Background:Despite currently inconclusive results there is evidence accumulating that ultrafine particles (UFP), a major portion of the particulate matter of air pollution, may play a role in the development of lung cancer. Ultrafine particles (UFPs), the finest particulate matter in air, vary with space and time which is a challenge in estimating levels of exposure for individuals in health studies. Therefore, we aim to evaluate robustness of two statistical regression models in estimating the levels of ultrafine particles in atmospheric air for Greater Montreal area.Methods:A fixed site monitoring campaign was designed to measure the levels of ultrafine particles with a dense monitoring network of 250 sampling sites in the Greater Montreal Area. Ultrafine particles were measured for 20 minutes at each sampling site. In order to derive average annual levels, three repetitions were performed in winter and three in summer. However, high spatio-temporal variation of UFP makes it difficult to fit a regression model with high predictive value.In this analysis, we attempt to derive regression model for UFP within a Bayesian framework by considering the unobserved latent effect and euclidean distance between the sampling points. The proposed model attempts to capture the dependency of ultrafine particles between the sampling points that are close to each other.Results:The efficiency of conventional land use regression models in predicting UFP levels will be compared with spatiotemporal modelling in Bayesian framework with non-stationary covariance. Regression models will be developed using 90% of the sampling points. The remaining 10% will be used to evaluate the robustness of the models in estimating the UFP levels.Conclusions/Next steps:The best model for estimating the UFP levels will be used to derive an exposure surface for the Montreal area. This will be used to evaluate risk within a case-control study of lung cancer conducted in Montreal.