Large Language Model-facilitated national review on the use of ecological tools and processes in Environmental Impact Statements (EIS) in the U.S. with demonstrated use cases for environmental planners, legal practitioners, and researchers

<title>Abstract</title> <p>Recent changes to the National Environmental Policy Act (NEPA) in the U.S. call for a more efficient environmental review process for Environmental Impact Statements (EISs). There is a need for a centralized, curated ecology data store that facilitates faster searches and the use of precedents from previous EISs. By harvesting PermitAI’s EIS corpus and employing a customized Retrieval-Augmented Large Language Model pipeline with quality control and accuracy benchmarking, we conducted a large-scale review of 2000 + EISs across the country. We systematically extracted and annotated uses of ecology ranging from databases, protocols, methodologies, maps, models, and policy processes that support biological resource management. Using the ecology data store, we found that EISs in the Western U.S. generally mention more uses of ecology than those in the Midwestern and Southern states. Meanwhile, offshore wind projects and the Bureau of Ocean Energy Management lead in the total number of uses of ecology in their respective EISs. To use the ecology data store for decision support, we conducted case studies to demonstrate that: 1. an environmental planner can compare past harbor improvement projects to conduct pre-NEPA risk mapping and anticipate additional ecological analyses needed among potential action alternatives; 2. a legal practitioner can establish an ecological tool benchmark for the Nevada metal mining industry to conduct an omission analysis of a recent salient Lithium project; 3. an ecological researcher can identify where and how DNA methodologies have been used to facilitate future project development and collaborations. The data store and scripts are made publicly available.</p>