The upcoming release of EPA's Lead and Copper Rule is top of mind for many in the drinking water industry. As part of its rule-making process, EPA received 80,000 public comments on its proposed rule. Using natural language processing (performing data science on textual data), BlueConduit, in collaboration with the Environmental Policy Innovation Center, explored the content and tone of those comments.

This provided an opportunity to understand what topics were important to stakeholder groups and how they communicated those messages as part of their submitted comments. By employing the techniques of topic analysis and sentiment analysis, we were able to identify the principle themes of different types of comments and the speakers' feeling (on a positive/negative spectrum). analyzing what different types of commenters had to say about the proposed regulations.

Machine learning found that service line replacement was a more frequent and detectable theme of public input than enhanced testing and sampling of drinking water.
Concern over children's safety is a central theme in comments and attachments.
The analysis found little to no evidence that financial themes were central in both comments and attachments. Additionally, there was no evidence of negative sentiments related to “unrealistic” financial demands, suggesting that the federal government should not exaggerate cost considerations in its final rule, compared to the draft.
Topic analysis found detectable differences between different commenter types, but that those differences do not always correspond to differences of opinion.

The insights that this type of analysis highlights for the Proposed Lead and Copper Rule Revisions show the use of these tools for government agencies in processing large volume of comments. The full report offers a deeper exploration of those opportunities.

Download PDF • 8.09MB

Updated: Sep 29, 2020

In light of the importance of lead service line inventories expected in the final Lead and Copper Rule Revisions, BlueConduit partnered with ASDWA to develop a white paper that outlines important considerations for state regulators and utility leadership when using statistical and predictive methods for LSL inventory and replacement.

Substantial uncertainty still surrounds the nation’s water systems regarding the number and locations of lead service lines (LSLs). The kind of uncertainty that the LSL question presents is well-suited for data science methods that have evolved in recent years. Given the significant public health, regulatory, and financial implications of these decisions, it is essential that regulators and utilities be aware of and adhere to some fundamental statistical methods when using predictive methods to inform SL work.

Download • 805KB

The paper includes five guiding principles for using data science to better characterize uncertainty LSL inventories:

  1. Clean data management and organization;

  2. Not accepting all historical records as truth;

  3. Conducting a representative randomized sample of service lines;

  4. Transparency in public outreach and reproducibility; and

  5. Accuracy on held-out sample.

These principles can be used by regulators to encourage water systems to plan strategically, make data-driven decisions, set budgets and requests for funds, build capacity in some skill areas, communicate with the public and build trust, and, most importantly, continue to protect the health of all individuals in the system.

The white paper draws on the BlueConduit team's experience in Flint where their statistical machine learning algorithm was used to guide pipe replacements.

For more information about BlueConduit and how we work with communities to reduce uncertainty around LSL inventory and replacement, connect with us at You can also find us on Twitter at @BlueConduitAI

U-M Office of Technology Transfer is celebrating the 31 start-ups that it helped launch this year. They highlighted BlueConduit as one of this year's notable companies.

U-M inventors went to market with a wide range of discoveries, including those from a company using machine learning predictive modeling to help cities like Flint replace their lead-tainted water pipes to another that pivoted from prostate cancer screening to rapid COVID-19 testing during a global health crisis.
  • LinkedIn
  • Twitter

2020 | Abernethy Schwartz Partners LLC