Title: AutoVL: A Nash-Equilibrium Engine for Intelligent Streamflow Separation 

Speaker: Professor Vincent Lyne

Brief Introduction to the Report

Rivers flow in pulses—quick bursts of surface flow from rainfall and slow releases from stored groundwater. For a century, hydrologists have struggled to separate these two rhythms objectively. AutoVL (Automatic Variable Leak - Lyne (2025)) offers a breakthrough: a self-learning algorithm that divides total streamflow into fast and slow components without any user-parameters, manual tuning, or rainfall input. It works as a deconstruction-reconstruction “game” between two reconstruction-players—fastflow and slowflow—each adjusting collaboratively to the other until they reach equilibrium, much like a Nash-Equilibrium game in economics.

This simple yet powerful process opens a transparent window into how rivers assemble their flow from rain-driven bursts and groundwater-fed persistence. Applied to 467 Australian catchments from 1976–2021, AutoVL shows that traditional fixed-parameter filters (like the long-used Lyne–Hollick method with an uncalibrated parameter) have severely underestimated baseflow, particularly in drier, more variable rivers.

Across the continent, AutoVL finds a consistent ceiling near a baseflow-to-streamflow index of 0.6 in wetter regions, suggesting an upper bound to how much groundwater flows in streams. By adapting automatically to each catchment’s behaviour, AutoVL provides the first objective, reproducible global benchmark-standard for streamflow separation—offering new insight into water resilience, drought vulnerability, and the dynamic memory of rivers in a changing climate. 

Lyne, V. (2025). AutoVL: Automated streamflow separation for changing catchments and climate impact analysis. Journal of Hydrology X, 26(1 January 2025), 100195. https://doi.org/10.1016/j.hydroa.2024.100195

Brief Introduction to the Speaker

Professor Vincent Lyne is an internationally recognised ecological systems scientist, oceanographer, earth systems scientist, and modeller with more than five decades of experience in environmental modelling and planning. He is best known for developing the Lyne–Hollick filter, a cornerstone method in global hydrology for streamflow and baseflow separation, and for creating its modern successor, AutoVL, an adaptive algorithm that has redefined objective hydrological analysis.

Host: Associative Professor FU Dongjie, Associate Professor YAN Fengqin

Time: 10:00 – 12:00am Octorber 17, 2025

Venue: Meeting Room A847

Hosted by State Key Laboratory of Geographic Information Science and Technology