Stepping Foundations Down a Sloping Site

Two falls, and only one of them is drawn

The gradient on the topographical survey is the site's, measured down the line of maximum fall, and it is almost never the gradient you build to. What the footing sees is the fall along its own line, and a wall set at an angle to the line of maximum fall sees the site gradient multiplied by the cosine of that angle. It is worth having the number rather than the intuition: at thirty degrees off the line of maximum fall a wall has lost only thirteen per cent of the gradient, but at sixty degrees it has lost half of it, and at ninety it is dead level and takes no steps at all. On a rectangle set square to the contours, two walls do all the stepping and the other two are level runs at different heights.

So level along the wall lines and do it after the setting out, because the building rarely lands where the survey assumed. Take a reading at every profile and every intended step position, work off one bench mark that will survive to the slab, and close the run back onto it. A traverse that closes badly over a 40 m perimeter is telling you something about the staff, the tripod or the ground, and finding out now costs a quarter of an hour.

Then take the second fall, the one nobody levels: the fall of the bearing stratum. Ground surface and founding stratum are two surfaces under no obligation to be parallel. A hillside with a clay head over rock frequently has a formation dipping steeper than the topsoil, and a made-up garden on a fall can have one running the other way entirely under a metre of imported fill. The steps follow the worse of the two, because the depth rule has to hold everywhere along the run rather than on average. Two trial pits, one at each end of the falling wall, settle it — and where they disagree with the ground surface by more than a step, the stratum is setting the step positions and the survey is not.

A step is a lap, and trench fill is lapped by its own rule

A step is a splice in a load path. The wall above runs continuously through it, but the concrete under that wall stops at one level and restarts lower down, so for a short distance the load has to travel through a region where the upper pour is bearing on the lower pour rather than on the ground. That is what the lap is: the length over which the two pours share, and it is dimensioned so the load spreads out of the upper section and into the lower one before the upper section runs out.

In England and Wales the arithmetic is published, and it is published twice — which is the part that gets missed. Approved Document A, Section 2E, caps the step the same way for everything: no step may be taller than the thickness of the foundation. Then it splits. A strip foundation is lapped by the greatest of twice the height of the step, the thickness of the foundation, or 300 mm. Trench fill is not: for trench fill the lap is the greater of twice the height of the step or one metre, and the thickness of the concrete does not enter it at all. Read the current edition rather than a memory of one, because the plain-concrete section is one of the parts that has been re-set between editions and the diagram numbers have moved with it.

Carrying the strip formula onto trench fill is the common error, and it goes wrong in both directions. On a conventional 225 mm strip the step is capped at 225 mm and twice the step is 450 mm, so the step drives the lap and the thickness never gets a look in. On 750 mm trench fill the strip formula would say 750 mm and the actual answer is a metre, so the run is short by 250 mm at every step. On 1.20 m trench fill the strip formula would say 1.20 m and the actual answer is still a metre, so the same habit now over-orders. The rule that matters on trench fill is the flat metre: a 225 mm step, a 300 mm step and a 450 mm step all lap one metre, and below a 500 mm step making the step shorter buys nothing at all. That runs against the instinct on almost every sloping-site job — on trench fill the cheap direction is a taller step and fewer of them, not a gentle staircase, and no amount of extra depth lengthens the lap.

The American codes handle the same problem with a different emphasis, and it is fairer to say so than to import the British formula and pretend. The International Residential Code, at Section R403.1.5, requires the top surface of a footing to be level, allows the soffit a slope of no more than one in ten, and requires stepping wherever the top has to change level or the soffit would exceed that slope. What it does not carry is a published lap length; that comes from the engineer of record or a local amendment, and none is invented here. Where the ground beyond the building is steeper than one in three, the IRC at Section R403.1.7 and the IBC at Section 1808.7 add something the code writers kept carefully distinct from stepping: the footing must sit back from the face of a descending slope by H/3, capped at 40 ft, and clear of an ascending one. That rule regularly drives the lowest pour deeper than any bearing check would, and it is a setback, not a step.

Choosing a step that divides the fall and lands on a course

The step height is not free to be any number, because whatever sits on the concrete is laid in courses and those courses have to arrive at the damp course together. British brickwork gauges at 75 mm and blockwork at 225 mm, so a 225 mm step is one block or three bricks, 450 mm is two blocks or six, and 300 mm is four bricks but no whole block. Pick a step off a round number in millimetres and somebody is cutting blocks below ground or making up 40 mm in a bed joint — which is exactly where a substructure wall loses its line.

The second constraint is that the fall rarely divides by the step. Our 1.204 m over the long wall against 225 mm steps gives 5.35 steps: five of them leaves 79 mm over, which is close enough to one brick course to be absorbed by making a single step 300 mm rather than 225 and letting the arithmetic land exactly. That is the fix worth learning — carry the remainder in one step, not spread across every bay as a slightly different depth, and never carry it at the shallow end, because the shallow end is the one with a minimum founding depth to satisfy and the deep end is not.

The third constraint catches trench fill on a steep fall, and it is not the one people expect. Bay length is the step height divided by the gradient, so 225 mm steps give bays of 2.25 m on a 1 in 10 fall, 1.125 m on a 1 in 5, and only 900 mm on a 1 in 4. Set that against a lap fixed at one metre and the arithmetic closes in without the depth ever being mentioned: the bays stop being longer than the laps once the step drops below the gradient written as a fraction of a metre — 200 mm on a 1 in 5 fall, 250 mm on a 1 in 4. Past that point every bay is lapping into the next one, the staircase has become a continuous double-depth mass, and the honest answers are a taller step, a level-soffit mass foundation, or somebody else's foundation type. The corollary is worth holding on to, because it is the reverse of the usual worry: depth is the one thing that does not hurt here. Near mature trees in shrinkable clay, where the founding depth is set by the tree rather than by the load, the thick section permits a taller step, and a taller step is exactly what pushes the bays back out past the metre.

The staircase in the ground is deeper than its design depth

Here is the arithmetic that goes missing from every sloping-site dig sheet. Within one bay the soffit is level and the ground is not, so the depth from ground to formation is smallest at the downhill end of the bay and largest at the uphill end, and the difference between the two is exactly one step. Since the minimum founding depth has to be met at the shallow end, the deep end of every bay is a full step deeper than the number on the drawing, and the average depth along the whole run is the design depth plus half the step. Nothing about that is optional and nothing about it is waste.

On the annexe that means 900 mm of design depth becoming an average of 1.013 m with 225 mm steps, and an average of 1.013 m over 24 m of falling wall at 600 wide is 14.6 cubic metres of dig where the drawing implied 12.96. Choose 450 mm steps to save boards and the average goes to 1.125 m, and the dig to 16.2. The sawtooth is bought either as concrete or as blockwork depending on where you stop the pour, but it is bought, and it is the single most reliable difference between a sloping-site groundworks price that stands up and one that does not.

The dig has two other habits on a fall. Spoil gathers at the bottom of the site because that is where gravity and the machine both want it, and the bottom is exactly where the deepest trench and the tallest unsupported face are — a heap surcharging the crest of the worst excavation on the job. Keep it uphill or keep it off site. And the trench is now a drain: groundwater moves along the formation and stacks up behind each riser, so the lowest bay stands in whatever the five above it have shed, which is a sump planned for rather than a bucket at eleven o'clock. Support on the deeper faces is a formal decision under CDM 2015 or OSHA 29 CFR 1926 Subpart P, and the trench guide below covers it properly rather than half-covering it here.

1. Set the step positions on the ground before anything is dug, and mark them on the profiles as well as on the ground, because the ground marks go first.
2. Dig each bay to its own level and check it against the bench mark, not against the bay behind it — errors in a staircase compound in one direction.
3. Cut the formation step where the lap begins, uphill of the riser, not under it.
4. Trim the last 75 mm by hand and do it on the day of the pour, not the week before.
5. Keep spoil above the deepest face, and put the sump in the lowest bay before you need it.

Every lap is concrete you pour twice

Over the lap length the concrete runs from the lower pour's soffit right up to the upper pour's top surface, so its depth there is the thickness plus the step. That is an extra block at every step measuring the lap by the trench width by the step height, and on the annexe — ten steps across the two falling walls, a metre of lap at each, 600 wide, 225 mm tall — it comes to 1.35 cubic metres on an 18 cubic metre run. Seven and a half per cent, and that is the floor rather than the answer: choose to bring the top of concrete up to ground level at the high end of each bay instead of stopping it level with the low end, and the sawtooth above the pour is another 1.6 cubic metres, taking the run to 21.0 and the overrun to a shade over sixteen.

The proportion moves with the shape of the job rather than with its size, which is why it cannot be carried as a habitual percentage. Shallow steps on a gentle fall barely register. Trench fill on a steep fall is the punishing case, because the lap stays at its metre while the bays get shorter, and it can put a third of the run into double depth — at which point the arithmetic has stopped being a waste allowance and started being an argument for a different foundation. The one number a ready-mix ticket cannot be corrected on is the one already discharged.

One pour downhill, or a joint you have to make on purpose

The sequence decides how much the stop-end boards are holding, and the two cases differ by nearly an order of magnitude. Fill the bays from the bottom of the hill upward on one visit and the board at each riser retains only the step: 225 mm of wet concrete across a 600 mm trench end, held by two pins driven into the sides. Start at the top with the bay below still empty and the same board is retaining the full 750 mm of the upper pour against nothing, which is a propped shutter, not a board. Decide at setting-out, because the answer changes what gets loaded onto the truck.

Consistence is the other half of the same decision, and it is a specification question rather than a site one. A trench-fill mix ordered for easy placement migrates downhill along a bay that is level only in intention, and keeps migrating past the riser if the board is not sealed to the trench sides. BS EN 206 with BS 8500 sets consistence as a class rather than a preference, so a stiffer class on a stepped pour is a conversation to have when the concrete is ordered. The same order settles exposure: BS 8500's designated concretes carry GEN1 for unreinforced strip foundations in non-aggressive ground and the FND range for ground that is not, with the aggressive chemical environment class coming from a soil test read against BRE Special Digest 1. A falling plot digs into strata a level one never reaches, so the test that described the top of it may not describe the bottom.

If the run cannot be poured in one visit, the joints stop being incidental and become part of the design. A construction joint through a foundation wants to be placed where the engineer puts it — normally at a step, where the section changes anyway — and it wants the treatment BS EN 13670 and ACI 301 describe for the execution of one: laitance removed, the surface roughened to expose aggregate rather than trowelled smooth, kept clean and damp, and any reinforcement carried through with its own lap detailed by whoever designed it. What ruins a daywork joint is not the delay; it is a stopped end that was tipped rather than shuttered, left to skin over, and then poured onto three days later with the debris still on it.

The formwork a stepped foundation consumes is small and awkward rather than large: a board per riser cut to the trench width and the step height, pins, and whatever side form the trench face refuses to stand up for after rain, which on a fall is the deepest bay. Order it as a list of pieces, not an area, and cut the risers to the step rather than to a stock width — a board 25 mm proud of the intended top of concrete becomes a level everybody works to by accident.

1. Fix the pour direction before ordering, and tell the plant: it changes the mix as well as the sequence.
2. Seal each riser board against both trench sides, not just against the formation.
3. Where two visits are unavoidable, put the joint at a step and treat it as a construction joint from the start.
4. Strike, clean and re-cut riser boards between bays rather than leaving them cast in.
5. Record the level of the top of concrete in every bay on the day, because the bricklayer sets out from it and it is buried by the time anyone asks.

Above the concrete, the courses finish the falling

Whatever the concrete does not step, the masonry does, and on a single-storey building with one floor level that share is heavily one-sided. The damp course sits at one level all the way round and the top of the concrete steps away from it going downhill, so the substructure wall grows by the full fall of the site between the high corner and the low one. On the annexe the high end carries about 300 mm from top of concrete to damp course and the low end about 1.5 m; count the returns and the perimeter needs near 36 square metres of walling per leaf where a level plot wanted twelve. In a cavity substructure that is two leaves — 70-odd square metres and around seven hundred blocks nobody drew.

Which raises the question the sequence actually turns on: fill that with concrete or build it in blocks. Filling it means carrying the pour up rather than stepping the top of it, and 36 square metres at 600 mm of trench width is more than 21 cubic metres of extra concrete — against seven hundred blocks and the labour to lay them. Concrete wins on speed and on not needing a bricklayer before the slab; blockwork wins on quantity by a wide margin, and it wins by more the more fall there is. Most jobs land on blockwork with the pour stepped, and the ones that do not are usually the ones where the programme, not the take-off, made the decision.

Below the damp course the units are not the units above it: dense aggregate concrete blocks to BS EN 771-3 or engineering brick, chosen for the exposure of buried masonry rather than matched to the wall above, with the cavity filled in weak concrete to a level below the lowest damp course that comes from the warranty provider's standard rather than from the Approved Document. Set the staircase out from the damp course downward so the odd course lands at the bottom, where it can be absorbed. What the damp course itself does as a plane in the wall is covered in the garden wall guide and not repeated here.