The Water Heater That Walked in the Quake

Why this matters

A tank water heater full of water is one of the heaviest single objects in a house, and it sits on a gas line. In an earthquake it does not fall over so much as walk: it rocks, its base slides, and every rigid thing connected to it takes the movement. The gas leak that follows is the reason seismic regions require straps at all. This case is a heater that had a strap and still leaked, and the interesting part is what each detail of the install contributed, because a reader who learns only "strap it" will strap it wrong.

First, the part that is not diagnosis

A moderate earthquake hit mid-afternoon. An hour later the homeowner walked into the garage and smelled gas. She did the right things without being told: left the garage door open, touched no switches, took the family out to the sidewalk, and called the gas utility from there. The utility's responder closed the meter, found the leak at the water heater with a detector and leak solution, capped the heater's gas supply at the appliance shutoff, and left the meter locked off with a tag: the house would be relit by the utility after a repair and a pressure test.

When the shop's tech arrived the next morning, the meter was off and tagged, and he left it that way. He confirmed with a combustible gas detector at the garage door before going in that there was no residual reading, opened the door fully, and did not operate the garage door opener or any light switch until the detector stayed silent at the heater itself. A gas leak after a quake is never only at the place it was found; if the detector had read anywhere else in the house, the call would have stopped and gone back to the utility.

What he found

A 50 gal gas water heater on an 18 in platform in the garage, the platform there because an appliance with an ignition source in a garage has to sit up off the floor. The heater was about 60 in tall. One strap, a band of light-gauge metal tape, wrapped it at about 30 in above its base and was screwed into a wall stud. The gas supply came down the wall as black iron pipe, through the appliance shutoff, and threaded directly into the gas control on the heater: rigid from the wall to the valve. The vent connector rose from the draft hood to the flue.

After the quake, the heater's base had slid about 3 in out from the wall and rotated a few degrees on the platform. The strap was intact and still screwed to the stud. The vent connector had pulled off the draft hood and was hanging an inch clear of it. The water connections were braided flexible connectors and had not leaked.

Four explanations, killed or kept on evidence

The strap failed. The first thought on seeing a heater out of position: the strap broke or pulled its screws. It had not. The band was intact, the screws were tight in the stud, and the band had not stretched. So the strap held, and the heater moved anyway. That moved the question from "did the restraint fail" to "what was the restraint actually restraining".

The answer was geometry. A heater held at one point at mid-height is free to pivot about that point. Its full mass, roughly 417 lb of water (50 gal at about 8.34 lb per gal) plus the tank, is shaken at the base by the floor, and with nothing holding the bottom, the base walks while the strap acts as a hinge. The 3 in slide at the base and the slight rotation are exactly what a single mid-height point allows. Kept, but as a contributor, not the cause of the leak.

The leak was in the house piping, and the heater was a coincidence. Plausible after a quake, since any joint in the system can loosen. The utility had answered this before the shop arrived: with the heater shutoff closed and capped, the rest of the system held pressure. And the leak solution had bubbled at one place: the threaded joint where the rigid supply entered the heater's gas control. Killed.

The gas control itself cracked. Possible, and serious, because a cracked control is a replacement. With the gas off and the meter locked, the tech unthreaded the supply from the control. The nipple's threads were distorted on one side, and the control's inlet showed a small witness mark where the nipple had been levered against it, but no crack in the body. The control was replaced anyway, because the manufacturer treats a control that has taken mechanical stress at its inlet as suspect, and a gas control is not a part to give the benefit of the doubt. Kept as damage, not as the origin.

The rigid connection turned the movement into a leak. This was the cause. Pipe threaded straight into the valve cannot take 3 in of base movement. As the base slid and the tank rotated about the strap, the pipe acted as a lever on the joint at the control. Threaded joints seal by interference, and a bending load opens one side of the thread enough to leak. A listed flexible appliance connector exists to take exactly that movement, and it would have bent instead of the joint.

What each detail contributed

Laid out the way the tech wrote it on the ticket, so the homeowner could follow the chain:

• One strap, at the wrong height. IRC M1307.2, as adopted by your jurisdiction, requires heaters in Seismic Design Categories D0, D1 and D2 (and townhouses in C) to be anchored or strapped in the upper third and the lower third of their height, with the lower point at least 4 in above the controls. California reaches the same result through its plumbing code and a state statute requiring water heater bracing. On this 60 in heater, the upper third is 40 to 60 in and the lower third is 0 to 20 in. The controls sat at 10 in, so the lower strap belonged between 14 and 20 in. The single strap at 30 in sat in the middle third, in neither zone. Contribution: it let the base walk.
• The rigid gas supply. Contribution: it converted the walk into a leak.
• The platform. It raised the heater's center of mass by 18 in, which increases the rocking a floor motion produces. It was required, and correct; it just made the strap placement matter more.
• The vent connector, held by friction. It came off the draft hood. Nobody noticed because the gas was already off, but had the leak not happened and the heater kept firing, the flue products would have spilled into the garage. Contribution: a carbon monoxide hazard that was masked by a louder one.
• The braided water connectors. They took the same movement without leaking. Contribution: the proof, in the same install, that flexible connections do what they are for.

The repair, and putting gas back

The repair was the detail list inverted:

1. A listed seismic strap kit, two straps, the upper at 48 in and the lower at 16 in on this heater, each fastened into studs with the kit's fasteners and with blocking where a strap fell between studs. The heater pushed back against the wall before tightening, with the kit's spacer filling the gap so the tank could not build momentum across it.
2. The rigid supply cut back above the appliance shutoff, a sediment trap installed as the fuel gas code (the IRC fuel gas chapter or NFPA 54, as adopted) requires at the appliance, and a listed flexible appliance connector of a length that reaches without stretch or kinking.
3. A new gas control, installed to the manufacturer's procedure.
4. The vent connector re-seated on the draft hood and fastened with screws at the hood and at each joint, per the vent manufacturer.
5. The utility called to relight. Before the meter was turned back on, every joint the tech had made was ready for a leak test; once the utility restored gas and the system held pressure, the tech leak-tested each joint with solution, lit the heater per its instructions, and stood beside it, not in front of it, through the first firing. Draft established at the hood within the first minute, measured with a smoke source; flue gas CO within the manufacturer's air-free limit; a personal CO monitor on the tech's collar showed no rise above background in the garage during a 15 minute run.

Only after all of that did he tell the homeowner the heater was back in service.

How he confirmed the reconstruction

A reconstruction is only a story until something checks it. Two things did. Before the repair, with the gas off, the tank drained to a floor drain through a hose (the water had cooled overnight, but he still ran the hose before opening the drain valve and kept his hands off the outlet), and the vent connector already off, rocking the heater by hand against the old single strap reproduced the pivot: the base moved, the top barely did. And the distorted threads sat on the side of the nipple facing away from the direction the base had slid, which is where a bending load from that movement would open the joint. Both pointed the same way.

What changes the call on the next one

A heater in a lower seismic category may carry no strapping requirement at all, and the rigid connection is then a fuel gas code question rather than a seismic one; in many jurisdictions a flexible connector is still the expected method, and in some, rigid piping to the appliance is allowed. Some cities also require a seismic gas shutoff valve on the service, which would have closed the meter at the moment of shaking. See related: Seismic Gas Shutoff and Excess Flow Valves. None of that changes the lesson here: a strap in the wrong place restrains the wrong thing, and a connection that cannot move will leak when the appliance does.

References

• International Residential Code M1307.2 (anchorage of appliances), as adopted by your jurisdiction, in the edition it adopted.
• Your jurisdiction's adopted fuel gas code (the IRC fuel gas chapter or NFPA 54) for appliance connectors and sediment traps.
• California Health and Safety Code provisions on water heater bracing, for work in California.
• See related: How to Strap a Water Heater for Earthquakes; Seismic Gas Shutoff and Excess Flow Valves; Anchoring Equipment and Ducts in Seismic Regions.