The explosion of PFAS testing and sampling in the past several years has led to increased urgency around source identification. Our technical teams have been fielding more questions around this topic than ever before, with some common themes emerging:
Where do PFAS come from?
Can they come from multiple sources?
Who should be responsible for removing them?
How long have PFAS existed in our community?
I don’t live near any factories; why are the PFAS levels in my community so high?
Borrowing from the field of forensics, environmental scientists have the ability to identify contamination events and determine their origins. It is useful in both remediation and litigation and involves all aspects of site characterization. Not surprisingly, fingerprinting, one of the most widely known tools in a forensics toolkit, has become increasingly relevant to the study of PFAS.
There are more than 4,000 PFAS globally, a seemingly insurmountable data set for any detective following contamination trails. To make matters worse, by the time communities report a positive identification of PFAS in mixed-use plumes, separation and identification becomes time consuming and potentially cost-prohibitive. Fortunately, sophisticated techniques like machine-learning and ever-expanding libraries of data have enabled environmental scientists to trace contaminants like never before.
As with any valid fingerprinting approach, there must be a comparison or suite of chemicals to evaluate against. When fingerprinting petroleum or petrochemicals, which are the most documented, petroleum sources, formulation, refining, alterations and degradations each play a key role in selecting a reliable forensic strategy that reflects complexities in the field. In the world of PFAS, however, simply identifying the best comparison suite can be a significant endeavor on its own.
Large, publicly available datasets, like California's GeoTracker, can help isolate relevant comparison groups. Geotracker is the state's data management system for sites that impact groundwater or have the potential to impact groundwater. At CDM Smith, our scientists and engineers have been working behind the scenes to build a robust and accurate PFAS metadata library by combining tools like GeoTracker with machine learning. Machine learning is a subcategory of artificial intelligence that enables computers to “learn” from large datasets and improve from experience without having to be specifically programmed. The models and tools developed using this approach can provide predictive analytics and support decision-making for site management of mixed contaminants.
In addition to active research initiatives conducted at CDM Smith laboratories, we have a number of ongoing projects in the field focused on the intersection of PFAS and AI. For more information reach out to email@example.com.
The future of PFAS is about options: optimizing upstream treatment technology, reducing downstream waste generation, and destroying PFAS.
Join CDM Smith at this year's virtual summit on PFAS, sustainability and water-borne pathogens.