Causal health impacts of power plant emission controls under modeled and uncertain physical process interference

• DOI: 10.1214/24-aoas1904
• Publication Date: 2024-10-31T17:54:54Z
• Authors: Nathan B. Wikle; Corwin M. Zigler
• Keywords: Methodology (stat.ME); FOS: Computer and information sciences; 13. Climate action; Applications (stat.AP); Statistics - Applications; Statistics - Methodology

Causal inference with spatial environmental data is often challenging due to the presence of interference: outcomes for observational units depend on some combination local and nonlocal treatment. This especially relevant when estimating effect power plant emissions controls population health, as pollution exposure dictated by: (i) location point-source well (ii) transport pollutants across space via dynamic physical-chemical processes. In this work we estimate effectiveness air quality interventions at coal-fired plants in reducing two adverse health Texas 2016: pediatric asthma ED visits Medicare all-cause mortality. We develop methods causal interference underlying network structure not known certainty instead must be estimated from ancillary data. Notably, uncertainty propagated resulting estimates. offer a Bayesian, mechanistic model mapping, which combine flexible nonparametric outcome marginalize estimates effects over interference. our analysis finds evidence that upwind reduce mortality; however, accounting renders results largely inconclusive.