Performance of Road-Traffic-Based Exposure Proxies Against Personal PM2.5 Measurements in Three Sub-Saharan African
Handsome Bongani Nyoni1,2, Terence Darlington Mushore2,3, Laura Munthali4,5
1Place Alert Labs, Surveying and Geomatics Department, Faculty of the Built Environment, Art and Design, Midlands State University, Gweru, Zimbabwe.
Medrxiv : the Preprint Server for Health Sciences
|March 27, 2026
Summary
Road-traffic proxies offer context-specific insights into particulate matter (PM2.5) exposure, especially where monitoring is limited. Combining these proxies in a hybrid model enhances predictive accuracy for PM2.5 exposure.
Area of Science:
- Environmental Health
- Air Quality Monitoring
- Spatial Epidemiology
Background:
- Particulate Matter (PM2.5) exposure is a significant global health concern.
- PM2.5 monitoring is often sparse and uneven, particularly in areas with limited ground-based networks.
- Road-traffic indicators can serve as valuable proxies for estimating PM2.5 levels in data-scarce regions, but require validation.
Purpose of the Study:
- To evaluate the effectiveness of three road-traffic related proxies for estimating personal PM2.5 exposure.
- To validate these proxies against high-resolution PM2.5 measurements in The Gambia, Kenya, and Mozambique.
- To assess the utility of combining proxies to improve PM2.5 exposure predictions.
Main Methods:
- Utilized personal PM2.5 exposure data from 343 postpartum participants over one year.
- Employed spatial mapping with 5 km hexagonal tessellations to analyze village-level PM2.5-proxy relationships.
- Assessed proxy-PM2.5 associations using Spearman correlation and Random Forest models, evaluating R², RMSE, and feature importance.
Main Results:
- Spatial patterns of PM2.5 and proxy relationships were heterogeneous across study sites.
- Population-Weighted Road Network Density (WRND) showed weak correlations with PM2.5 overall.
- Euclidean distance proxies (EH, EM) exhibited variable and sometimes negative correlations with PM2.5 across countries; combining proxies improved predictive performance in Kenya and Mozambique.
Conclusions:
- Road-network proxies provide context-dependent indicators of personal PM2.5 exposure.
- The predictive performance of PM2.5 exposure estimation is enhanced by combining multiple proxies within a hybrid model.
