Application of kriging to air quality mapping at different time and spatial scales
Résumé
The development of geostatistics to map air quality is mainly driven by two objectives : meeting the information requirements of the European directives ; increasing knowledge about population exposure to air pollution. The present review outlines the methodological studies conducted by INERIS to support this evolution. It is focused on the estimation of nitrogen dioxide concentration patterns which can be derived from numerous monitoring surveys. The analysis is organized by time and space scales, pointing out the most significant issues. The geostatistical kriging methods can serve two purposes : interpolation from sampling data collected for several weeks in summer and/or in winter ; daily correction of deterministic model output using data from automatic monitoring stations. The first point deals with mapping time-integrated concentrations over a region or a city. Several questions have been investigated : which sampling periods and sampling mesh are most suitable ? Which is the best way to select and introduce auxiliary variables ? What is the uncertainty of a map ? How roadside concentrations can be incremented to account for small-scale pollution due to traffic ? Most of those preoccupations concern the second application as a rigorous selection of the data points and a careful adjustment of all kriging parameters are a prerequisite to any operational use of geostatistics.