Choosing Between Warm, Cold and Inverted Flat Roofs

The records give physical tests. Between a warm deck and an inverted roof, the question is which side of the insulation the waterproofing is on: on a warm deck the membrane is the surface underfoot, while on an inverted roof the surface underfoot is ballast, paving or planting and the layer resisting water is somewhere below.

Between a cold deck and a warm deck, the test is where the insulation sits relative to the structure and whether a clear air space is maintained above it. Between a cold flat roof and a ventilated pitched roof, the record notes the difference that matters: here the space is shallow, horizontal and closed at the top by an impermeable deck and membrane, so there is no drying route upwards at all.

Because the waterproofing is the outermost layer, rain lands on it, sunlight falls on it, and it can be walked over and inspected. The record treats that as an advantage for inspection and a liability for damage, so access routes, protection where people walk and keeping outlets clear become ongoing duties rather than one-off provisions.

The arrangement also puts the exposed membrane through the widest temperature range in the assembly while the deck beneath it, being inside the envelope, sees very little. The layers move differently, and how that difference is absorbed at laps, upstands and changes of direction is part of the design.

Falls belong to the build-up rather than to the structure. Where the fall is formed in the insulation, the outlet layout and the insulation layout become the same drawing, and moving an outlet after the tapered layout is drawn redesigns the thermal layer at the same moment.

The insulation and the void compete for the same joist depth, and the record states that the competition is not neutral: push the insulation up to the deck and the ventilation path disappears while the roof looks fully insulated. Because the deck above is impermeable, everything that condenses in the void has to leave the way it came, as vapour in moving air.

How much work the void is given is settled at the ceiling rather than at the openings. A continuous plane there leaves the bays with only slow seepage to deal with; a recessed fitting, an unsealed hatch or a cable run opens that plane at the warmest and most humid face of the assembly, and warm air then rises into the bays carrying moisture to the coldest surface in the roof.

The record also names the verification problem. Holding insulation clear of a deck across a whole roof, and keeping every bay connected to the openings, is difficult to achieve and almost impossible to check once the ceiling is up, and it notes that some jurisdictions and clients restrict or discourage the arrangement.

Putting the waterproofing on the deck and the insulation above it keeps the membrane at a stable temperature, out of ultraviolet light and out of reach of foot traffic and dropped tools. The cost is that once the build-up above it is complete it cannot be seen, tested by eye or repaired without removing everything over the affected area.

Two consequences follow that have no equivalent elsewhere. The insulation is repeatedly wetted and is expected to keep working while wet, which narrows the material families found in that position. And the stack has to be held down by weight, because insulation floating on water is a real condition: the ballast or paving resisting it is a structural part of the assembly rather than a finish, so removing surfacing even temporarily removes restraint from everything beneath it.

Locating a fault behaves differently too. Damage traced by dampness inside the building cannot be located by walking the roof, because the leak is at a level that is not the surface and water can travel along it before appearing.

On a warm deck the horizontal build-up has to turn up the inner face of a parapet and stop somewhere defined, and whether the insulation and vapour layer turn with the membrane or stop short decides whether the perimeter is part of the envelope or a gap in it.

On an inverted roof the upstand has to be measured from the concealed membrane rather than from the visible surface, so a build-up of insulation and paving raises the finished level relative to a termination that stays where it was, and the perimeter also has to stop the loose stack from migrating.

On a cold deck the edge serves the weathering, the ventilation and the wall head thermal continuity at once, which is why the record says the perimeter is where the assembly is usually decided.

An outlet is a hole made deliberately through every layer at the point water is being directed towards, which is why outlet positions are agreed with the falls rather than after them. On an inverted roof the outlet has to collect at membrane level while remaining reachable through insulation and surfacing above, so the point of highest maintenance need is the point hardest to get to.

Later penetrations are harder in every case and hardest in the inverted arrangement, where a new penetration crosses ballast, filter, insulation and membrane, each of which has to be reinstated in an order nobody can check afterwards. The records treat provision for future penetrations as a design question rather than something to improvise into a finished roof.

Where a roof is walked on, the wear surface and its supports sit above the membrane and conceal it, so the layer order below is unchanged but the membrane stops being visible and the outlets stop being reachable without lifting something. On an inverted roof used as a terrace, everything above the waterproofing becomes a usable deck with its own thresholds, access and guarding.

Paving carried on adjustable supports adds a further question, because pedestals concentrate the restraining weight into points rather than spreading it, and how load passes down through a loose thermal layer to the deck is raised early rather than late.