The National Environmental Policy Act (NEPA) requires impact assessments before work proceeds on federally funded projects like highway construction, oil exploration and more. The results of the assessments inform decision makers of possible harm to environmental considerations like animal populations and watersheds.
Since NEPA’s implementation in 1970, the findings from those studies have been stored in hard-to-access archives, placing a half century of knowledge – thousands of information-rich documents – beyond the reach of the public and modern data science.
Now the NEPAccess project, led by the University of Arizona, is unlocking data embedded in environmental impact statements (EIS) and environmental analyses required by NEPA. The goal is to create a more efficient, transparent and accountable tool for science-based, democratic environmental governance.
Using machine learning, the project is turning unstructured documents into structured, searchable data in a model that can also be adapted for transforming other troves of government documents into information ripe for meta-analyses.
Devised by a team of data scientists, environmental researchers and scholars in public policy and law, the process begins with manually pulling content to train machine-learning algorithms, enabling automated natural language processing to take over the work of knowledge extraction.
The searchable NEPAccess repository launched in late 2021, and today provides free, full-text and filtered searching of EISs with continued work to provide new functionality in response to stakeholder input and user-experience data.
For the first time, researchers around the world can apply AI, machine learning, georeferencing and other computing tools to study NEPA information across agencies, action types, regions and sectors.