Machine learning predicts which rivers, streams, and wetlands the Clean Water Act regulates
Creators
-
1.
California Institute of Technology
Abstract
We assess which waters the Clean Water Act protects and how Supreme Court and White House rules change this regulation. We train a deep learning model using aerial imagery and geophysical data to predict 150,000 jurisdictional determinations from the Army Corps of Engineers, each deciding regulation for one water resource. Under a 2006 Supreme Court ruling, the Clean Water Act protects two-thirds of US streams and more than half of wetlands; under a 2020 White House rule, it protects less than half of streams and a fourth of wetlands, implying deregulation of 690,000 stream miles, 35 million wetland acres, and 30% of waters around drinking-water sources. Our framework can support permitting, policy design, and use of machine learning in regulatory implementation problems.
Copyright and License
© 2024 the authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original US government works. https://www.sciencemag.org/about/science-licenses-journal-article-reuse
Acknowledgement
Participants at AERE, the California Water Boards, the Energy Institute at Haas, EPA, Global Policy Lab, Stanford, and TWEEDS seminars, E. Benami, M. Burke, J. Corona, C. Costello, T. Doley, J. Hewitt, S. Hsiang, N. Kohli, R. Kwok, M. Massey, A. McGartland, S. Mullainathan, K. Swann, and C. Taylor provided useful comments and discussions. We thank B. Walker for research assistance.
Funding
This material is based upon work supported by the following: Ciriacy-Wantrup Postdoctoral Fellowship (S.W.); The Giannini Foundation (J.S.S.); Google Cloud Research Credits Program (H.D., D.A.K., J.S.S.); NIH grant 1R01AG079914-01 (D.A.K., J.S.S.).
Contributions
Conceptualization: J.S.S. Data curation: S.G., H.D., S.W., D.A.K., J.S.S. Formal analysis: S.G., H.D., S.W., D.A.K., J.K.M., N.Y., A.T., J.S.S. Funding acquisition: H.D., D.A.K., M.G., A.T., J.S.S. Investigation: S.G., H.D., S.W., D.A.K., M.G., J.K.M., N.Y., A.T., J.S.S. Methodology: S.G., H.D., S.W. Software: S.G., H.D., S.W., D.A.K., J.S.S. Visualization: S.G., H.D., S.W., J.S.S. Writing – original draft: S.G., H.D., S.W., D.A.K., J.S.S. Writing – review and editing: S.G., H.D., S.W., D.A.K., J.S.S.
Data Availability
All code is available on Zenodo (29). All data are publicly available online (15–17, 19, 20, 30–34). All data used to train the models are on Dryad (35), as is a subset of the prediction data, as well as model weights and all data required to produce the results presented here (36). The model is freely available for research and noncommercial purposes (29, 35, 36), but its commercial use is limited (patent pending).
Conflict of Interest
D.A.K. serves on the scientific advisory board of the EPA and serves as an expert witness on water pollution litigation. During the research, J.S.S. served as an adviser on trade and environment to the European Commission’s Directorate General of Trade. D.A.K. and J.S.S. wrote a public report on EPA’s analyses of the CWR and the NWPR as part of the External Environmental Economics Advisory Committee, funded by the Sloan Foundation. All authors of this paper are inventors on a patent pending with serial number 63/513,464, submitted by UC Berkeley, which covers WOTUS-ML. The software has a Creative Commons Attribution Non Commercial No Derivatives 4.0 International license, which freely allows use for research and other noncommercial purposes. Potential commercial users should contact the Office of Technology Licensing at UC Berkeley.
Files
science.adi3794_sm.pdf
Additional details
Identifiers
- ISSN
- 1095-9203
Funding
- University of California, Berkeley
- Ciriacy-Wantrup Fellowship
- A.P. Giannini Foundation
- Google (United States)
- National Institutes of Health
- 1R01AG079914-01