Roofs That Drain at a Level You Cannot See

An inverted roof puts the waterproofing directly on the structural deck and the insulation on top of it, so rain lands on a surface that is not the waterproofing, passes down through or around the thermal layer, and reaches the membrane from beneath the insulation. A trafficked deck carries a wear surface on supports or bedding above the waterproofing, so water passes through the joints into a void and meets a plane with its own falls. An attenuation roof holds water on purpose, in a void within the build-up or on its surface, and releases it slowly through a restricted outlet.

Pedestal paving is the same idea made explicit. The record says water arriving on the surface passes straight through the open joints into the space beneath and runs down the deck falls to the outlets, so the trafficked layer is not the layer that resists water at all; it is a wearing surface and a shading layer over the real one.

Both records state this as a common misunderstanding rather than a subtlety. The inverted record says the surfacing is not the waterproofing, and that a roof which looks sound underfoot tells an observer nothing about the layer actually keeping water out. The trafficked record says the paving is not the roof, and that the layer keeping water out is beneath it and is doing so while concealed and loaded.

That has a direct consequence for inspection. Walking a terrace establishes the condition of paving. It establishes nothing about the membrane, the upstands or the outlets, all of which sit under a layer that has to be lifted to reach them.

Drainage at the visible surface does not drain these roofs, because water that has passed into the void has to reach an outlet at the level below. The trafficked record describes what happens when the two levels disagree: if the surfacing sheds towards a perimeter while the waterproofing falls towards a central outlet, the roof has two drainage layouts working against each other rather than one working in layers.

The inverted record puts the same requirement on the outlet itself. If an outlet collects only at the surface, water that has already passed into the build-up never leaves it; if it collects only at membrane level, the surface has to be able to shed into it. The arrangement of the outlet, the depth of the insulation and the finished level of the surfacing are therefore settled as a single detail.

The trafficked record contains an observation worth holding onto. The level surface and the falling roof beneath it only coexist because the support layer varies in height across the roof, and that variation makes the void deepest where the roof is lowest, which is exactly where the outlets are. The deepest, least visible part of the assembly is therefore also the part that has to stay clear for the roof to drain.

It is also where anything washed through the joints ends up. The pedestal record notes that debris falls through the joints and collects where nobody looks, and that the first sign of a blocked outlet is often water appearing somewhere that has nothing to do with where the blockage is.

On an attenuation roof, water standing is the design working. The record says so explicitly, and adds that clearing the restriction to make the roof drain faster defeats the purpose. It also names the consequence of nobody knowing that: a future owner or contractor arriving at a roof holding water has no reason not to remove what looks like a blockage.

On an ordinary drained roof, a persistent pool reports something else, usually a plane set wrong or an obstructed outlet. And on an inverted or trafficked roof, a pool on the visible surface may report nothing at all about the membrane, because the two levels are separate. In each case the observation is the same and the meaning is different, which is why the arrangement has to be known before the pool is interpreted.

The inverted record states that a leak in this build-up rarely appears inside the building beneath the point where the membrane is damaged, because water travels along a concealed plane first. Damage traced by dampness inside cannot be located by walking the roof, because the leak is at a level that is not the surface.

That is what makes the investigation process a design question rather than a maintenance one. The record asks, before the roof is built, what the process would be for locating a fault beneath the surfacing, and how the waterproofing will be verified and recorded before it is covered. Where neither question was asked, the search begins by removing surfacing in areas chosen by guesswork.

On an inverted roof the surfacing is doing mechanical work. The record describes ballast as resisting wind from one direction and flotation from the other, because water standing on the waterproofing surrounds loose-laid boards and boards surrounded by water want to rise. Stripping surfacing from part of a roof to carry out unrelated work leaves the layers beneath restrained by nothing.

That is why the ability to open the surface is treated as a design provision rather than an operational matter. Which areas can be opened, how surfacing is returned afterwards without losing restraint, and how debris is removed from the void all belong in the design. A terrace laid as a continuous bonded field over its outlets has, in the record's phrase, been designed to be opened up destructively.