Scalable penalized spatiotemporal land-use regression for ground-level nitrogen dioxide

Abstract

Nitrogen dioxide (NO 2 ) is a primary constituent of traffic-related air pollution and has well-established harmful environmental and human-health impacts. Knowledge of the spatiotemporal distribution of NO 2 is critical for exposure and risk assessment. A common approach for assessing air pollution exposure is linear regression involving spatially referenced covariates, known as land-use regression (LUR). We develop a scalable approach for simultaneous variable selection and estimation of LUR models with spatiotemporally correlated errors, by combining a general-Vecchia Gaussianprocess approximation with a penalty on the LUR coefficients. In comparison to existing methods using simulated data, our approach resulted in higher model-selection specificity and sensitivity and in better prediction in terms of calibration and sharpness, for a wide range of relevant settings. In our spatiotemporal analysis of daily, US-wide, ground-level NO 2 data, our approach was more accurate, and produced a sparser and more interpretable model. Our daily predictions elucidate spatiotemporal patterns of NO 2 concentrations across the United States, including significant variations between cities and intra-urban variation. Thus, our predictions will be useful for epidemiological and risk-assessment studies seeking daily, national-scale predictions, and they can be used in acute-outcome health-risk assessments.

Correlation-based sparse inverse Cholesky factorization for fast Gaussian-process inference

“…Special cases of our general CVecchia idea have already been successfully employed in several applications (which were started later but completed earlier than the present paper): Katzfuss et al (2022) used the idea to approximate anisotropic GPs for computer-model emulation in high input dimension; Messier and Katzfuss (2021) approximated spatio-temporal land-use regression for ground-level nitrogen dioxide; and in the context of nonparametric inference (Kidd and Katzfuss, 2021), ideas related to CVecchia were used with sample correlations instead of parametric correlations.…”

A geospatial modeling approach to quantifying the risk of exposure to environmental chemical mixtures via a common molecular target

“…Another data gap is the accurate determination of human exposure to multiple chemicals from environmental sources such as ambient air. Advancements in methods used to determine human exposure to environmental chemicals include biomonitoring and environmental modeling techniques such as land-use regression and geostatistical models ( Dennis et al, 2017 ; Katzfuss, 2017 ; Katzfuss et al, 2020 ; Messier et al, 2014 ), mechanistic chemical transport and dispersion models ( Appel et al, 2021 ), and hybrid mechanistic and geostatistical models ( Cleland et al, 2020 ; Messier and Katzfuss, 2021 ; Wang et al, 2016 ). Additionally, exposure models have been used to estimate simple exposure heuristics of near-field (i.e., personal and consumable product) and far-field (i.e., ambient environment) chemical sources ( Wambaugh et al, 2014 ) and can estimate exposure and internal concentration by chemical class and real-world demographic characteristics (e.g., age, sex).…”