Basin Futures · Explorable models

How GR4J works

GR4J is the rainfall–runoff engine inside most Basin Futures models: two water stores and four numbers — x1…x4 — turn daily rain into river flow. This page runs it live on a real catchment: the upper Goulburn River at Dohertys, in the Victorian high country, against fifteen years of daily gauge record.

Everything on this page is the live model, re-simulated in your browser as you drag. Each section widens the view: a single day, a full history, a family of behaviours, the whole parameter landscape.

Ground · one day inside the model

Water enters as rain P, leaves as evapotranspiration AET and river flow Q. Between them sit two tanks: a production store (soil moisture, capacity x1) and a routing store (slow drainage, capacity x3). Press ▶ and watch the tanks breathe through the seasons — or drag anywhere on the charts to scrub time. On the third strip the shaded band is the satellite-derived credible range for actual ET (TerraClimate and GLEAM) — Basin Futures uses the same envelope to constrain calibration. The window chips zoom the whole instrument in: the record includes the Millennium Drought and the 2010–12 La Niña floods, and each window rescales the axes to its own flows — drought baseflow is invisible at flood scale. In the September 2010 close-up you can watch the day-by-day lag between a storm and its flood peak — that lag is x4.

Live model — Goulburn at Dohertys

Goulburn River at Dohertys
694 km² of forested high country — the headwater reach of the Basin Futures Goulburn case-study model (20 reaches, shown grey).
Gauge 405219 · record 1967–2024 · forcing MSWEP v2 rain + CRU PET, catchment-averaged · window 2000–2014
rain P 0.0 demand PET 0.0 AET 0.0 production store soil moisture · cap x1 S 0 Pn−Ps perc 0.0 Pr 0.0 90% 10% UH1 · x4 days UH2 · 2·x4 routing store cap x3 R 0 exchange F 0.0 groundwater · x2 Qd 0.0 Qr 0.0 Q 0.00 mm/d
window
production store S / x1
routing store R / x3
monthly actual ET (solid) vs PET demand (dashed) · grey band: satellite AET envelope (TerraClimate–GLEAM)
gauge 405219 (observed) your simulation rain 2000 = warm-up (unscored)

Rung 1 & 2 · the four numbers, and families of rivers

The grey hydrograph is the real gauge record at Dohertys. Your job is the modeller's job: make the teal simulation match it. Drag a slider and the whole fifteen-year history re-simulates instantly. Press sweep on any parameter to widen the view — seven ghost rivers appear, one per value across its range, showing the family of behaviours that one number controls. When you want the answer, show calibrated fit jumps to the parameters Basin Futures' automated calibration found for this reach.

Parameters — drag to calibrate

NSE— KGE— 0.75+ is a good fit

Rung 3 · the parameter landscape

Calibration is a search over this landscape. Each pixel below is a complete fifteen-year simulation scored against the gauge — about 2,700 model runs, computed when you press the button. Bright ridges are parameter sets that reproduce the river; the marker is where your sliders sit now. Click anywhere to step back down: the sliders, hydrograph and tanks above all jump to that point.

NSE over x1 × x3 (x2, x4 held at slider values)

x1 = 20production capacity x1 (log) →1500 mm

~2,700 full simulations.