How Flat Roof Waterproofing Fails and What Repair Reinstates

A built-up bituminous roof is assembled on the roof itself from successive reinforced sheets bonded to one another, with the laps of each set out so that they do not coincide with the laps below. A liquid-applied membrane does not exist until it is made: a primer conditions the substrate, liquid is applied, reinforcement fabric is laid into it and further liquid encapsulates the fabric. A single ply roof is laid from rolls and adjacent sheets are welded or bonded so that the finished layer is continuous across the field.

Each of those descriptions already contains its own failure mode. Where redundancy comes from a relationship between plies, the failure is the loss of that relationship. Where the layer exists only because it is bonded, the failure is adhesive. Where continuity is achieved at a seam, the seam is where continuity is lost. None of this can be read off a material name, which is why the records describe the assembly rather than the product.

The record is explicit that redundancy is not a property of any sheet. It exists only in the relationship between plies: how far the laps are offset, how completely each ply is bonded to the one below, and whether that bond is continuous enough to stop water entering one lap from travelling sideways until it finds a second weakness. Plies laid to the same setting-out align their laps, and the assembly quietly becomes as strong as a lone sheet along every one of those lines while still looking layered from above and in the specification.

Bonding between plies then decides what a defect can do. Where the plies are fully bonded, water entering a lap is trapped locally and stays close to where it entered. Where the bond is patchy, the same water travels between the layers and emerges somewhere with no relation to the entry point, which makes diagnosis and repair far harder. A partial bond that was designed as a partial bond is a considered decision about movement and venting; a partial bond that happened is a path.

With a liquid-applied membrane, the chain from substrate to primer to coating is what carries the load, and it is only as good as its weakest link. A sound coating over a poor primer, or a good primer over a friable surface, produces a layer that looks complete and can lift as a sheet later. That is why so much of the design and inspection effort on this type sits below the layer anyone can see, and why moisture in or on the substrate is the usual reason an applied layer fails to bond or blisters afterwards.

Reinforcement is the second half of the same story. The fabric gives the cured film the ability to spread small cracks opening in the substrate across the layer rather than letting them concentrate at one line, and where the fabric stops short of a junction the film is asked to take that movement unaided at precisely the point where the substrate is most likely to move. The conditions during application and cure become permanent properties, so an assembly whose coats cured out of sequence can behave as separate films rather than as the single layer it was intended to be.

The single ply record puts it plainly: with a lone waterproofing layer, continuity has nowhere to hide. A seam is not a lap over another ply, because at that line the seam is the waterproofing. The same is true at every upstand, corner, outlet and penetration, which is why detailing effort on this type concentrates at the perimeter and around openings rather than in the open field, and why the count of details rather than the roof area predicts the effort involved.

How the sheet is held down reaches all the way down the build-up and becomes part of the failure picture. Mechanical fastening drives fixings through everything between the membrane and the structure, so each layer passed on the way is perforated. Adhesion asks the layer immediately below to be sound, compatible and prepared. Ballasting asks the structure to accept weight and puts the layer that needs inspecting out of sight. Wind acts on the assembly rather than on the sheet, and it concentrates at edges and corners.

All three records converge on the same place. The bituminous record notes that an upstand carried up with fewer plies than the field, or a corner without its additional reinforcing piece, removes the redundancy exactly where movement, ponding and mechanical damage are concentrated, so the field of the roof ends up better protected than the parts that will be tested first. The liquid-applied record says the type is chosen because of details rather than in spite of them, and that its value disappears if the detail reinforcement is omitted. The single ply record locates the same difficulty at corners and outlets, where seams turn, cross other seams and change plane.

The published junction records say the same thing from the other side. An upstand raises the waterproofing above the level water can reach and then still has to end, and that top edge is where water can get behind the whole layer regardless of how high the upstand went. A membrane turning through a right angle is stressed hardest at that angle, which is why a tapered fillet turns one sharp change of direction into two shallower ones, and why its omission is usually discovered only through the failure it was there to prevent.

Layered weathering relocates water rather than releasing it where it entered. On a roof whose plies are patchily bonded, water travels between the layers. On an inverted roof, the record states that a leak rarely appears inside the building beneath the point where the membrane is damaged, because water travels along a concealed plane first. Where the waterproofing sits under paving, ballast or a terrace, the surface someone walks on says nothing at all about the layer actually resisting water.

That is why an investigation that starts at the stain is starting in the wrong place, and why the records treat the process of locating a fault beneath a covered surface as a question to ask before the surface goes on. Where the membrane is concealed, the available evidence is the record made before it was covered, and where no such record exists the search runs over the whole roof.

A repair is not finished when the hole is covered. On a built-up roof, the record treats reinstating the layering, including its offsets, as what restores the property that defines the assembly, so a patch laid over damage leaves a roof that is layered everywhere except at the point that failed. On a liquid-applied roof, a repair can be made by preparing an area and reapplying without lifting a sheet, but that advantage depends on knowing which product was used and whether a compatible material is still available. On a single ply roof, a repair reinstates a seam that stands alone.

Across all three, compatibility with existing waterproofing, sealants and adjacent materials has to be confirmed for the specific combination against the documentation belonging to the products involved, and that documentation is the governing reference rather than any general account. Which is why the most useful thing an owner can hold is not a description of the roof but a record of what was actually applied, over what, and in what conditions.