École des mines de Paris, Centre de géosciences/géostatistique, 35, rue Saint-Honoré, 77305 Fontainebleau, Geovariances, 49 bis, avenue Franklin Roosevelt, 77212 Avon, Air Normand, 21, avenue de la porte des champs, 76000 Rouen
- Key words: air pollution, geostatistics, models, statistical, nonlinear dynamics, risk assessment, threshold limit values
- DOI : 10.1684/ers.2007.0088
- Page(s) : 207-18
- Published in: 2007
How to evaluate the risks of exceeding limits: Geostatistical models and their application to air pollutionGeostatistics is increasingly applied to the study of environmental risks in a variety of sectors, especially in the fields of soil decontamination and the evaluation of the risks due to air pollution. Geostatistics offers a rigorous stochastic modelling approach that makes it possible to answer questions expressed in terms of uncertainty and risk. This article focusses on nonlinear geostatistical methods, based on the gaussian random function model, whose essential properties are summarized. We use two examples to characterize situations where direct and thus rapid methods provide appropriate solutions and cases that inevitably require more laborious simulation techniques. Exposure of the population of the Rouen metropolitan area to the risk of NO
2 pollution is assessed by simulations, but the surface area where the pollution exceeds the threshold limit can be easily estimated with nonlinear conditional expectation techniques. A second example is used to discuss the bias introduced by direct simulation, here of a percentile of daily SO
2 concentration for one year in the city of Le Havre; an operational solution is proposed.