Air pollution modelling for birth cohorts: a time-space regression model

Area of Science:

• Environmental Epidemiology
• Air Quality Monitoring
• Spatio-temporal Modeling

Background:

• Investigating air pollution's impact on pregnancy and early life requires models capturing spatial and temporal exposure variability.
• Existing models may not adequately address the dynamic nature of air pollutants during critical developmental windows.

Purpose of the Study:

• To develop and validate a time-space exposure model for ambient nitrogen dioxide (NO2) concentrations.
• To assess the model's suitability for epidemiological studies focusing on vulnerable populations.

Main Methods:

• Utilized passive NO2 monitoring data from rural and urban sites in Bern, Switzerland (1998-2009).
• Employed multivariable regression, incorporating spatial (land use, traffic, population) and temporal (meteorology, continuous monitoring) predictors.
• Performed internal cross-validation and external validation using participant home measurements.

Main Results:

• Traffic-related variables and fixed-site NO2 measurements were key predictors in both rural and urban models.
• Achieved robust model performance with R-squared values of 0.63 (rural) and 0.54 (urban) for internal validation.
• External validation yielded R-squared values of 0.54 (rural) and 0.67 (urban), confirming model reliability.

Conclusions:

• The developed time-space NO2 model effectively predicts air pollution exposure.
• This approach is well-suited for epidemiological research on time-sensitive health effects during pregnancy and infancy.