Real time modelling as an emergency decision support system for accidental release of air pollutants

In the framework of the project ‘Windbank’, wind field patterns in an area of 30km×30km in the Swiss Plateau between the Alps and the Jura were measured with 22 temporary meteorological stations and two SODARs during 4 months in 1997. Hourly averages from this high resolution network were combined with meteorological information from routine stations and from a weather prediction model. This data-set comprises all available parameters influencing the complex wind flow in the investigated area between the Alps and the Jura. A cluster analysis for this data-set lead to 12 classes with a high separation quality. It is demonstrated, that an on-line acquisition of meteorological data from routine stations and from a weather prediction model can be used to diagnose the recent wind field class with a probability of 96% to hit the correct wind field class. This diagnosis reveals wind fields with a very high spatial resolution in a very short time. Consequently, it is useful as a contribution to a decision support system for safety management after accidental releases of nuclear or chemical air pollutants. Further, a method is outlined to use the weather prediction model to forecast the wind field class. An average probability of 79% to hit the correct wind field classes for a forecast time of up to 24h is evaluated.