Sampling design for air quality measurement surveys: An optimization approach

Abstract

Measurement surveys using passive diffusion tubes are regularly carried out to elaborate atmospheric concentration maps over various areas. Sampling schemes must be designed to characterize both contaminant concentrations (of benzene or nitrogen dioxide for example) and their relations to environmental variables so as to obtain pollution maps as precise as possible. The concentration variable is interpolated by external drift kriging, with the help of exhaustively known covariates. The quality of a sampling scheme is quantified by the spatially averaged external drift kriging variance, which incorporates the drift estimation error as well as the spatial interpolation error. A weighted criterion is also introduced. Optimizing this criterion by simulated annealing provides an optimal sampling scheme. A preliminary study is performed on concentration data available from previous surveys on two different agglomerations so as to determine the covariance model and the relevant covariates to be used on a third agglomeration. It does not reveal a single model but a whole parametrized family of relevant models. The method is then applied to the third agglomeration with different values of the model parameters. The results are discussed and finally compared to those obtained with a pragmatic approach, described in a previous paper.