Automatic Mapping Algorithm of Nitrogen Dioxide Levels from Monitoring Air Pollution Data Using Classical Geostatistical Approach: Application to the French Lille City

This work aims to test a new method of automatic mapping of gaseous nitrogen dioxide in an urban area. These maps have to be realised based on the data provided by automatic on-site monitoring stations, taking into account their scarcity that could hamper a correct spatial interpolation of the pollution levels. In the first part of this study, we propose a new methodology to generate additional data, based on several previous field campaigns performed by passive sampling. Among these passive sampling sites some of them (henceforth called “virtual stations”) are time-correlated to a given fixed station (called “reference station”). In the second part of this study, we have tested the suitability of our method for the automated generation of variograms on a case study as well as the quality of the estimations calculated based on these data. For mapping, geostatistical methods were applied, particularly the cokriging one. This multivariable method exploits the additional information given by auxiliary variables; in the case of nitrogen dioxide, variables depicting the area, such as the population density or the emissions inventory, may therefore be used. In order to take into account the uncertainty of the data generated in the virtual stations, we included in the variance-covariance kriging matrix an additional component called the variance of measurement error (VME); a methodology to calculate this component is described. Finally, the resulting maps are well detailed and do show the main features of the pollution due to nitrogen dioxide on the considered domain