It got to 24 degrees overnight and nothing happened. The bay sensor did its job, the alert went out at two in the morning, and by breakfast everything looks exactly as it did. Then at ten o’clock, with the sun on the side of the rig and the bay back up to 38, a fitting lets go and starts spraying.
Freeze damage almost never announces itself during the freeze. Ice expands, splits a fitting or a line, and then holds its own leak shut. It is the thaw that opens it, which is why the expensive hour is the middle of the following morning and why so many people find the water after they had decided they got away with it.
That timing is the useful part. Freeze protection is the article about not getting here. This one assumes the cold already happened, and the question is how much of the rig it costs you.
The history is a prediction
Look at the bay chart for the shape rather than the alert: an overnight minimum below 32°F, then a climb back through 34. Expect the leak at the climb, not at the minimum. If that shape is in last night’s history, this morning is the morning to be near the rig, or to have somebody near it.
It is not an alert and it should not be one — a threshold at 34°F rising would fire on every ordinary winter morning the rig never froze. It is a thing you read, on the mornings after a night the alert already told you about. Any of the ambient sensors give it to you at 60 to 120 seconds, and a sensor on the bay floor reads colder than the line it is standing under, which makes it usefully pessimistic.
What the history cannot say is whether anything actually broke. A dip to 31 for twenty minutes with heat tape on the line is usually survivable. It says the conditions happened, and it puts you in front of the readings below on the right day.
Catching the water
a. A puck where the water would land
The only direct reading, and the cheapest. A leak puck reports every 120 seconds, or 60 on a BTHome device, and three of them cover the places that actually split: under the pump, under the water heater, and at the low-point drains.
Each one sees one puddle. A split inside a wall cavity that runs down into the underbelly never touches any of them, and in winter the false positives are their own season — condensation on a cold bay floor, a wet dump hose put back in the bay, rain through a door seal. Where the pucks go is the longer argument.
b. Fresh against gray, which separates the two stories
This is the pair worth having and it is free on a coach with a Bluetooth SeeLevel board. Water leaving the fresh tank overnight with nobody aboard has two innocent explanations and one expensive one, and the gray tank tells you which.
Fresh falling and gray rising by roughly the same amount is water going down a drain — the tap you left dripping deliberately for freeze protection, or a toilet flapper that does not seal. Fresh falling and gray not moving at all is water going somewhere it should not. On a Mopeka the water reading is raw millimetres rather than a percentage until the water curve ships, which does not matter here: a falling number is still a fall.
c. The pump, which should be switched off anyway
A demand pump chasing a split runs and runs. On a shunt that is a sustained 4 to 8 amps above your night-time baseline rather than the short bursts of ordinary use, and if the fresh tank empties underneath it the pump runs dry and destroys itself — which is the failure that turns a plumbing repair into a plumbing repair plus a pump.
This method is silent by design if you did the right thing. A rig left in freezing weather should have the pump switched off at the panel, which makes the current reading useless and prevents the worst version of the damage. That is the correct trade: you lose a detection method and keep a floor. The pump as a leak detector is a year-round argument and it assumes the pump is on.
d. Line pressure, with a bad winter false positive
Most rigs hold 40 to 55 psi at rest, so pressure below 20 for more than a few polls means the system is not holding. The hub carries a generic water-pressure layout that reads every 120 seconds — a placeholder rather than a product we have verified, exactly as with the DC monitor, so treat an implausible number as the layout not matching rather than as a plumbing event.
The bigger problem is seasonal: campgrounds shut the spigot off in a hard freeze, and that reads identically to a frozen or split hose. In the one week this method is most relevant it is also least trustworthy.
The combinations
The pairs here are unusually decisive, because each one has an innocent twin that a single reading cannot rule out. This is the cookbook at its most useful.
- Bay minimum under 32 overnight, and the puck fires as it climbs past 34 — a fitting split in the freeze and let go on the thaw. Shut the pump and the city water. The split is at the coldest spot the history shows.
- Fresh falling, gray rising by about the same — the drip you left running, or a toilet that does not seal. Not a leak.
- Fresh falling, gray unchanged, nobody aboard — water going somewhere it should not. Get the pump off; a puck should confirm within a couple of minutes if it is anywhere near.
- Sustained pump current and pressure staying under 20 psi — the pump cannot build pressure. A burst, an open tap, or a dry tank; if fresh is also falling it is a burst, and if fresh is already at zero the pump is destroying itself now.
- A puck wet, all tank levels unchanged, and no pump current — not the plumbing. Rain through a bay seal, condensation, or a wet hose put away.
What none of this tells you
Where the split is. Every method here says water is going somewhere; not one of them says which fitting. A puck reports one puddle at one spot, tank senders are coarse and lag by hours, and a current reading names a load size rather than a load. The coldest point in the bay history is the best guess you get, and it is a guess.
Nothing here shuts a pump off or closes a valve either. Everything on this page is a record being written while the water runs.
What to do about it
If you are leaving a rig in freezing weather, the protection is physical and it is not us. Switch the pump off at the panel, disconnect and drain the city-water hose, and drain the low points. Do that and a split fitting costs you a fitting instead of a fresh tank on the floor and a burned-out pump. The sensors then become a record of what happened rather than the thing standing between you and it, and that is the honest description of what they are.
With that done, one leak puck under the pump is the highest-value thing left, because the pump area is where the fittings actually are. Add the fresh-versus-gray comparison if the coach already has a Bluetooth tank board, since it costs nothing and it is the only reading that separates a burst from the tap you left dripping on purpose.
And on the morning after a night that went below freezing, look at the chart before you go anywhere. That is a two-minute habit that puts you in the rig at ten o’clock rather than at six in the evening.
Where to get one
Pucks first, and more than one. Everything else on this page is a reading you already own or a habit that costs nothing. The compatibility database has what each device reports.
SwitchBot Water Leak Detector
A puck you drop under the wet bay, behind the water heater, or beside the fresh tank. It broadcasts leak state, tamper state, and battery.
We may earn a commission if you buy through this link. It does not change your price.
Garnet SeeLevel II 709-BTP3
The earlier SeeLevel Bluetooth board. It cycles one sender per broadcast rather than sending them together, so readings arrive staggered.
We may earn a commission if you buy through this link. It does not change your price.
Victron SmartShunt and BMV
Victron's Instant Readout broadcast. The battery monitors decode fully. The rest of the range is recognized, and that is all we will claim for now.
We may earn a commission if you buy through this link. It does not change your price.