PNNL

Abstract

The National Environmental Policy Act of 1969, as amended (NEPA), mandates that federal agencies assess and document potential environmental impacts before deciding on proposed actions. While significant progress has been made in cataloging and standardizing NEPA documents themselves, the legal challenges that frequently arise from these decisions remain poorly cataloged and largely inaccessible for systematic analysis. Litigation challenging NEPA compliance can substantially delay project timelines, reshape agency decision-making, and establish precedents that influence future environmental reviews — yet no standardized, machine-readable corpus exists that links litigation records to the NEPA projects they contest. The absence of such a resource limits empirical understanding of litigation patterns, impedes risk assessment, and constrains evidence-based efforts to modernize the permitting process. In this work, we publicly release PermitTEC v0.1, a curated metadata corpus of 761 federal court litigation cases related to NEPA and adjacent environmental statutes. The corpus is constructed through an NLP pipeline that employs large language models (LLMs) for extracting contested project references from litigation text and few-shot classification to categorize each case by the nature of its legal challenge — whether it contests a specific NEPA document (e.g., an EIS, EA, or FONSI) or the absence of a required environmental review.%, or compliance with adjacent statutes such as the Clean Water Act (CWA) or Endangered Species Act (ESA). To bridge the gap between litigation records and NEPA project data, we develop and evaluate three complementary matching approaches — LLM-based keyword extraction, fuzzy matching with composite metadata keys, and semantic retrieval — for linking litigation cases to their corresponding project records in NEPATEC v2.0, a corpus of over 140,000 NEPA documents spanning 60,000 projects across more than 60 federal agencies. Together, PermitTEC v0.1 and NEPATEC v2.0 form an integrated permitting-to-litigation data infrastructure that will enable downstream applications, including litigation trend analysis, project-level risk prediction, identification of recurrent grounds for legal challenge, precedent retrieval for legal practitioners, and AI-assisted compliance review — advancing the broader effort to modernize federal environmental permitting through data-driven insights.