A ground-level ozone concentration prediction procedure is tested under real-world circumstances. Models are constructed for measured meteorological variables to predict ozone concentrations. Subsequently, a weather forecast is used as input to the prediction in order to obtain an estimate of the next day's maximum ozone concentration. The work investigates whether the additional variance, introduced by the use of weather forecasts, still allows accurate ozone concentration predictions. Two simulation studies were performed in order to check the predictive ability and the effect of the weather forecast accuracy. Good prediction accuracy was obtained. Only 2.6% of the predictions lead to residuals exceeding 40 μg/m 3 . In addition, the influence of the uncertainty in the weather forecasts is generally low in these cases (3.39 to 3.75 μg/m 3 ). Real predictions in the region of Grenland, Norway, for the summer of 1996 showed good agreement between measured and calculated maximum ozone levels. Only 3.5% of the predictions yielded residuals exceeding 40 μg/m 3 .