Water standing on the upslope side of a rooflight upstand is a drainage fault in the roof, not a fault in the rooflight. The upstand is a solid obstruction planted in the middle of a shallow drainage plane, and unless the falls were set out to run water past it, everything arriving from above collects along its uphill face and stops. If that pool is still there 48 hours after rain in drying conditions, the roof ponds. Left alone it will degrade the membrane, the bond line and the timber underneath, in that order, over roughly three to eight winters. The correction happens during replacement, by rebuilding the falls, not by sealing anything.
This page covers the engineering: what physically happens in a pond, how the damage accumulates, and the three ways falls get corrected when we take a failed unit off. If you want the short causal explanation of why flat rooflights pond in the first place, that is on why flat rooflights pond water. The fall figures themselves, 1:80 finished and why we design to 1:60 or 1:40, are set out on minimum fall for a flat rooflight. Neither is repeated at length here.
The upstand is a dam, and dams collect things
Picture the roof as a very shallow ramp. Water lands anywhere on it and travels downhill to an outlet or a gutter. Now put a rigid, watertight, 150mm-tall rectangle in the middle of that ramp. Water arriving from the upslope side cannot pass through it, so it divides and travels around. That is the intended behaviour, and on a correctly set out roof it works.
What goes wrong is the geometry of the division. Water splitting around an obstruction slows at the point of the split, in the centre of the upstand’s uphill face. Slow water deposits what it is carrying: grit off the membrane, moss spores, leaf fragments, roofing granules from a felt cap sheet. Over a couple of seasons that builds a low silt bank along the base of the upstand. The bank is only a few millimetres tall. On a roof with 12mm of fall per metre, that is a significant fraction of the drop available, and it now holds a shallow pool that never fully drains even though the roof technically falls correctly.
This is why a roof can pass a laser check at handover and pond by year four. The fall did not change. The effective fall did, because the roof grew a kerb it was not designed with.
The three geometries that guarantee it
- Upstand set square across the fall. A long side facing directly uphill presents the maximum obstruction to the maximum flow. Turning the unit so the fall runs along its length rather than into its face costs nothing at design stage and halves the problem.
- Upstand immediately upslope of the only outlet. Everything the roof collects has to get round the unit before it can leave. If the route round is tight on one side, that side backs up.
- Upstand landed in a dished bay. Common on retrofits into 1970s and 1980s roofs, where firrings were cut on site and the deck has since deflected between joists. The installer measures the fall once, across the roof, gets a good number, and puts the unit into the one flat pocket.
What standing water actually does, in order
Ponding does not cause a leak on Tuesday. It runs a slow chemical and mechanical attack on four separate things, and the sequence is consistent enough that we can usually date a pond from the damage.
Year one to two: the surface
A permanent pool on a dark membrane is a warm, still, nutrient-rich environment. Algae and then moss establish in it. On EPDM that is mostly cosmetic at first. On a felt cap sheet the moss roots into the mineral granule layer and lifts it, exposing bitumen to the UV those granules were there to block. On GRP the pool holds fine grit that abrades the pigmented topcoat off the laminate.
Year two to four: the bond line
This is where a pond becomes a rooflight problem rather than a roof problem. The point where the covering turns from horizontal to vertical is a bonded joint on almost every system. On a dry roof that joint sees water for minutes at a time. In a pond it is submerged continuously.
Standing water on a black membrane in July reaches temperatures well above ambient air. Adhesives are specified against a service temperature range, and sustained heat plus permanent immersion degrades them far faster than either does alone. In January the same pool freezes. Water expands roughly nine per cent on freezing, into the weakest available direction, which is the peel direction at a bonded lap. Each freeze lifts the edge of the bond fractionally and each thaw lets water into the space that opened.
Suffolk and north Essex see enough freeze-thaw cycles per winter for this to be a real mechanism rather than a theoretical one. It is also why ponding damage shows faster on shaded and north-facing roofs, where the pool cycles repeatedly instead of freezing once and staying frozen.
Year three to six: the termination and the seals
Once the bond at the base of the upstand starts to peel, water tracks up behind the covering by capillary action and wind pressure. It reaches the termination at the top of the upstand from the wrong side, the one direction no termination is designed to resist. A mechanical termination bar with sealant behind it holds for a while. An adhesive-only detail does not. Meanwhile the rooflight’s own perimeter gasket sits in a humid microclimate the manufacturer never tested for. EPDM and silicone gaskets cope reasonably. The mastic bedding used on 1980s and 1990s dome frames does not, and it shrinks back at the corners first.
Year four onward: the timber
By the time water shows on a ceiling, the upstand timber has usually been damp for years. Treated softwood held at high moisture content, in still air, with no drying cycle, goes soft on the top edge and at the corners where end grain is exposed. That finding decides most replacement jobs, and it is not visible until the old unit comes off.
Inspect from the ground or a window, never from the roof. A flat roof with standing water on it is slippery, and an old dome will not carry your weight. A photograph taken from an upstairs window two days after heavy rain tells an installer more about your job than any measurement you can send.
| Covering | First visible sign | The failure mechanism | Rough time to a wet ceiling |
|---|---|---|---|
| EPDM, 1.2 to 1.5mm | Algae film, then edge lift at the upstand base | Adhesive hydrolysis under warm immersion, then freeze peel at the bond line | 5 to 10 years |
| GRP fibreglass | Topcoat worn matt in the pond footprint | Abrasion exposes the laminate, then crazing at the internal angle where the fillet flexes | 4 to 8 years |
| Single-ply PVC or TPO | Plasticiser bloom, staining in the pond outline | Weld seams tolerate it well, but any cold weld in the pond opens quickly | 6 to 12 years |
| Polyester-reinforced felt | Granule loss, moss rooted at the upstand base | Moss lifts granules, UV attacks bare bitumen, laps soften and creep | 3 to 7 years |
| 1980s oxidised bitumen felt | Blistering and crazing across the pond | The material is already past service life. Ponding simply finishes it | Already happening |
Figures are working estimates from what we find on site in Suffolk and north Essex, not manufacturer data. Shaded and north-facing roofs run at the faster end of every range.
Correcting falls during a replacement
When a rooflight comes off a ponding roof, we have the covering back and the deck accessible around the opening. That is the only sensible moment to fix the drainage, and it is why we survey the fall on all four faces before quoting rather than after. There are three corrections, and the right one depends on how much of the roof is wrong and how much height is available at the perimeter.
1. Re-firring a local area
Where the roof falls correctly overall and the problem is a dished pocket around the opening, the fix is local. The deck comes up in the affected bay, tapered timber firrings are cut and fixed over the joists, and the deck goes back. This is the cheapest correction and it suits the 1960s to 1980s garage and utility roofs at Kesgrave, Rushmere St Andrew and Chantry, where the original firrings were often skimped or laid the wrong way from new.
The limit is height. Firring raises the deck, which eats into the upstand height above it. If reinstating the fall would drop the effective upstand below 150mm at the low point, the upstand gets rebuilt taller at the same time.
2. Tapered insulation
Where the deck itself is flat and cannot be re-boarded, or the correction spans more than one bay, tapered insulation is laid over the existing deck to a designed scheme. Boards are manufactured to a set taper, typically 1:60 or 1:40, and set out to a drawing so the falls converge on the outlet and run clear of the upstand on all four faces. We quote it separately by area. It is a permanent correction rather than a deferral, and it lifts the thermal performance of the roof at the same time.
3. Moving the unit or splitting it
The correction people never think of is geometric. On a narrow side return, two 1.0 x 1.0m units with a clear drainage channel between them drain better than one 2.2m unit spanning nearly the full width. Rotating a rectangular unit so the fall runs along its length rather than into its face does the same for no extra cost. If the rooflight is being replaced anyway, the layout is back on the table.
Outlets
Correcting the fall is pointless if the water has nowhere to go. A large glass rooflight sheds everything landing on it to its perimeter, which raises the peak flow the outlet has to take in a downpour. A single 63mm outlet on a small roof that has just gained 2.5m² of glass can be marginal. Adding a second outlet while the covering is back costs very little. Doing it after a flood costs a ceiling. Ipswich has a lot of mature street trees, so a leaf guard is not optional here. A blocked outlet turns a roof with correct falls into a tank, and the upstand is the tallest thing in it.
A worked example
A rear extension roof in central Ipswich, 4.2m by 3.0m, felt laid about 1998, a single 63mm outlet at the front left corner and a 900 x 900mm polycarbonate dome near the centre. The owner reports a brown ceiling stain each autumn and again in spring.
Laser check at survey: the roof falls 55mm over 3.0m from the back wall to the front edge. That is 1:55 and correct on paper. Measured on the four faces of the upstand individually, the uphill face has 3mm of fall over the first 400mm and the left face has none. Photographs from the bedroom window two days after rain show a pool roughly 700 x 400mm against the back and left faces. The deck has deflected between joists in that bay and a silt bank has built at the upstand base. The dome is not the problem. It is where the water became visible.
The job that follows: strip the covering back around the opening, lift the deck in that bay, re-firr to reinstate fall on all four faces, rebuild the upstand insulated to 150mm above the new finished level, add a second outlet with a leaf guard, recover locally and set a flush glass unit. The felt is polyester-reinforced rather than 1980s oxidised and is sound outside the pond footprint, so a full renewal is not required yet. Prices for the unit sit in the published ranges on the flat roof rooflight installation cost page. The upstand rebuild is a provisional item, confirmed with photographs once the old unit is off.
Why resealing does not touch this
The trade default on a rooflight sitting in a pond is a bead of mastic round the frame. It addresses none of the four things above. The bond line is still submerged, the timber is still wet, the fall is still wrong. The mastic closes the visible route, which stops the staining for a season or two while the water finds another way and the timber carries on decaying out of sight.
We do not offer resealing or patching on any rooflight. On a unit sitting in standing water it delays the only intervention that works, which is getting the covering back and putting the falls right. That argument in full is on reseal or replace a dome rooflight.
The test that settles it. If the staining follows rain from one direction and shows at one corner, it is water ingress. If it appears in cold weather, worsens after cooking or showering and forms evenly on all four reveals, it is condensation from an uninsulated upstand, and no drainage work will change it. The building physics of that is on cold bridging and damp around flat rooflights.
Where this sits in the rest of the work
Ponding is one of three things that decide whether a flat rooflight lasts. The other two are the upstand itself and how the covering is dressed at it, covered on rooflight kerb construction and rooflight flashing and membrane detailing. Where the water has already got in around the kerb, the diagnosis is on leaks at the rooflight kerb. The wider picture, including structure, glazing specification and published prices, is on the main flat roof skylight installation page. If you are building new rather than correcting, fitting a flat rooflight without leaks covers getting it right first time.
Get the falls checked before you buy a rooflight
Every survey we do includes a laser check of the fall on all four faces of the opening, not once across the roof. If your roof already ponds, we will tell you what correcting it costs before you commit to a unit, because fitting a rooflight into a roof that holds water is spending money on a problem nobody has solved. Every installation carries a 10-year workmanship guarantee, and the installers have 25 years fitting rooflights behind them.
Send a photograph of the standing water taken from a window two days after rain, along with the roof size and roughly where the property sits, and request a fixed quote. We cover Ipswich, Suffolk and north Essex, roughly 30 miles out.
Thinking about this job? We survey across Ipswich, Suffolk and north Essex and give you a fixed price before any work starts, with a 10-year workmanship guarantee.