U of A-Affiliated Hydrologic Modeling Project Selected for DOE Genesis Mission
University of Arizona researchers are helping design a new large-scale hydrologic model driven by AI.
Hydrologic models help describe and predict how water moves through the environment by simulating the physical interactions of plants, land, water, and people. Communities and industries use these models to prepare for droughts and floods, and to understand their changing water availability. Laura Condon, a professor of hydrology and atmospheric sciences and interim director of Biosphere 2, uses “physically based equations” to create models that can answer questions about how a certain stream or groundwater system might behave under various conditions.
Most non-hydrologists think a river is one thing and groundwater is another, explained Condon. But in fact, rivers and the earth and water beneath them are all “one connected, continuous system.” Modelling the complexly interrelated physical processes of this system demands large amounts of data and computational power.
Condon is partnering with the Argonne National Laboratory to lighten the computational load of hydrologic simulations using AI, for a “smarter, more flexible forecasting system.” The Argonne-led project will use a technique known as “multi-fidelity fusion” to blend continental, regional, and local data, including human impacts and real-time observations.
“Rather than running our physically based model at every time step,” she said, “the AI model will be learning how our model behaves ... so that it can generate predictions with a very limited number of data points.”
Lowering the computational intensity of a model doesn’t only mean it can produce answers faster—it also means the model can run more comprehensive simulations that take wider sets of variables into account. “There’s a whole bunch of what-ifs that can happen along the way,” said Condon. “What if we get a different kind of storm? What if we divert water differently? What if the plants did something different?” A multi-fidelity fusion approach will be able to address these questions, while using less computer power.
The project’s first phase is funded through the US Department of Energy’s (DOE) Genesis Mission, which President Trump established via Executive Order in November 2025. The mission’s aim, according to a July press release, is to support “AI-enabled scientific discovery,” particularly in the realms of national security and energy infrastructure.
The DOE announced its first round of Genesis Mission awards on July 22, 2026, in Washington DC. The 278 awards were distributed across 342 participating institutions, including national laboratories, companies, nonprofit organizations, and universities—168 of the awarded projects were university-led. Five U of A affiliated projects, including Condon’s, received funding through the program.
Condon’s team is getting started right away. The Genesis grant projects are “really intended to be a sprint,” she explained. In Phase One, she and her team will be preparing a pilot version of the model. She anticipates seeing the model operational on a national scale in Phase Two. “We have only about nine months—six months for some of our targets—to show that we’re making adequate progress and to write the next proposal to get a Phase Two grant.”