SPATIAL AND TEMPORAL VARIABILTY IMPACT ON AIR POLLUTION INTERPOLATION: A CASE STUDY ON PM10 ESTIMATION IN NORTHEN-FRANCEKHAOULA KARROUM, ANTON SOKOLOV, YANN BEN MAISSA, MOHAMED EL HAZITI AND HERVÉ DELBARREAbstract In this work, we compare the performance of a set of spatial interpolation techniques in estimating PM10 concentrations for Hauts-de-France region. Estimations of Coefficient of Determination, Root Mean Square Error and corresponding 95% confidence intervals show that classic and optimized version of Inverse Distance Weighting method and Gaussian Process Regression with two different kernels give comparable results. The spatial distribution of the error shows the high dependence on industry and coastal atmospheric phenomena. To assess the influence of the local meteorological effect on pollution dispersion, we estimated the Coefficient of Determination for the interpolation of time-averaged pollution data. It has a clear 24-hour maximum, corresponding to periodic atmospheric effects, such as the sea breeze. The sensitivity of interpolation techniques to the noise in measurements and to the data density shows that methods behave in the same way, leading to a bigger Root Mean Square Error following the magnitude of data perturbation and decreasing with the number of stations. In addition to that, regions with higher error are less sensitive to the perturbation. |