Local ozone prediction with hybrid model

Tropospheric ozone in high concentrations can cause health problems. A reliable alerting system is needed. In this paper we present the hybrid model that can be used for ozone forecasting in urban microlocations. The hybrid model is combined from meteorological and air-quality models (covering large geographical 3-dimensional space), and empirical model (offering good local forecasts), implemented as a Gaussian-process model. Prediction model for the city of Koper in Slovenia that has Mediterranean climate and problems with the ozone pollution is presented and used for improved one-day-ahead forecasting of the maximum hourly value within each day. The model validation results show that hybrid model improves ozone forecasts and provides better alert systems for the selected location.