We developed an exposure estimation model for an epidemiological study on the effect of traffic-related air pollutants on respiratory diseases. The model estimates annual average outdoor concentration of nitrogen oxides (NOx) and elemental carbon (EC). The model is composed of three nested plume dispersion type submodels treating different spatial scales from a few meters to tens of kilometers. The emissions from road traffic was estimated at high spatial resolution along the paths of roads taking into account the effects of individual building shape and traffic signals to secure accuracy near trunk roads where most of the subjects of the epidemiological study resided. Model performance was confirmed by field measurements at permanent local government stations and purpose-built temporary stations; the latter supplemented roadside monitoring points and provided EC concentrations, which are not measured routinely. We infer that EC emissions were underestimated by using the available database because there were significant contributions to EC concentrations from sources that did not emit much NOx. An adjustment concentration yielded good agreement between model estimates and field measurements.