Skip to Main content Skip to Navigation
Conference papers

National scale LUR model over France for PM10, PM2.5, NO2 and O3: development, validation and back-extrapolation from 2010 to 1990

Abstract : BACKGROUND AND AIM: In the context of a case-control study nested in a French cohort of about 10400 subjects, on air pollution exposures and breast cancer risk, we reconstructed subjects' exposures to four pollutants (PM10, PM2.5, NO2 and O3) between 1990 and 2010, using land use regression models. METHODS: We used databases of ambient air measurements, a large number of geographical variables and background pollution data provided by a country-wide chemistry-transport model (CTM) with an 8x8km resolution. We first built baseline models for the most recent year of the study period (2010) and validated them via cross-validation (pooling of predictions from leave-20%-out). In a second step, we extrapolated the baseline models to 1990. For each pollutant, we tried multiple backextrapolation approach (trends established with concentrations measured from remote location or simulated by the CTM). We selected the best method, based on the performance compared to the measurement data available between 1990 and 2009 (and non-used for the backextrapolation). RESULTS:The cross-validation of the baseline models showed robust results (CV-R²; CV-RMSE): NO2 (0.67; 7.5), PM10 (0.64; 3.3), PM2.5 (0.51; 2.6) and O3 (0.60; 6.2). Based on the baseline models, we produced high spatial resolution maps of the French territory (50x50m). The extrapolation of the baseline models to 1990 seems to produce better results if based on trends established at the local scale (8x8km) with the CTM data. For NO2 and O3, the statistical indicators remain fairly stable over the period. For PM10 and PM2.5, we observe a drop in performance over the period 1990-1998. This can partly be explained by the very low number of PM10 and PM2.5 measurement data available in France before 2000, which limits extensive comparisons. CONCLUSIONS:This work shows that it is possible to accurately reconstruct exposures over very large territories and long periods thanks to the input of simulated background pollution data.
Document type :
Conference papers
Complete list of metadata
Contributor : Gestionnaire Civs Connect in order to contact the contributor
Submitted on : Thursday, January 6, 2022 - 1:40:12 PM
Last modification on : Friday, January 7, 2022 - 3:22:53 AM


  • HAL Id : ineris-03514688, version 1



Thomas Coudon, Lény Grassot, Florian Couvidat, Béatrice Fervers, John Gulliver. National scale LUR model over France for PM10, PM2.5, NO2 and O3: development, validation and back-extrapolation from 2010 to 1990. 33. Annual Conference of the International Society for Environmental Epidemiology (ISEE 2021): Promoting Environmental Health and Equity in a Shifting Climate, International Society for Environmental Epidemiology, Aug 2021, New York, United States. ⟨ineris-03514688⟩



Les métriques sont temporairement indisponibles