Virginia Tech
Integrated assessment modeling to characterize environmental, health, and economic tradeoffs in electricity grid decarbonization in the northeastern United States
Abstract
dc:description.abstractElectricity grid decarbonization is a cornerstone of climate change policy; however, proposed pathways frequently face delays stemming from disagreements and controversies over social, economic, and environmental impacts. In the northeastern United States, grid transformation is shaped not only by technical and economic constraints, but also by contested causal claims, uncertainty regarding system performance, and heterogeneous public valuations of environmental, health, and economic tradeoffs. These tensions are compounded by limitations in commonly used analytical tools, which often struggle to resolve disagreement in socially credible ways. Conventional electricity planning relies heavily on capacity expansion models and environmental impact assessments. While essential for evaluating reliability and cost, these tools often depend on deterministic assumptions, simplified treatments of uncertainty, and indirect incorporation of environmental and health impacts. As a result, they may inadequately characterize indirect causal relationships, obscure uncertainty in renewable-dominated systems, and fail to capture distributional and health consequences central to public concern. In this dissertation, we develop and apply an integrated assessment framework to characterize environmental, health, and economic tradeoffs associated with electricity grid decarbonization in the northeastern United States. The framework integrates three analytical components addressing interrelated sources of disagreement. First, we apply Bayesian causal inference methods to evaluate contested claims regarding indirect environmental impacts of transborder electricity transmission. Opponents of transborder interconnections have argued that generation-side impacts of hydroelectric reservoirs in Canada should be counted as impacts of increased transborder transmission. Using historical data from 1979–2021, we find no empirical support for the claim that expanded transmission capacity directly stimulates new hydroelectric generation in Canada. Generation expansion is primarily associated with domestic Canadian demand and US--Canada price differentials, with transmission acting through indirect price-signal pathways rather than as a direct cause. We acknowledge that these price differences are not independent of the existing import relationship: the long-standing presence of import infrastructure has shaped Canadian export pricing, demand projections, and investment decisions over decades, so the price differential and the import relationship are partially co-determined rather than fully separable. What our analysis rules out is the narrower causal claim that any specific new transmission project triggers new dam construction. This distinction matters for environmental impact assessment scope and for the policy levers available to opponents of cross-border transmission (e.g., environmental non-profits), who would need to address price differentials and demand growth rather than transmission capacity itself to influence Canadian generation decisions. Whether advocacy groups would in practice redirect attention to these less visible levers is an open political question; the visibility of transmission infrastructure makes it a more salient organizing target than wholesale price design, regardless of the underlying causal structure. These findings inform the appropriate scope of environmental impact assessment and clarify the range of impacts properly considered within a life-cycle assessment (LCA) framework. Under current attributional LCA practice, a fraction of reservoir emissions is charged to transmission infrastructure, inflating its indirect carbon footprint in ways that have been used to argue against projects in regulatory filings. Our findings support a consequentialist LCA interpretation: because transmission is not causally linked to new dam construction, dam-related emissions are not properly chargeable to transmission, which narrows the appropriate scope of environmental review under NEPA. Second, we develop the Probabilistic Hourly Assessment of System Electricity (PHASE) model to compare alternative decarbonization pathways under probabilistic hourly fossil fuel and renewable generation. By integrating hourly system operations, correlated uncertainty, monetized air pollution health impacts, and non-monetized ecological effects, we show that pathways with similar mean net present costs differ substantially in cost uncertainty bounds and in the intraregional distribution of avoided air-pollution mortality. Monetized air-pollution damages are one to two orders of magnitude smaller than direct system costs across all pathways, but the underlying physical mortality differences between pathways translate to differences in avoided premature deaths that are policy-relevant in their own right and are unevenly distributed across rural and urban communities. Under optimistic SMR deployment assumptions, the SMR-inclusive pathway (C3) achieves the lowest mean cost ($471 billion) but with the widest uncertainty range, while the High Electrification reference pathway (B1) achieves comparable mean cost ($477 billion) with narrower uncertainty. Constraining transmission with Canada increases pathway costs by at least $19.4 billion with greater than 90% confidence, despite overlapping 90% confidence intervals on absolute costs; a tradeoff that is visible only when correlated uncertainties are tracked across pathways. These results demonstrate how deterministic least-cost rankings can mask meaningful tradeoffs under uncertainty. Third, we evaluate the ability of large language models to predict attitudes of demographically and politically diverse individuals, the extent to which adding information about the individuals improves the models' performance, and the performance of different models. We treat this evaluation as exploratory: our human benchmark is a non-representative sample of New England residents recruited via social media, which serves as a methodological reference point against which to compare LLM-generated responses rather than as ground truth for U.S. public opinion. AI tools have been proposed as one method of screening public attitudes in response to infrastructure or other projects, but their performance in the setting of renewable energies is unknown, as is the extent to which demographic and attitudinal covariates such as political lean affects model performance. Across configurations, most LLMs reproduced the core ordinal structure of human responses, correctly ranking environmental health and economic impacts above localized siting-related visual and ecological impacts (e.g., viewshed disruption, bird and bat mortality, project-scale land use change). However, performance varied by architecture: while several models captured the hierarchy, only one LLM replicated the parity between environmental and economic domains, and others showed systematic deviations or failed to reproduce structured tradeoffs (i.e., the ordinal ranking of environmental health and economic impacts above localized siting impacts, and the statistical parity between those environmental health and economic domains). Incorporating demographic and attitudinal covariates, including political lean, generally improved alignment in some domain means but did not consistently recover the full tradeoff structure. Conversely, open-ended responses diverged substantially from the human benchmark, with LLMs systematically shifting narrative emphasis toward energy independence and reliability rather than the economic and climate themes observed in human responses. Across these analyses, we show that persistent disagreement over electricity grid decarbonization pathways reflects unresolved causal questions, incomplete treatment of uncertainty, divergence in both quantitative tradeoff prioritization and qualitative framing within a complex sociotechnical system (e.g. differences in how stakeholders rank the relative importance of environmental health, economic, and visual or ecological impacts), and inability to anticipate social attitudes (e.g. differences in how stakeholders narrate and frame these tradeoffs in their own words which in turn fails to capture true public sentiments). By integrating causal inference, uncertainty-aware system modeling, and empirical analysis of public valuation, this dissertation advances an analytically rigorous and socially responsive framework for electricity grid planning.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Biomedical and Veterinary Sciences
- Department dc:contributor.department
- Biomedical and Veterinary Sciences
- Grantor dc:publisher
- Virginia Tech
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Mortazavigazar, Amir
- Chair dc:contributor.committeechair
-
- Calder, Ryan Spencer Dyas
- Committee members dc:contributor.committeemember
-
- Gohlke, Julia M.
- Grady, Caitlin Anne
- Marston, Landon Todd
- Cohen, Alasdair Gordon
Subjects
dc:subject × 5Rights
dc:rights- Statement dc:rights
-
- In Copyright
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Other
- vt_gsexam:46605