SMOGSTOP: a model for forecasting maximum daily ozone concentration in Belgium

Abstract

Every summer, ground level ozone concentrations rise in Belgium and cause episodes of photochemical summer smog. This phenomenon is the cause of well recognised public health distress, especially for people suffering from respiratory diseases. To warn groups of sensitive people against forthcoming smog episodes, VITO (the Flemish Institute for Technological Research) and VMM (the Flemish Environmental Agency) have joined forces to create an ozone pollution forecasting model, called SMOGSTOP (Statistical Model Of Groundlevel Short Term Ozone Pollution). SMOGSTOP is used by Belgian government agencies to issue ozone reports and warnings in the media (TV, radio, internet, etc.). Ozone pollution levels, as well as concentrations of other air pollutants, are monitored in Belgium by the telemetric air quality measuring networks of the three Belgian regions: Flanders, Wallonia and Brussels. The major meteorological variables, such as the windvector, temperature, pressure, humidity and precipitation, are also monitored by the same networks. The historical time series of those variables, generated by the networks, are the source of input data for SMOGSTOP. Photochemical ozone pollution is the result of complex non‐linear interactions between atmospheric pollutants and meteorology. As a consequence, it is extremely difficult to determine relationships between source emissions (ozone precursors) and ambient pollutant concentrations (ozone). To deal with this complexity, SMOGSTOP was constructed as an empirical model, applying a VITO tailor‐made methodology called stratified pattern matching, to link meteorological and precursor information into ozone forecasts. In this paper, an overview of the process of ozone formation in Belgium is given, followed by the definition of the explanatory variables which will be used in the model. Then the methodology behind the model is reported and finally the results of the forecasting efforts in the period of 1.5.1995 till 31.8.1995 are presented. Copyright © 2000 John Wiley & Sons, Ltd.

Spatial interpolation of ambient ozone concentrations from sparse monitoring points in Belgium

“…At these fixed locations an historical dataset is available, and based on these data a statistical short-term ozone forecasterz (for daily ozone maxima) has been operational since 1996. 5,6 However, both the measurements and the forecasts are restricted to the sparse monitoring points and some inductive model is still needed to estimate the full spatial ozone state. Since this interpolation should complement the prediction model, it has to be fast in operational mode.…”

A neural network based model to forecast hourly ozone levels in rural areas in the Basque Country

“…In the last decades cause/effect models and statistical models have been developed with the purpose of forecasting O 3 hourly levels [1]. Recently, the artificial neural networks have become very useful in the elaboration of prognostic models to forecast air quality levels [2,3].…”