Automated Adaptive Baseflow Separation Across Australia's Catchments Using AutoVL

Abstract

Baseflow represents streamflow sustained between rainfall events and is fundamental to water security, ecosystem resilience, and drought management. However, widely used digital filters such as the Lyne–Hollick (LH) algorithm, when applied with a fixed, uncalibrated parameter, can substantially underestimate baseflow and obscure non‐stationary streamflow responses. Changes in climate and catchment conditions alter streamflow behavior and recession dynamics, yet many baseflow separation methods rely on fixed parameters, motivating the need for adaptive approaches. This study applied AutoVL, an automated self‐calibrating baseflow‐separation algorithm, based upon LH, to 467 Australian catchments spanning diverse hydroclimatic regimes over 1951–2021. AutoVL reconstructs streamflow components using a variable‐leak signal formulation coupled with a statistical recession model that iteratively estimates governing parameters from the discharge record. Variations in AutoVL‐derived flow diagnostics were evaluated relative to LH to assess systematic differences in flow partitioning across catchments. Compared with the conventional LH configuration ( = 0.925 with three passes), AutoVL consistently identified larger baseflow contributions, often nearly doubling Baseflow Index (BFI) where LH yielded BFI < 0.5. Hydrological variability was examined using clustering of interannual streamflow as a similarity construct to group catchments with comparable temporal behavior. These clusters delineated distinct streamflow regimes, from stable perennial systems to highly variable flood–drought responses and shifts in runoff efficiency. Systematic variation in AutoVL parameters across these regimes demonstrated the method's capacity to adapt to differing streamflow dynamics. Together, these results provide a framework for interpreting spatial and temporal variability in streamflow behavior and highlight the need for adaptive approaches under changing conditions.