The Lake That Won't Come Back
Boardhysterezeshallow lake
This is a lake with one dial: how much phosphorus runs into it from the land around it. Phosphorus feeds algae, and the indicator below tells you whether the water is clear or turbid. Turn the dial up slowly and watch what the lake does.
● draft, deployed. Model parameters are pedagogical, not calibrated to a real lake.
The lake instrument
Picture of the lake. Its colour and speckle follow the phosphorus level; the state is also given in the readout.
- Water
- Loading L
- Phosphorus P
- Model time
Arrow keys change it by 0.01.
Before you look
You're the new watershed manager. You cut phosphorus loading to 0.50, below the level where the lake flipped. What will the water look like once it settles?
What the model did
The lake does come back, but not where it left. Keep lowering the dial and find the second threshold. The dial is yours again.
Why the way back is different
Phosphorus leaves the water by settling into the sediment. But once the water is rich enough, the sediment starts releasing phosphorus back, and the more there is, the more it releases. That feedback is weak in a clear lake and strong in a turbid one.
So between two loadings there are two lakes that can each hold themselves in place: a clear one where settling wins, and a turbid one where recycling keeps the phosphorus up. Which one you get depends on where you came from. The diagram shows every resting level the model allows at each loading. Solid lines are states the lake returns to after a nudge; the dashed line is the boundary between them.
Going up, the clear branch ends at and the lake has nowhere to go but up. Coming down, the turbid branch holds until . Your path is drawn on top.
| Resting level P | Kind | Water |
|---|
Show the equation
dP/dt = L − s·P + r·Pq / (mq + Pq)
Resting states solve dP/dt = 0. A resting state is stable when the slope of the right-hand side is negative there. A fold is where a stable and an unstable state meet and vanish, which happens where dP/dt = 0 and its slope is zero together.
| Symbol | Meaning | Value | Source |
|---|
Where else this shape might appear
Each case names what carries over and where the mapping breaks. Tags say how much evidence stands behind the mapping, not behind the lake model.
Where the lake stops being a guide
Evidence
The claim, model notes and limitations open after you commit or skip, so they don't give the answer away. Version and sources are below.
Claim
Status: Tests:
Model
dP/dt = L − s·P + r·Pq / (mq + Pq)
Integrated with fourth-order Runge-Kutta at a fixed step of 0.01 model time units, independent of frame rate. No randomness in the dynamics. The speckle in the lake picture uses a fixed seed (1) and is decoration only.
Limitations
- Parameters are chosen to show the mechanism, not fitted to any lake. Real lakes add vegetation, fish, wind, and seasons.
- With these values the lake recovers below a loading of about 0.39. Some real lakes are thought to have a lower threshold below zero, which this version does not show.
- The equation form follows Carpenter, Ludwig & Brock (1999) from memory; it has not yet been checked line by line against the paper.
- Not yet completed by a reader outside the author. Not yet deployed.
Record
Version , status , frame .
Sources
Errata
No corrections yet for version .
Found a mistake? Open an issue in the project repository, or write to the author. Corrections ship as a new version that names what it replaces; nothing is edited silently.
Your record
Your predictions stay in this browser. Nothing is sent anywhere, and there is no account.