Bayesian spatio-temporal modeling of fine particulate matter in data-scarce urban environments: a case study of Bujumbura, Burundi

Résumé

Abstract Sub-Saharan Africa, along with other data-scarce urban environments, suffers from a lack of reliable air quality data, limiting evidence-based actions for public health and urban sustainability. Bujumbura, Burundi’s largest and fastest-growing city, lacks systematic air quality monitoring despite rapid urbanization. This study presents a city-scale, high-resolution assessment of fine particulate matter (PM 2.5 ) in Bujumbura using a hybrid deployment of low-cost sensors that combines fixed and rotating sites across 27 locations representing four land-use categories. Spatial and temporal dependencies were modeled using a Bayesian hierarchical framework based on the stochastic partial differential equation (SPDE) approach implemented with integrated nested Laplace approximation. The results reveal pronounced spatial contrasts, with lower concentrations in administrative, business, and industrial areas (48.3 μ g m − 3 ) and higher levels in high-density residential and town background zones (59.5–60.6 μ g m − 3 ). Diurnal and daily patterns show persistent morning–evening peaks, exceeding World Health Organization guideline values. Model validation based on out-of-sample cross-validation indicates moderate-to-strong predictive skill, with a root mean square error of 0.35 on the logarithmic scale ( μ g m − 3 ) and a correlation coefficient of r = 0.77 . These findings demonstrate that combining cost-effective sensor networks with Bayesian SPDE-based modeling can yield actionable insights into urban PM 2.5 exposure in data-limited cities.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0