Wednesday, Aug 7: 10:30 AM - 12:20 PM
5155
Contributed Papers
Oregon Convention Center
Room: CC-D140
Main Sponsor
Section on Statistics and the Environment
Presentations
Wildfire events have been increasing in frequency, duration, and severity. Coupled with these events is poor air quality, which is known to be detrimental to health. Air pollution is a complex mixture, and traditional approaches to understanding exposure-health associations, which typically examine one pollutant at a time, do not capture the multidimensional dynamics and correlational structure of its multiple co-existing components. While some multi-pollutant modelling strategies have been developed, there remain shortcomings. We adapt a principal stratification approach so that its application can be generalized to a broad range of environmental and health data. We demonstrate these developments using chemically speciated particulate matter air pollution concentrations coupled with satellite-derived wildfire smoke plumes in California, which have been matched to participants in the California Teachers Study (CTS) cohort. We examine the association between different biomarkers of inflammation with the multi-pollutant mixture under wildfire and non-wildfire conditions.
Keywords
Principal Stratification
Multi-Pollutant Models
Propensity Score Matching
Sliced Inverse Regression
The Council on Environmental Quality's Climate and Economic Justice Screening Tool defines "disadvantaged communities" (DAC) in the USA, highlighting where benefits of climate and energy investments are not accruing. Understanding the impact of economic factors such as income and employment on DAC is crucial for addressing economic well-being and reducing inequalities. However, investigating the individual impacts of income and employment categories is challenging due to their highly intercorrelated nature, influenced by numerous hidden factors. To address this, we employ principal component generalized linear regression to model their relationship to DAC status. We (1) identify the significant income groups and employment industries impacting DAC status, (2) predict DAC distribution spatially across census tracts, comparing predictive accuracy with conventional machine learning methods, (3) project historical DAC probabilities, and (4) spatially downscale DAC across block groups. Our study provides valuable insights for policymakers and stakeholders to develop strategies that promote sustainable development and address inequities in climate and energy investments in the USA.
Keywords
Disadvantaged communities
Socio-economic challenges
Principal component generalized linear model
Spatial distribution
Spatial downscaling
Temporal trend
Public transport is commonly connected to advantages such as mitigating traffic congestion and improving air quality. The Elizabeth Line, introduced in 2022, represents the most significant single increase in London's transport capacity in over 70 years. Connecting surrounding cities, a major airport, and central employment centres, this line is anticipated to increase the rail capacity in central London by 10%. Using meteorological normalisation for confounding control, repeated change point detection for response identification via hypothesis testing, and a regression discontinuity design for causal inference, our study finds heterogeneous responses in air pollution across different places in London. Changes in NO2 concentrations ranged from -9% to 0% in the short run and -15% to 0% in the long run. The comparison across different regions reveals more significant pollution reductions in inner and outer London at the town-wide level, whereas central London experienced greater decreases near roads. Our findings highlight the potential of public transport improvements in mitigating air pollution while emphasising the importance of accounting for the spatial heterogeneity of effects.
Keywords
Air Pollution
Causal Analysis
Public Transport
Meteorological Normalisation
Change Point Detection
Given the increasing amount of evidence suggesting a connection between climate change and health disparities, this study utilized a statistical approach to explore the relationship between climate policies in the United States, past temperature changes, and vulnerable populations. By analyzing data from the United States excluding D.C., Hawaii, and Puerto Rico, we assessed the relationship between minority demographics, temperature changes (2011-2021), and climate policies (2021). We found no direct link between policy adoption and prior temperature experiences or minority percentages. However, states with climate policies in place had consistently higher Asian American populations than those without such policies. Additionally, we found significant differences in the demographic composition of Black or African American or below the poverty line populations in states with electricity policies in place compared to states without such policies. Lastly, we found that states with higher percentages of people below the poverty line were less likely to have carbon pricing in place suggesting a nuanced relationship between policy types and demographic compositions within states.
Keywords
Environmental Justice
Health Policy
Climate Change
Public Policy
The association between long-term exposure to fine particulate matter (PM2.5) and the risk of various health outcomes such as mortality and major adverse cardiovascular events has been extensively documented over the past several decades. However, these previous studies often lack comprehensive evaluations regarding how a proposed policy implementation might influence this prevalent public health concern. In response, we propose a balancing weight framework to estimate and assess counterfactual outcomes under the assumption that the distribution of exposures has been shifted through policy interventions. The weights utilized for evaluating these counterfactual outcomes are designed to optimally balance the moments and correlations of the covariates with the factual exposures, and match them with those derived from the shifted counterfactual exposures. To illustrate the applicability of our method, we evaluate the potential impact of several policy interventions to PM2.5 on lowering the incidence of major adverse cardiovascular events in a randomly selected 10% cohort of high-risk Medicare recipients. We support our application with a numerical study of the proposed methodology.
Keywords
Causal Inference
Stochastic Interventions
Calibration Weights
Air Pollution Epidemiology
Cardiovascular Disease
Ambient air pollution measurements from regulatory monitoring networks are routinely used to support epidemiologic studies and environmental policy decision making. However, regulatory monitors are spatially sparse and preferentially located in areas with large populations. Numerical model output can be leveraged into the inference and prediction of air pollution data combining with measurements from monitors. Nonstationary covariance functions allow the model to adapt to spatial surfaces whose variability changes with location like air pollution data. In the paper, we employ localized covariance parameters learned from the numerical output model to knit together into a global nonstationary covariance, and use this nonstationary covariance in a fully Bayesian model in which the unknown spatial process has a Gaussian process prior distribution. We model the nonstationary structure with greatly reduced number of parameters to make it computationally feasible.
Keywords
spatial model
Bayesian model
non-stationary model
air pollution
environment