This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.
This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.
Regulatory agencies are beginning to recognize per-and polyfluoroalkyl substances (PFAS) as concerning and bioaccumulative compounds of which the use and environmental burden must be reduced. In the specific context of groundwater management, testing and mitigation strategies are desirable but can be time and cost-intensive processes. As a result, only a fraction of all groundwater wells has had testing performed to determine the abundance of PFAS compounds. In this study, we build machine learning models based on real-world groundwater databases from California to accurately predict PFAS levels in the absence of testing. We project that this machine learning model can predict individual PFAS compound abundances with and R2 of 0.72. It can also predict groundwater wells likely to have concerningly high overall levels of PFAS with an accuracy of 91% and an AUC of 0.93. We propose a new regulatory paradigm in which prioritization of PFAS testing in groundwater wells can be supported by such a machine learning approach. Additionally, we believe this approach may have widespread applicability for other hazardous anthropogenic compounds in groundwater.
https://doi.org/10.31223/X5JS3F
Environmental Engineering, Environmental Health and Protection, Environmental Monitoring, Geochemistry, Hydrology, Risk Analysis, Water Resource Management
groundwater, chemical fingerprints, PFAS, open data
Published: 2020-10-19 19:41
Last Updated: 2020-10-19 22:19
CC BY Attribution 4.0 International
Conflict of interest statement:
A.D. is a founder of and equity holder in Coral Genomics. The authors alone are responsible for the views expressed in this publication and they do not necessarily represent the views, decisions or policies of the institutions with which they are affiliated.
There are no comments or no comments have been made public for this article.