Chaopeng Shen
Pennsylvania State University, PA·Earth & Environmental Science
When rivers run low during droughts, cities, farms and ecosystems all compete for the remaining water. Predicting how these low-flow periods will change as the climate warms requires global hydrologic models, which are computer simulations of how water moves through land, soil and rivers.
- Fields
Earth & Environmental Science
Engineering
Computer Science
- Topics
hydrology
differentiable modeling
differentiable programming
- Grants
- 1 known grant, through Aug 2027
What they work on
- Build differentiable, learnable process-based global hydrologic models that keep physical laws such as mass conservation while learning their equations and parameters from global data.
- Develop global parameter schemes for groundwater and address scale mismatches, so that the models can better reproduce river low flows and dry-season trends seen in observations.
- Measure how much prediction error comes from flaws in model structure, and improve the reliability of low-flow predictions in data-sparse regions.
- Use the improved models to estimate how global low flows will change under climate change, and share the results with climate impact assessors and non-profits in Africa.
More about the research
Today's models often struggle with dry periods. Some even predict trends opposite to what river gauges actually record. Machine learning models tend to be more accurate, but they are hard to interpret and do not necessarily obey physical rules such as conservation of mass, the principle that water cannot appear or vanish.
Why it matters
Better estimates of future river low flows could help communities and governments prepare for drought and plan climate adaptation. They could also help aid groups direct disaster relief to the right places. Improved predictions in places with few river measurements could especially help regions in Africa, where non-profit partners will receive the project's low-flow forecasts.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Stefano Galelli
Cornell University, NY·Earth & Environmental Science
Characterize the structural uncertainty that comes from choosing one dam operation model over another, using a computational framework applied to 18 large river basins spanning every major U.S. hydrologic region and about 200 dams.
Hoori Ajami
University of California-Riverside, CA·Earth & Environmental Science·Early-career
Measure water balance dynamics, water residence times and bedrock groundwater's contribution to streams in the Kaweah River watershed using hydrometric observations and hydrochemical tracers.
Nicoleta Cristea
University of Washington, WA·Earth & Environmental Science
Measure sap flux, soil moisture and snow across forest stands of different ages in western Washington to test how forest regrowth after clear-cut harvest affects summer low streamflows.
The school
Pennsylvania State University: 60 labs on PI Pages, $1.3B a year on research, 743 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.