Bayesian spatiotemporal modeling for estimating short‐term exposure to air pollution in Santiago de Chile

Abstract

Spatial prediction of exposure to air pollution in a large city such as Santiago de Chile is a challenging problem because of the lack of a dense air‐quality monitoring network. Statistical spatiotemporal models exploit the space–time correlation in the pollution data and other relevant meteorological and land‐use information to generate accurate predictions in both space and time. In this paper, we develop a Bayesian modeling method to accurately predict hourly PM2.5 concentrations in a 1‐km high‐resolution grid covering the city. The modeling method combines a spatiotemporal land‐use regression model for PM2.5 and a Bayesian calibration model for the input meteorological variables used in the land‐use regression model. Using a 3‐month winter‐time pollution data set, the output of sample validation results obtained in this paper shows a substantial increase in accuracy due to the incorporation of the linear calibration model. The proposed Bayesian modeling method is then used to provide short‐term spatiotemporal predictions of PM2.5 concentrations on a fine (1 km2) spatial grid covering the city. Along with the paper, we publish the R code used and the output of sample predictions for future scientific use.

Spatial prediction of air pollution levels using a hierarchical Bayesian spatiotemporal model in Catalonia, Spain

“…There are numerous studies that propose models to estimate the levels of air pollutants, explicitly incorporating both spatial and temporal dependence (Cameletti et al ., 2011, 2013; Pirani et al ., 2013; Shaddick et al ., 2013; Liang et al, 2015, 2016; Calculli et al ., 2015; Cheam et al ., 2017; Mukhopadhyay and Sahu, 2018; Chen et al ., 2018; Clifford et al, 2019; Nicolis et al, 2019; Wan et al, 2021) (to refer to only some of those that have appeared in the last ten years). However, we must point out that very few studies attempt to predict air pollution levels in locations where there is no monitoring station (ie, spatial prediction) (Cameletti et al, 2011, 2013; Pirani et al, 2013; Shaddick et al, 2013; Mukhopadhyay and Sahu, 2018; Nicolis et al, 2019), or, having them, to perform out-of-sample temporal predictions (Wan et al, 2021).…”

A spatio‐temporal model for the analysis and prediction of fine particulate matter concentration in Beijing

“…1), as well as account for the influences of various covariates. The literature abounds with applications of spatio‐temporal models to air quality data (i.e., the recent ones including Calculli, Fassò, Finazzi, Pollice, and Turnone (2015), Cheam, Marbac, and McNicholas (2017), Nicolis, Diaz, Sahu, and Marin (2019), Clifford et al (2019), and Padilla, Lagos‐Álvarez, Mateu, and Porcu (2020)). For example, Cheam et al (2017) applied an expectation‐maximization (EM) algorithm to the inference of a parametric spatio‐temporal mixture model used for clustering the air quality data.…”

“…These studies put more concerns on flexibility of models and computation, but they did not consider the environmental covariates which play an important role in triggering air pollution (Cai, Li, Liao, Wang, & Wu, 2017; Chen, Lu, Li, & Wang, 2015), and they did not apply the model to generate predictions. In addition, some studies developed spatio‐temporal models with environmental covariates involved and used them for space‐time predictions (Calculli et al, 2015; Nicolis et al, 2019; Padilla et al, 2020), however, in contrast to our proposed model, they did not take varying coefficients into account.…”

“…These studies put more concerns on flexibility of models and computation, but they did not consider the environmental covariates which play an important role in triggering air pollution (Cai, Li, Liao, Wang, & Wu, 2017;Chen, Lu, Li, & Wang, 2015), and they did not apply the model to generate predictions. In addition, some studies developed spatio-temporal models with environmental covariates involved and used them for space-time predictions (Calculli et al, 2015;Nicolis et al, 2019;Padilla et al, 2020), however, in contrast to our proposed model, they did not take varying coefficients into account. Specifically, the model in our study is set up based on Fassò and Finazzi (2013) and Finazzi and Fassò (2014), which contains a spatially correlated random field and a separate latent temporal dynamic, and allows for stochastically varying coefficients to incorporate the interaction between spatial random effects and time-invariant covariates.…”

Quantifying the impact of the modifiable areal unit problem when estimating the health effects of air pollution

“…Data on air pollution concentrations are available as point‐level measurements and/or grid‐level modeled concentrations, and average concentrations for each areal unit have been estimated using simple averaging (Lee & Sarran, 2015) or spatial prediction techniques such as block Kriging (Zhu, Carlin, & Gelfand, 2003). A number of different statistical issues have been addressed in the literature in relation to modeling air pollution and its health effects, including different approaches for (i) spatiotemporal pollution prediction (Gilani, Berrocal, & Batterman, 2019; Nicolis, Díaz, Sahu, & Marín, 2019); (ii) estimating individual‐level exposures (Clifford et al, 2019); (iii) the impacts of preferential sampling of pollution (Lee, Szpiro, Kim, & Sheppard, 2015); and (iv) allowing for errors in exposure variables (Keller & Peng, 2019; Strand, Sillau, Grunwald, & Rabinovitch, 2015).…”