An ensemble air-quality forecast over western Europe during an ozone episode

Multi-model ensemble averaging is tested as a technique for improving the accuracy of photochemical air-quality modeling. The performance of an ensemble-mean of four different photochemical models is tested against the individual models and the observations for a 6-day episode (31 July–5 August 1990), when high ozone levels were observed (up to 140 pbbv) over five monitoring stations in northwestern and central Europe. These models are the European Monitoring and Evaluation Programme (EMEP) Model, the European Air Pollution Dispersion (EURAD) Model, the Long-Term Ozone Simulation (LOTOS) Model and the Regional Eulerian Model with three different chemistry schemes (REM3), and each model has different initial and boundary conditions. The ensemble approach shows promising results, giving an overall better performance than any single model, including good performance for ozone peak-value prediction.