Background and aims: Spatio-temporal methods have been developed for the estimation of concentrations of pollutants such as particulate matter (PM) and nitrogen dioxide (NO2) for application in epidemiological studies. A very limited number of city-specific spatio-temporal ozone (O3) models have been proposed recently. Our aim was to develop a spatio-temporal land use regression (LUR) model that estimates daily concentrations of O3, for the whole year, warm (April 1st to 30th September) and cold season (October 1st and 31st March), within the greater Athens area. Methods: We developed models using a semiparametric approach including linear and smooth functions of spatial and temporal covariates and a bivariate smooth thin plate function. The final set of explanatory variables was selected based on the adjusted-R2. We tested the final model in temporal and spatial terms following a leave-one out monitor approach.Results: The adjusted-R2 of the developed annual model was 0.76, while for the warm and cold season it was 0.70 and 0.71, respectively. The spatial terms in our annual model explained 32.9% and the temporal 63.2% of the variability in O3. There was no remaining temporal or spatial autocorrelation in the residuals. The adjusted-R2 in the leave-one-out cross validation was 0.73 for the annual model (warm: 0.65 and cold: 0.70). The developed models showed good validity when comparing predicted and observed measurements for the 2015 data.Conclusions: Spatio-temporal LUR modeling provides a useful tool for estimating O3 spatio-temporal variability with adequate accuracy for subsequent use in epidemiological studies.