Estimates of population exposure to atmospheric pollution and health-related externalities in a real city: The impact of spatial resolution on the accuracy of results.

Health impacts of atmospheric pollution is an important issue in urban environments. Its magnitude depends on population exposure which have been frequently estimated by considering different approaches relating pollutant concentration and population exposed to it. However, the uncertainties due to the spatial resolution of the model used to estimate the pollutant concentration or due to the lack of representativeness of urban air quality monitoring station (AQMS) have not been evaluated in detail. In this context, NO2 annual average concentration at pedestrian level in the whole city of Pamplona (Spain) modelled at high spatial resolution (~1 m) by Computational Fluid Dynamic (CFD) simulations is used to estimate the total population exposure and health-related externalities by using different approaches. Air pollutant concentration and population are aggregated at different spatial resolutions ranging from a horizontal grid cell size of 100 m × 100 m to a coarser resolution where the whole city is covered by only one cell (6 km × 5 km). In addition, concentrations at AQMS locations are also extracted to assess the representativeness of those AQMS. The case with a spatial resolution of 100 m × 100 m for both pollutant-concentration distribution and population data is used as a reference (Base case) and compared with those obtained with the other approaches. This study indicates that the spatial resolution of concentration and population distribution in the city should be 1 km × 1 km or finer to obtain appropriate estimates of total population exposure (underestimations <13%) and health-related externalities (underestimations <37%). For the cases with coarser resolutions, a strong underestimation of total population exposure (>31%) and health-related externalities (>76%) was found. On the other hand, the use of AQMS concentrations can induce important errors due to the limited spatial representativeness, in particular in terms of population exposure.