The Early Warning System You've Never Heard of is About to Get a Lot Smarter

Dr. Adam Gushgari, Senior Director of Emerging Contaminants, Eurofins Environment Testing USA

I distinctly remember the first time I was introduced to the power of modern generative AI. I had asked one of my employees to help me draft a press release we were planning, and while I normally would have written it myself, time got away from me that day. After an hour or so, she brought me one of the most polished, well-crafted drafts I'd seen in years. Eloquently written, brief yet encompassing everything we wanted to relay. I laid the praise on thick. I told her that she had a genuine gift, that I was blown away by what she'd produced. Her response?

"I used ChatGPT!"

I was astounded. I had never used an AI model before myself, and couldn't believe the quality of work she was able to produce with it. This was back in 2023, and nearly three years later, generative AI has reshaped the landscape of virtually every profession and discipline in existence. What is absolutely wild to me is that despite the extraordinary disruption generative AI has already caused, and despite attracting hundreds of billions of dollars in venture capital funding, we are still watching a technology truly in its infancy. Individuals have used it to become dramatically more productive. Companies have used it to sharpen their competitive edge, overhaul marketing strategies, make sense of vast and complex customer datasets, and refine sales strategies. The CEOs of OpenAI and Anthropic are reportedly running a betting pool on when the first billion-dollar, single-employee company will emerge, with the leading guess being 2026. And despite all of this, I think a strong argument can be made that everything we have accomplished with these tools to date represents only a sliver of what they are ultimately capable of.

AI'S ROLE IN ENVIRONMENTAL DATA

Across the environmental sciences, researchers and practitioners of all kinds have begun exploring how artificial intelligence, machine learning, and digital twin technologies can sharpen both the data we generate and the systems we use to act on it. Several converging realities are driving this shift. Environmental science datasets are inherently complex and subject to significant noise, which can be partially offset by automating data analysis tasks that have become routine in the field. An LLM can analyze in seconds what might take a data scientist hours or days to do manually, and it can handle incredibly large and complex datasets, surfacing correlations that would otherwise be extraordinarily difficult to identify. This improves both immediate project costs and the downstream utility of the data itself. There is also the matter of transparency. Explainable AI, or XAI, addresses this directly by making the reasoning process of AI models visible and auditable rather than treating outputs as conclusions to be accepted on faith. The approach has gained significant traction in healthcare AI applications, likely due to the liability stakes around errors in clinical settings, but its relevance to environmental science is equally compelling for the same reasons. Ultimately, the work we do in environmental science exists in service of public health, and the stakes are no less significant for it.

So what does all of this mean for wastewater surveillance, "the early warning system you've never heard of?" If you've followed my work at all, you've almost certainly heard of it, but I needed a hook to draw you in. Forgive me.

There have been a number of interesting developments in the application of AI models specifically to wastewater surveillance, and the rationale is straightforward. These datasets are inherently complex and growing more so, but an added layer of complexity arises from the fact that wastewater surveillance data was never meant to be viewed in a vacuum. It is designed to augment traditional methods of public health data collection, not replace them. That integrative purpose makes artificial intelligence, machine learning, and digital twin technologies a logical next step for the field. At Eurofins Environment Testing USA, our wastewater surveillance programs span a broad and growing domestic and international footprint, generating the kind of large, methodologically consistent datasets that AI-driven analysis is best positioned to work with.

PUTTING AI TO WORK IN WASTEWATER SURVEILLANCE

One area where these tools could deliver remarkable value is the back-calculation of analyte degradation rates within the sewer system. The field has broadly acknowledged that in situ degradation affects virtually every wastewater surveillance dataset. Our approach to handling this has generally fallen into one of two camps: ignore it entirely, or apply a basic first-order decay equation to estimate the originating concentration. The problem with the latter is that temperature-induced degradation and enzymatic hydrolysis can vary dramatically based on the physical characteristics and microbial communities unique to each wastewater system. The kinetics of one system simply do not translate to another, and the challenge compounds when you consider that individual target analytes may behave very differently across varying environmental conditions. This is not a theoretical concern. It touches every biological target we monitor via RT-qPCR and ddPCR, and every chemical target we quantify through LC-MS/MS, whether we are tracking a pathogen, a pharmaceutical, or an emerging contaminant of concern.

A combination of artificial intelligence and digital twin technologies could change this entirely, enabling the real-time calculation of sewershed-specific degradation rates. What that would ultimately achieve is something significant: the same analytical certainty we associate with building-level or neighborhood-level sampling programs, but applied at the treatment plant scale for an entire city's population.

Another major application is data analysis itself. Wastewater surveillance is arguably the closest thing we currently have to real-time public health tracking, but a significant amount of value remains locked until we can move the technique from retrospective analysis to predictive modeling. Artificial intelligence is not just valuable in that transition, it is essential, driven by both the speed at which it can operate and the cost savings realized by reducing the personnel hours currently devoted to manual data review and modeling. The analytical diversity of a modern wastewater surveillance program, spanning targeted quantitation, non-targeted analysis, and biological surveillance across hundreds of sites, generates a volume and complexity of data that manual review simply cannot keep pace with. Without AI, the scalability of wastewater surveillance as a mainstream public health tool is genuinely in question.

A third application worth highlighting is explainable AI, and the same value it brings to healthcare settings applies directly to wastewater surveillance. Historically, one of the most persistent barriers to adoption has been end user utility. This data was developed specifically for public health and healthcare professionals, but those professionals are not accustomed to working with environmental datasets. That disconnect creates a meaningful gap between what the data can tell us and what decision-makers are actually able to do with it, and it remains one of the largest obstacles to broader adoption of the field. Explainable AI platforms built around wastewater surveillance analytics could close that gap considerably, enabling end users to leverage the data independently and confidently in pursuit of their own public health objectives. Greater utility drives greater demand, and greater demand drives the field forward.

TOWARDS THE GRAND VISION

The vision outlined here is ambitious, and none of it will be achieved overnight. But the foundation is already being laid. Eurofins Environment Testing brings to this space not only the analytical infrastructure, from RT-qPCR and ddPCR to LC-MS/MS and non-targeted analysis, but a domestic and international laboratory network capable of generating the geographically diverse, high-volume datasets that meaningful AI development in this space will require. What is essential now is the continued expansion of wastewater surveillance programs and the continued push for greater analytical breadth and depth across both chemical and biological monitoring. Only by doing so will we develop the datasets necessary to maximize the value of artificial intelligence within wastewater surveillance, and move the field closer to its grand vision – a true, global public health observatory.