Abstract: |
Mapping atmospheric pollutants in urban areas is a growing requirement from municipalities, because of obvious health-related issues. Such mapping frequently relies on a limited monitoring network that might be combined with physicochemical models. Classical interpolation approaches are usually based on background measurements only, due to significantly higher spatial variability of near-road pollution; ignoring the highest concentration levels in the produced maps has several drawbacks: tricky communication, inability to provide relevant estimates for population exposure. On the other hand, numerical models are pushed towards local scales, leading to intensive computation needs. In various contexts, geostatistics may provide appropriate methods to bridge the gap between background and near-road pollution during air quality mapping. Two typical cases are presented: (1) When a physicochemical model is available for the road network only, a first approach consists in: (i) mapping classically the background pollution using related measurements and potential cofactors (emission cadastre, land use), (ii) correcting the possible bias between near-road measurements and the output of the physicochemical model on the road network, (iii) spatializing the traffic concentrations consistently with the distance of impact of the roads. If no road model is available, the near-road measurements can be spatialized thanks to knowledge of the road network, its pollutant emissions and certain traffic conditions. (2) Physicochemical models commonly simulate the concentrations of chemical compounds on a regular grid which is refined along the road network. The outcome of the numerical simulation is then linearly interpolated. This post-processing usually creates artefacts around the road network, due to the distribution of the simulation nodes. Accounting for locally varying anisotropies improves this post-processing and reduces the number of computation nodes (and consequently the computation time) which is required to guarantee the quality of the physicochemical model output. Both approaches are illustrated on real cases coming from several French cities, together with a discussion about their advantages/drawbacks for air quality mapping in urban areas. |
