Why a Fragile Roof Surface Is Its Own Hazard Class
Why this matters
If nobody can establish what carries the load, nobody goes on the surface: the work gets done from a platform, from a lift, from staging with its own load path, or from underneath. A fragile roof breaks the one assumption every other fall control quietly depends on, which is that the surface will hold you. Guardrails are bolted to it. Anchorages are fixed to it. Restraint keeps you away from the edge and standing on it. Arrest assumes the fall starts at the edge. Take away the surface and each of those controls is either useless or is itself the thing that puts the concentrated load into the sheet.
There is no paragraph in either Part titled "fragile surfaces." The duty reaches you through the hole and unprotected-edge paragraphs of 29 CFR 1926.501 or 29 CFR 1910.28 once someone has already gone through, and through the employer's duty at 29 CFR 1926.20(b) to run frequent and regular inspections of the site by a competent person. That gap is precisely why this hazard gets planned around instead of planned for: the fall-protection plan has a form to fill in and the surface question does not.
One call, and the four seconds it turned on
A service crew was sent to a light industrial building to replace a failed exhaust fan on an older sheeted roof. Corrugated fibre cement sheeting over steel purlins, with translucent panels every few bays for daylight. Eave height about 18 ft outside, interior clear height about 22 ft to a concrete slab.
The fall-protection plan was good and it was reviewed. The trigger Part was identified correctly as construction, because the fan was being replaced rather than serviced, so 29 CFR 1926.501 applied at 6 ft. A harness and lanyard were specified. An anchorage was identified. Clearance was computed. The crew was briefed on the translucent panels and told to stay off them, and they did.
The technician stepped off the crawl board to reach a fastener, put his weight on a sheet that looked identical to the ten he had walked over, and the sheet failed at the fastener line. He did not go through, because he caught the purlin with his hip. What stopped him was luck, and luck is a control that works once.
Working back: three things that were true, and one that was assumed
True: the plan protected against the edge. The perimeter was addressed and the technician never got near it. The edge was never the hazard on this roof.
True: the translucent panels were identified. Everyone knew where they were and nobody stepped on one. That briefing was correct and it was also the thing that made everyone confident, because it created a mental model in which the hazard was the panels and the rest of the roof was floor.
True: the crawl board was on the roof. It was there, it was the right idea, and it was 5 ft long on purlins at 4 ft centres, which means when it was placed it spanned two purlins with a few inches of bearing at each end, and any shift left it bearing on one. A board that spans two supports has no redundancy and no tolerance for placement error. A board long enough to reach three supports at those centres is over 8 ft, and the difference between the two is the difference between a crawl board and a plank lying on sheeting.
Assumed: that the sheet would carry a person. Nobody wrote that down, because nobody was asked to. Fibre cement sheeting is a cladding element. It is designed for a distributed load - wind, snow, its own weight - spread across its whole area between supports. A person is a concentrated load applied through two contact patches the size of a boot sole, at whatever point in the span they happen to land, and a uniform design load stated in pounds per square foot does not convert into a permissible point load by multiplying by area. The conversion is a structural calculation that belongs to the panel manufacturer or to an engineer looking at that panel, that span and that condition. Nobody in the field owns it.
Why the failure is at the fixing and the span, not in the middle of the sheet
The intuition that kills people here is that a fragile sheet looks fragile. It usually does not, because the two places these roofs fail are the two places you cannot see.
The fixing. A sheet is only as good as what holds it down. Fasteners corrode, wash out, and get overdriven at installation so the washer has crushed the sheet around the hole. A sheet in perfect condition attached by a fastener that has rusted through will lift, slide or tip under a foot at the edge of the sheet, which is exactly where the overlap is and exactly where people put their weight because the overlap looks like the strongest place.
The span between supports. The safe line on this roof runs directly over the purlins, and "walk the purlin line" is a technique that survives right up until someone misjudges by four inches, or until the purlin itself has section loss at the bearing. It also fails completely when a person has to reach, kneel, or take a load, all of which move the contact patch off the line.
Add to that the materials that are fragile because of age rather than type: aged plastic and glass-reinforced rooflights that have gone chalky, thin profiled metal with perforation corrosion at the laps, wood plank with rot at the bearing, and insulation board with no structural deck beneath it, which is a full-thickness trap that reads as a solid floor underfoot right up until it is not.
Where the building is old enough that the sheeting may contain asbestos, the surface question acquires a second one: breaking a sheet releases fibres, and that turns a fall event into an inhalation exposure for everyone in the building. Establish the material from the building's records before anyone disturbs it, and where it is or may be asbestos-containing, that work is not yours to do without the specific programme and respiratory protection the applicable asbestos standard requires.
What a fragile surface disqualifies
This is the part that surprises crews, because these are all normal, good controls that become the hazard here.
- An anchorage fixed to the sheeting. A fall-arrest anchorage under 29 CFR 1926.502(d) has to hold at least 5,000 lb per attached worker, or be part of a designed system carrying a safety factor of at least two under a qualified person's supervision. A load in that range applied to sheeting or to light purlins does not have a load path. The anchor will hold the technician standing still and will not hold the arrest.
- A weighted or mobile anchor. These put a large concentrated deadweight onto the surface, and their published ratings are for specific substrates. On a fragile roof they are a load the roof was never checked for, parked in one place all day.
- A ladder laid on the roof. Unless it bears on structure someone has verified, it distributes load onto the sheeting and gives the user the confidence to move faster.
- Staging or a platform footed on the surface. Same problem, larger load, and now several people are relying on it.
- Restraint alone. Restraint stops you reaching the edge. It does nothing at all about a fall through the middle of the roof, and it makes people comfortable moving around inboard, which is where the exposure actually is.
The re-plan, with the numbers that decided it
The crew stopped, came down, and re-planned. Three options were costed in labour hours rather than technique preference, and the arithmetic that decided it was not the hours.
Option 1, work from the surface with better boards. Crawl boards long enough to span three purlins at 4 ft centres, over 8 ft each, moved by a worker who is on a board at all times. Rejected. It halves the exposure and does not remove it, because there is still a person on a fragile roof and the failure mode is a placement error, which is the thing that just happened.
Option 2, arrest from a rated anchorage. Rejected on two grounds, and the clearance one is worth printing:
| Line | Value | Note |
|---|---|---|
| Fall distance available below the walking surface, if the fall is off the roof | 18 ft | To grade at the eave |
| Fall distance available below the walking surface, if the fall is through the roof | 22 ft | To the interior slab, which is the case the roof surface creates |
| Which one governs the clearance calculation | the one for the actual exposure | For a through-roof fall the drop is measured to the interior floor, and there may be equipment in that path that shortens it further |
| Anchorage available on this roof | none rated | Nothing on the sheeting or the light purlins has a 5,000 lb load path |
| Result | Not available | The clearance question never gets asked, because there is no compliant anchorage to ask it from |
The order there matters. People compute clearance first because it feels like the technical step, and then discover the anchorage does not exist. The anchorage question comes first, and on a fragile roof it usually ends the conversation.
Option 3, do not load the surface. A boom lift positioned outside, with the technician working from the platform at the fan, never transferring to the roof. Selected. It carried its own cost: the whole setup then turned on the ground, because the lift's outrigger loads land on a yard surface with a backfilled service trench crossing it, and that is a separate question with its own answer.
That trade is the honest one and it is worth naming. Choosing not to load the roof does not remove risk, it moves it to a hazard that has published numbers, a load chart and a bearing capacity somebody owns. A risk with an owner and a number beats a risk with neither.
What would change the answer
Documentation exists. If the building's records, the panel manufacturer, or an engineer establish that this deck carries a concentrated worker load at any point in the span, the roof is not fragile and the ordinary fall-protection plan applies. That is a document, not an opinion, and it is specific to the condition the roof is in now rather than the condition it was built in.
A structural deck under the covering. A membrane over rigid board over steel deck is not a fragile surface, and it is common to meet both on the same building where an addition was built differently. The boundary between the two is invisible from above and it is where people walk from one onto the other.
Netting or decking below. Where the work genuinely has to happen on the surface, protection installed underneath changes the arithmetic, because it converts a 22 ft fall to a slab into a short arrested fall. That is a designed installation with its own criteria, not a tarp.
Confirming you have actually established the surface
The test is whether you can name the source. Ask whoever says the roof is walkable where they got it: a drawing, a manufacturer, an engineer, a records file, or their impression from standing on it last year. Only the first four are answers.
The second check is the boot count. If your plan has a person on the surface at any point, including the transition on and the transition off, the surface question applies to that person and has not been avoided by keeping the work off the roof.
References
- 29 CFR 1926.501 and 29 CFR 1910.28, for the construction and general industry fall protection duties that reach a roof surface once it fails, at 6 ft and 4 ft respectively.
- 29 CFR 1926.502(d) and 29 CFR 1910.140, for the anchorage criteria that a light sheeted roof cannot satisfy.
- 29 CFR 1926.20(b), for the employer's duty to run frequent and regular site inspections by a competent person.
- Panel manufacturer documentation or a structural engineer, which owns any claim that a given deck carries a concentrated worker load.
- See related: the library's cards on why a skylight is treated as a hole, on computing fall clearance, on fall restraint, and on what ground conditions decide about a lift you are standing in.