within, but varied between areas (e.g., 38–43% Turin; 25–31% Sabadell). Robustness of predictions within areas was high (ICC 0.73–0.98). External validation R2 was 53% in Basel and 50% in The Netherlands. Combined area models were robust (ICC 0.93–1.00) and explained UFP variation almost equally well as local models. In conclusion, robust UFP LUR models could be developed on short-term monitoring, explaining around 50% of spatial variance in longer-term measurements.

BackgroundLand Use Regression (LUR) is a popular method to explain and predict spatial contrasts in air pollution concentrations, but LUR models for ultrafine particles, such as particle number concentration (PNC) are especially scarce. Moreover, no models have been previously presented for the lung deposited surface area (LDSA) of ultrafine particles. The additional value of ultrafine particle metrics has not been well investigated due to lack of exposure measurements and models.MethodsAir pollution measurements were performed in 2011 and 2012 in the eight areas of the Swiss SAPALDIA study at up to 40 sites per area for NO2 and at 20 sites in four areas for markers of particulate air pollution. We developed multi-area LUR models for biannual average concentrations of PM2.5, PM2.5 absorbance, PM10, PMcoarse, PNC and LDSA, as well as alpine, non-alpine and study area specific models for NO2, using predictor variables which were available at a national level. Models were validated using leave-one-out cross-validation, as well as independent external validation with routine monitoring data.ResultsModel explained variance (R2) was moderate for the various PM mass fractions PM2.5 (0.57), PM10 (0.63) and PMcoarse (0.45), and was high for PM2.5 absorbance (0.81), PNC (0.87) and LDSA (0.91). Study-area specific LUR models for NO2 (R2 range 0.52–0.89) outperformed combined-area alpine (R2 = 0.53) and non-alpine (R2 = 0.65) models in terms of both cross-validation and independent external validation, and were better able to account for between-area variability. Predictor variables related to traffic and national dispersion model estimates were important predictors.ConclusionsLUR models for all pollutants captured spatial variability of long-term average concentrations, performed adequately in validation, and could be successfully applied to the SAPALDIA cohort. Dispersion model predictions or area indicators served well to capture the between area variance. For NO2, applying study-area specific models was preferable over applying combined-area alpine/non-alpine models. Correlations between pollutants were higher in the model predictions than in the measurements, so it will remain challenging to disentangle their health effects.Electronic supplementary materialThe online version of this article (doi:10.1186/s12940-016-0137-9) contains supplementary material, which is available to authorized users.