Europe-wide high-spatial resolution air pollution models are improved by including traffic flow estimates on all roads

Youchen Shen*, Kees de Hoogh, Oliver Schmitz, John Gulliver, Danielle Vienneau, Roel Vermeulen, Gerard Hoek, Derek Karssenberg

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Road traffic is an important source of noise and air pollution. Modelling of air pollution and noise therefore requires detailed information on annual average daily traffic (AADT) flows on all roads. Europe-wide estimates on traffic intensity are however not publicly available. This has hampered previous Europe-wide air pollution and noise modelling, used extensively in Europe-wide epidemiological studies of morbidity and mortality. We aim to estimate Europe-wide AADT and quantify potential improvements of previous Europe-wide air pollution models. We built separate random forests (RF) models for different road types in OpenStreetMap (highway, primary, secondary and tertiary, and residential roads). We collected observations on annual average daily traffic (AADT) from six European countries. We evaluated our AADT models using 5-fold cross-validation (CV) and by comparison of our Europe-wide traffic flow estimates with national traffic model estimates for Switzerland and the Netherlands. We evaluated whether adding our estimated AADT as predictors for Europe-wide air pollution models trained by more than 2000 routine monitoring sites improved the performance of the models based upon major road length in different buffer sizes. The 5-fold cross-validation result showed our estimates overall captured variations in AADT between road types (R2 = 0.82). Our result showed variability in AADT within and between road types, documenting the benefit of our model framework at a continental scale. Our AADT estimates modestly improved model performance of previous Europe-wide air pollution models for NO2, PM10, PM2.5, and O3, especially for NO2 (3% improvement of geographically-weighted regression model). Improvement of model performance was larger in urban areas (5% and 8% increases in R2 for NO2 and O3). Importantly, more detailed intra-city near-road variations were captured for traffic-related air pollution. The resulting AADT estimates of all roads across Europe will be useful for further improving air pollution modelling and facilitating harmonized road traffic noise modelling in Europe.

Original languageEnglish
Article number120719
Number of pages16
JournalAtmospheric Environment
Volume335
DOIs
Publication statusPublished - 15 Oct 2024

Bibliographical note

Publisher Copyright:
© 2024 The Authors

Funding

This work was supported by EXPANSE and EXPOSOME-NL projects. The EXPANSE project is funded by the European Union's Horizon 2020 research and innovation programme under grant agreement No 874627. The content of this article is not officially endorsed by the European Union. The EXPOSOME-NL project is funded through the Gravitation programme of the Dutch Ministry of Education, Culture, and Science and the Netherlands Organization for Scientific Research (NWO grant number 024.004.017) . The authors declare no competing financial in-terest. We thank technical and information support from Benjamin Flueckiger from Swiss Tropical and Public Health Institute (Swiss TPH) and Marta Cirach from Barcelona Institute for Global Health (ISGlobal) .

FundersFunder number
European Union874627
Gravitation programme of the Dutch Ministry of Education, Culture, and Science and the Netherlands Organization for Scientific Research (NWO)024.004.017
EXPANSE project
EXPOSOME-NL project

    Keywords

    • Air pollution
    • Geographic information system (GIS)
    • Land-use regression
    • Machine learning
    • Road traffic intensity

    Fingerprint

    Dive into the research topics of 'Europe-wide high-spatial resolution air pollution models are improved by including traffic flow estimates on all roads'. Together they form a unique fingerprint.

    Cite this