A skeptical embrace: Using AI to understand changing rivers and lakes and improve decision making

2026 Harlow Summer Seminar Series: June 25, 2026

Talk Title: A skeptical embrace: Using AI to understand changing rivers and lakes and improve decision making
Speakers: Dr. Matt Ross, Associate Professor of watershed science in the Department of Ecosystem Science and Sustainability at Colorado State University

We will host a BBQ at 5:30 followed by the seminar at 6:30, with a $10.00 suggested donation for attending the event. The event is open to the public and reservations are not required.

Attend Virtually: The talk will also be live on Zoom. The zoom link will be made available through our email list. If you haven’t already, join our mailing list!

Abstract

Artificial Intelligence is, and will continue to be, a revolutionary force in how we conduct science, teach, share information, and understand the world. In the midst of such a profound transformation, it is vital to embrace these tools—but to do so with a refined skepticism regarding their outputs, design, and potential for corruption. In this talk, I will focus on how my research group has skeptically embraced AI and its humbler cousin, Machine Learning (ML), to build a bedrock of datasets, tools, and philosophies that help us better understand changing rivers and lakes  and guide real-world decision-making.

The bedrock of AI and ML in the environmental sciences relies on benchmark datasets: shared resources where we agree on a foundational "truth" for our models to emulate. To this end, we have built several open, discoverable datasets, including AquaSat for remote sensing of water quality and MacroSheds for small watershed biogeochemistry. These datasets provide the basis for models that allow us to peer into the history of water quality change and make continental-scale discoveries across our freshwater landscapes, from river sediment dynamics to lake algae blooms.

Crucially, these benchmark datasets and research models also serve as the foundation for our engaged work. We take these baseline models and refine them to meet the specific needs of water resource partners in Northern Colorado. By building forecasting tools that directly enable both short-term water treatment plant operations and long-term planning, we ensure our models solve actual problems. We bring our partners directly into the model design process, building trust through education and highlighting the importance of rigorous testing, such as spatial and temporal holdouts.

Lastly, I will discuss how our projects—and our overarching approach to science—will evolve as AI-based systems directly replace our own coding. This "post-code" world requires even greater degrees of subject-area expertise and careful review. However, if embraced thoughtfully, it is a shift that is already increasing both the quantity and quality of our work.

Bio

I grew up in Colorado  and  went to CU-Boulder where I studied Ecology and French. After a brief stint as a teacher and translator in France, I went to grad school at Duke University where I studied watershed biogeochemistry, and the impacts of mining on water quality. I returned to Colorado at Colorado State University, where I have built a lab that uses data science techniques to understand freshwater change and to use that understanding to drive water resource decision making.