Product
HydroAgent-FF
Agentic workflow for flood forecasting: organized, reviewable, reproducible.
What It Gives You
Organized, reviewable, reproducible, with experts in control.
An agentic workflow product for flood forecasting operations. Climate change is driving more extreme floods, and forecasting is the first line of defense. Operational forecasting today starts from a hydrological model, but the final bulletin is never the model's raw output. Experienced forecasters remain responsible for review and release decisions, combining rainfall and water-regime information with local experience before the forecast is revised and released.
HydroAgent-FF is not about letting an LLM replace the hydrological model, nor about automating the forecaster away. It organizes real flood forecasting work into a structured, reviewable, and reproducible process, while keeping forecasters in control.
It works alongside your existing hydrological models and forecasters to make the whole forecasting effort more organized, easier to review, and easier to reproduce. Expert review stays built in, so the people responsible confirm the key results and keep full control of what gets released.
What HydroAgent Does
A workflow layer that keeps people in charge.
Brings scattered forecasting information into one organized, reviewable view.
Keeps expert judgment visible and reusable, instead of locked in individual experience.
Helps teams revise, compare, and explain forecasts as conditions change.
Prepares clear, review-ready materials for the final decision and release.
Preserves a traceable record for review, training, and institutional learning.
Current Validation Signals
Specific enough to discuss, early enough to test with partners.
Real basin tested
Demonstrated end to end on a real river basin (South Yamhill River, Oregon, USA) using historical flood events, including a March 2022 event.
Independent validation
On independent validation events, the system reproduces observed flood peaks and volumes well.
Multi-LLM ready
It performs consistently across several leading large language models.
Peer-reviewed
Methods paper under submission; presented as an oral talk at EGU 2026.
Beyond accuracy
Evaluation looks beyond accuracy: whether the process is clearer, the judgment is reviewable, and the result is reproducible.
Pilot Scenarios
Worth discussing if any of these sound familiar.
Experience that is hard to record
You already run a hydrological model, but the forecast-revision process relies on experience and is hard to record.
Updates that are hard to track
Across repeated forecast updates, version changes and the reasons behind them are hard to track.
Release requires clearer review materials
You need clearer review materials and handoff notes before a bulletin is released.
Turning history into learning
You want to turn historical forecasting into training and review material.
HydroAgent-FF turns flood forecasting from “model output + verbal experience + manual release” into one organized, reviewable, and reproducible process. Experts stay in control throughout.
Next Step
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.