Fencing a Site Perimeter

Eleven o'clock on the Monday, and the hire desk wants a number

The fence is always the first order and the least prepared one. Ground investigation has a report behind it, the crane has a lift plan, the piling rig has a mat design — and the fence gets ordered off a phone call by whoever is standing on the site at the time, from a plan they have scaled with a thumb. Then a wagon turns up with a number of panels somebody guessed at, a pallet of feet, a bag of couplers, and a driver who wants to be away in forty minutes.

Two numbers decide whether that call was any good. How many panels the line actually takes, which sounds trivial and is where most of the money leaks; and what has to sit on the ground to keep a run of them upright when the first proper wind arrives, which is where the risk sits. Everything else on the docket — anti-lift couplers, stays, netting, gate leaves, signage — hangs off those two.

Get either wrong and the failure is quiet. Under-order and you finish the line on a Friday afternoon with forty metres of open boundary and a security guard you did not budget for. Under-ballast and nothing happens at all, for six or eight weeks, until a named storm crosses and two hundred metres of fence lies flat across a footway with the netting still attached to it.

What arrives on the wagon, and what actually holds it up

A hired perimeter is five separate products that the hire desk sells as one line item, and they fail independently. The panel is a welded mesh sheet in a tubular frame, typically 3.5 m by 2.0 m (about 11 ft 6 in by 6 ft 7 in), weighing under twenty kilos and carried by one person. The feet are moulded blocks the panel legs drop into, and they do two jobs badly described by their name: they hold the panel vertical, and they contribute a fraction of what the wind needs. Couplers clamp two adjacent uprights so a line of panels behaves as one thing rather than as a row of separate ones. Stays or stabiliser struts run from the panel back to a base plate and are the only component with a lever arm worth anything. Netting or shade cloth is optional and changes the structural problem completely. And the ground underneath is a component too, whether anybody drew it or not.

The order that list comes apart in is the order it goes wrong in. Netting is the first thing a gale takes, then the panels lift out of their feet where nobody fitted anti-lift couplers, then the whole coupled run pivots about the downwind edge of the feet and lies down as one piece. Each of those is a different fix, and only one of them is bought by the tonne.

The line you are allowed to fence

Nobody fences the boundary. They fence a line set inside it, and the gap between those two changes the length before a single panel is counted. Site boundaries carry easements, oversail rights, root protection areas, adopted verges, and neighbours who will measure. A fence set on the title line across an adopted footway is not a fence but an obstruction, and in England and Wales work adjoining a street that needs a hoarding is dealt with under section 172 of the Highways Act 1980 and a licence from the highway authority.

The fence line is also a services question before it is a security one. The moment the answer to soft ground is a driven anchor pin, a ground stake or a screw-in stay, you are putting steel into the ground on a line nobody has traced. HSE guidance HSG47, Avoiding danger from underground services, exists for exactly this, and a temporary fence is the classic job where the permit-to-dig discipline gets skipped because the hole is only 400 mm deep. Trace the line, mark it, and where you cannot get a clean trace, ballast instead of pinning.

Then decide what the fence is being asked to do, because that decision is not yours alone. In the United Kingdom the Construction (Design and Management) Regulations 2015 put a duty on the principal contractor to take reasonable steps to prevent access by unauthorised persons to the site, and HSE guidance HSG151, Protecting the public: your next move, is the document that turns that duty into decisions about lines, hoardings, sight lines and out-of-hours condition. United States federal construction rules carry no blanket requirement to fence a site at all — 29 CFR 1926.202 governs barricades for the protection of employees — so the requirement usually arrives instead from the local jurisdiction, from the fire code's construction chapter, or from the owner's own contract. ANSI/ASSP A10.34, Protection of the Public on or Adjacent to Construction Sites, is the reference worth having in front of you when that requirement is vague.

1. Walk the whole line with a measuring wheel and a can of spray, marking the fence line on the ground rather than agreeing it on a drawing.
2. Set the line clear of the boundary by whatever the easements, adopted verge and root protection areas demand, and record where you have set it and why.
3. Trace for buried services along the marked line before committing to any pinned or staked stability solution.
4. Mark every opening — main gate, pedestrian gate, emergency egress, the crane's delivery gap — and take those lengths out separately.
5. Note the ground condition at each change: hardstanding, verge, crossfall, the top of a batter, the wedge behind a new retaining wall.
6. Wheel the line a second time in the opposite direction and reconcile the two totals before anything is ordered.

Netting is a structural decision that gets made by a marketing department

Debris netting, shade cloth and printed banner mesh arrive on site for reasons that have nothing to do with structure: dust, screening the works from a neighbour, a client who wants their branding on the hoarding. Each takes a panel that was mostly hole and makes it mostly surface. The scaling is not subtle — the loading standards handle porous fences by scaling the force coefficient with the solidity of the face, so going from an open mesh face to a sheeted one is a change of the same order as the difference between a fence and a wall.

So make it a decision with a paper trail. Ask the supplier what their ballast schedule assumes about netting, because most schedules publish two sets of figures and the sheeted one is the one nobody reads. Where the branding is a client requirement, price the extra ballast and bracing into the quote as its own line rather than absorbing it. And write the strip-down trigger into the site's severe weather procedure with a wind speed against it — netting comes off before a forecast storm, not during one, because a crew unclipping a flapping banner in a gale is a worse problem than the one you were trying to avoid.

The ground the feet actually stand on

The moment calculation assumes the pivot edge stays put. Everything that lets it move is a ground problem. A verge that has been wet for a week, or soft made ground worked by the repeated loading of every gust, lets the downwind edge of a foot sink a few centimetres, and the panel is leaning before the wind has done anything. Where a run has to cross ground like that, spread the load with a timber sole or a mat under the feet, or move the line onto the hardstanding and accept a longer perimeter.

Crossfall is the one that catches experienced people, because a fence on a slope looks fine and is not. A moulded foot sitting on a camber has one edge bearing and the other proud, so the effective lever arm on the downhill side is shorter than the arm the calculation used, and the panels above it are already carrying a lean. Pack the low side, or step the fence at the change of grade rather than running it through.

Two locations deserve an outright refusal rather than a mitigation. The first is the surcharge zone at the top of an excavation or a batter, where the mass of the fence and its ballast becomes a surcharge on the slope — the reason temporary works design puts a setback there, and the reason ballast blocks are the last thing that should sit on it. The second is the fill wedge behind a new retaining wall before it has been signed off. Both are cases where the fence is safe and something else is not, which is the version of this failure that nobody sees coming.

Before the first gale

Every site gets a storm eventually, and the difference between the ones that lose their fence and the ones that do not is whether anybody did anything in the forty-eight hours before it. Wind warnings are published days ahead; the failure is never a lack of information, it is that nothing in the site routine converts a forecast into an action with a name against it.

The list below is short on purpose — an hour on a Friday with a bag of couplers and a telehandler, not a procedure that needs a method statement written first.

1. Strip the netting and any banner mesh from the exposed runs, and do it while the forecast is still a forecast.
2. Walk the line and fit the anti-lift couplers that were left out at handover — they are almost always the missing item, and they are what stops a panel being lifted clear of its feet.
3. Re-stay the end panels either side of every gate and every return, and check the stays that were removed when a delivery needed the space and never went back.
4. Look at what is stacked against the fence. Sheet material, formwork panels and stillages leaned against a run turn a permeable fence into a solid one at exactly the spot with the least bracing.
5. Ballast to the supplier's sheeted-condition figures on any run that has to keep its netting for dust or screening reasons.
6. Photograph the line before the weather arrives, so the condition it was in is a record rather than a recollection.
7. Agree who walks it afterwards, at what hour, and who they call if a run is down across a footway or a road.

It is temporary works, and somebody has to own it

A hired fence is bought like a consumable and behaves like a structure, and that mismatch is where the responsibility gets dropped. In British practice the route out is well established: BS 5975, Code of practice for temporary works procedures and the permissible stress design of falsework, sets the procedural framework, a temporary works coordinator is appointed, the fence has a design or a documented justification, and somebody checks it against the design before it goes into use. That framework was written for falsework and it applies just as cleanly to two hundred metres of ballasted panel standing beside a public footway.

What goes in the file is short and specific: the walked perimeter length and where the line was set, the supplier and the panel, foot, coupler and stay references as hired, the design wind pressure used and where it came from, the ballast and bracing schedule the supplier published against that condition, the stay positions as fitted rather than as intended, and the severe weather trigger with a name against it. That is a page. It is also the only thing that answers the question an insurer asks after a run goes over, which is never whether the fence was strong enough but how anybody knew it was.

Off-hire, and the count that closes the account

The account stays open long after the fence is useful. Hire runs on a daily rate against a delivered count, and that count is the one on the driver's docket, not the one you ordered — reconcile them at the gate while the wagon is still there, because a discrepancy found in month four is one you will pay for. Panels that go into internal splits and never come back out are the usual source of a collection short.

Count and stack for collection rather than letting the collection count for you. Damaged panels get charged at a replacement rate, so separate them, photograph them, and get the damage agreed at collection rather than argued at invoice. Feet and couplers are the items that vanish: couplers walk off into other uses within a fortnight, and feet get buried under a spoil heap and found by a machine. Both are charged for. A perimeter fence is a cheap item that becomes an expensive one entirely through paperwork, and the last hour of it is worth as much attention as the first phone call.