Development of European NO2 Land Use Regression Model for present and future exposure assessment: Implications for policy analysis

Abstract

A new Land Use Regression model was built to develop pan-European 100 m resolution maps of NO₂ concentrations. The model was built using NO₂ concentrations from routine monitoring stations available in the Airbase database as dependent variable. Predictor variables included land use, road traffic proxies, population density, climatic and topographical variables, and distance to sea. In order to capture international and inter regional disparities not accounted for with the mentioned predictor variables, additional proxies of NO₂ concentrations, like levels of activity intensity and NO_x emissions for specific sectors, were also included. The model was built using Random Forest techniques. Model performance was relatively good given the EU-wide scale (R² = 0.53). Output predictions of annual average concentrations of NO₂ were in line with other existing models in terms of spatial distribution and values of concentration. The model was validated for year 2015, comparing model predictions derived from updated values of independent variables, with concentrations in monitoring stations for that year. The algorithm was then used to model future concentrations up to the year 2030, considering different emission scenarios as well as changes in land use, population distribution and economic factors assuming the most likely socio-economic trends. Levels of exposure were derived from maps of concentration. The model proved to be a useful tool for the ex-ante evaluation of specific air pollution mitigation measures, and more broadly, for impact assessment of EU policies on territorial development.

Keywords: Air pollution; EU policies; Exposure; Future-extrapolation; Nitrogen dioxide; Random forest.