Events / Community
Where to meet us
HydroAgent-Lab shares its work through conferences, workshops, informal meetups, and open community initiatives. This page records the initiatives we take part in, where we have presented, and where you can meet us next.
Collaborations
Collaborative research projects
HydroTuring — A Conservation Benchmark for AI Hydrological Models
Deep learning models can now predict streamflow more accurately than calibrated conceptual models. Accuracy alone, however, does not tell us whether their predictions respect the conservation of mass, energy, and momentum. HydroTuring, led by Prof. Zhi Li, is developing a standardized benchmark to answer precisely that question.
Any model can participate—including deep learning, machine learning, process-based, and hybrid models—regardless of the programming language in which it is written. A Docker container and a lightweight adapter are all that is required.
Models run in sealed containers on test cases generated on the fly from random seeds and never stored, preventing them from memorizing the answers. Each test returns a pass-or-fail result together with an explanation of any failure.
HydroAgent-Lab is collaborating on the benchmark and contributing new tests. If you work on snow, groundwater, evapotranspiration, soil moisture, or channel routing, we invite you to contribute a physics-based test that hydrological models should pass.
Status — Version 0.1.0 includes seven mass-conservation tests. Tests for energy conservation, momentum conservation, extrapolation, and real-world data are currently in development.
Write a probe · Propose a model · Join the paper
Events
Where we have been
EGU General Assembly 2026 · Vienna
Baoying Shan presented “Is it ready to apply Large Language Models to frontline hydro practice?” on behalf of HydroAgent-Lab. The talk described a human-in-the-loop approach to applying large language models in flood forecasting, rather than a chatbot. Several members (Qingyi, Shunan, Tiantian, xiaohuan) were on site too.
AI Builders Meetup · Huilongguan, Beijing
Siqian Qiu joined the OpenClaw community for an in-person exchange on agentic AI in water. The discussion covered how HydroAgent handles never-before-seen extreme floods, why it should provide a usable risk floor instead of promising a perfect peak, how forecasters' experience can be captured and reused, and why first-hand operational data matters for domain AI.
Tsinghua University · Beijing
Prof. Xudong Zhou (founder of the Hydro90 community, which several HydroAgent-Lab members come from) presented Hydro90 at Tsinghua University and used HydroAgent as a case study, introducing the human-in-the-loop approach to students and young builders.
Start a focused discussion about product fit, workflow design, or research collaboration.
HydroAgent-Lab works with institutions, forecasting teams, and research partners that need operationally credible hydrologic systems.