Sizing a Rainwater Harvesting Tank: Bounded by the Roof at One End and the WCs at the Other

A Year Off the Roof, Not a Storm Off It

Yield here is an annual figure, which makes it a different calculation from the one that answers what a downpour puts in a butt. Three numbers produce it: the plan area of every roof plane routed to the tank, the long-term annual rainfall at the site, and a pair of coefficients that account for what never arrives. The plan area is the flat footprint and not the tiled surface, for the reason rain falls vertically — the roof harvesting guide derives that, along with the millimetre-to-litre identity, and there is no sense in laying it out twice.

Take the rainfall from a record rather than from memory: the Met Office UK Climate Averages for the nearest long-running station, or the NOAA National Centers for Environmental Information 1991–2020 U.S. Climate Normals. Seven hundred millimetres is used below as an example, not a default: the same roof in Cumbria and in Essex differs by a factor approaching two, and the tank with it. Then apply the two reductions the calculation names: a yield coefficient on the catchment, for what the roof surface and the gutters lose, and a hydraulic filter efficiency, for what the filter turns away. Approved Document G works its rainwater figure by BS 8515's intermediate approach, takes the two together and gives 0.7 as its example — seven-tenths of gross — and that is the figure used below, the same one the site's Rainwater Tank and Cistern Size Calculator opens on. Where the filter's maker publishes an efficiency for your filter, use it rather than this sentence.

The house first. Seventy-four square metres of roof plan routed to one buried tank, at 700 mm, is 51,800 litres gross and 36,260 net. The school block: 1,150 square metres of plan area, same rainfall, 805,000 litres gross and 563,500 net. Neither figure is the tank. Both are the ceiling the tank has to sit under, and the second half of the arithmetic is what the building would drink if it could.

The Demand Is Flushes, and a School's Year Is 190 Days Long

Non-potable demand in a building is dominated by WCs and urinals to an extent that surprises people who have only ever thought about the garden. Everything else on the list — washing machine, outside tap, floor washing — is real but secondary, and irrigation is seasonal in a way that makes it a poor basis for a tank that has to earn its excavation in February. Size the store on the flushing and treat the rest as what fills the margin.

The flush volume itself is one of the few figures on this page that is fixed by law rather than by habit, which is what makes the demand side tractable. In England and Wales the Water Supply (Water Fittings) Regulations 1999 cap a WC installed since 2001 at a six-litre flush, with dual-flush cisterns rated at that full flush and a shorter one below it; a mixed-use average of 4.5 to 5.5 litres is a reasonable working figure for a modern installation. In the United States the Energy Policy Act of 1992 fixed 1.6 gallons per flush for tank-type toilets manufactured from 1994, and the EPA WaterSense Specification for Tank-Type Toilets sets 1.28. Where the building is a new dwelling, Approved Document G of the Building Regulations for England and its companion methodology, The Water Efficiency Calculator for New Dwellings, give both the whole-house consumption target of 125 litres per person per day and the fixture-by-fixture method behind it — which is a better starting point for a house than any rule of thumb.

For the house, then: four people, five flushes each per day at an average 5.5 litres, is 110 litres a day and 40,150 a year. A washing machine at four cycles a week adds roughly 10,400 at the 50 litres a cycle used here as a placeholder, and the machine's own energy label carries a litres-per-hundred-cycles figure that should replace it. Total non-potable demand about 50,600 litres against a yield of 36,260, so on this house yield is the lesser and the tank is eighteen days of it — 36,260 × 18 ÷ 365, or 1,788 litres. In a catalogue of moulded shells that means the 2,000 or the 2,700, and emphatically not the 6,000 the brochure's family-home page offers, which the roof would refill six times a year instead of twenty.

For the school, the household average is no use and the standards say so: BS 8515 and BS EN 16941-1 both expect non-domestic demand to be built from the building's own fixture schedule and occupancy pattern, with CIBSE Guide G, Public Health and Plumbing Engineering, the usual source for per-occupant benchmarks in the UK. This page will not publish a per-pupil figure it cannot cite, so the worked example uses a labelled assumption instead: 460 occupants at two flushes each per school day, at a 4.5-litre average across WCs and controlled urinals, is 4,140 litres a day. Multiply by 190 school days — not 365 — and the year is 786,600 litres. That is more than the roof yields, so once again the yield governs, and the tank is eighteen days of 563,500 litres: 27,800 litres. Replace the two-flush assumption with a survey and the demand moves; it will not change which of the two numbers is the lesser, because the roof would have to grow by two-fifths before it did.

What Is on the Moulding and What You Can Draw

The floor of the answer is partly physical rather than statistical, and it catches people out at commissioning. The number stamped on a shell is its brimful capacity, and you can use neither the top of it nor the bottom. Above, the overflow invert has to sit below the inlet so that a full tank passes flow onward instead of backing up the downpipe, and there is freeboard above that again. Below, the draw-off cannot sit on the floor: the whole point of a floating suction hanging a hand's width under the surface is that it takes the cleanest water in the vessel and leaves the settled solids alone, and everything under it is dead volume by design.

So the usable working volume is the depth between the overflow invert and the lowest draw-off level, multiplied by the plan area of the shell — and both of those levels are dimensioned on the manufacturer's installation drawing, which is where to get them rather than from a percentage quoted in an article. Do this before the tank is ordered, not after: if the working volume comes up short of the sizing figure, the fix is the next shell up, and that is a different hole.

The Store Is Sized on the Year; the Top-Up Is Sized on Twenty Past Twelve

A tank sized at eighteen days of average throughput will still be empty sometimes, and that is not a failure — it is the condition BS 8515 and BS EN 16941-1 both design around by requiring the system to satisfy demand with no rainwater available at all. What that means in practice is that the mains backup, not the tank, is what stops a school's WCs failing in the second week of a dry September. And the backup has to meet the building's peak, not its average, because the average is a fiction spread over twenty-four hours while a primary school does most of its flushing in two twenty-minute windows.

The peaking factor is the standard way of getting from one to the other for a collection vessel: multiply the average daily flow by a factor chosen from the building's occupancy pattern. A school with concentrated break-time use sits high in the usual range; an office or a dwelling with use spread through the day sits low. The result is a design daily throughput rather than an instantaneous flow, and it is the right basis for the top-up leg's duty and for the pump set's turndown. It is not the right basis for pipe sizing at the WC manifold, which is a separate calculation done in loading units under BS EN 806-3 and BS 8558, or in water supply fixture units against Hunter's curve where the International Plumbing Code applies. The two answer different questions and neither substitutes for the other.

One consequence for the tank itself is worth stating plainly, because it is another instance of the ceiling. Mains top-up belongs in a small header volume, not in the whole vessel — a day or two of demand, admitted through a Type AB air gap of the kind specified in BS EN 13077, so there is no continuous path from a non-potable store back to the main under any failure. Sizing the top-up to fill the whole tank turns a harvesting system into an expensively buried mains cistern, and it does it silently: the WCs keep flushing, the meter keeps turning, and nobody notices until the bill is read against the rainfall record.

Everything You Store, You Will One Day Have to Pump Out

Cleaning is where an oversized tank sends its bill, and it is almost never in the appraisal. Harvesting standards expect periodic inspection and desludging of the sediment zone, and a buried vessel has exactly one access route: the turret at the top. Emptying means pumping out through a hose, against the lift from tank invert to ground and into whatever receives it, and against the friction of every metre of hose.

The arithmetic is unkind to the big shell in a way volume alone does not convey. Take the school's 27,800-litre store, a submersible rated 100 litres a minute at zero head with 30 metres of maximum head, three and a half metres of lift and twenty-five metres of 32 mm hose. The pump does not deliver its rated flow — it settles at about 81 litres a minute where the head it can still produce matches the lift plus the friction — and the tank takes just under six hours. The 124,200-litre tank from the month-of-autonomy row takes over twenty-five, which is a two-day attendance and a different quotation. That is the same decision costing money for a second time, years after the hole was backfilled.

1. Isolate the pump set and the mains top-up before anything is drawn down, so the top-up float does not spend the clean refilling behind you.
2. Divert the downpipes at the filter, or pick a settled dry spell — a shower part way through empties several hours of pumping back into the vessel.
3. Pump from the tank floor rather than through the floating suction, which is designed to avoid exactly the sediment you are trying to remove.
4. Step the hose bore up rather than the pump duty if the time looks unacceptable: friction falls with almost the fifth power of the bore, and a bigger pump through the same narrow hose spends the extra output making heat.
5. Treat the confined space properly. A buried tank below ground with a single top opening is a confined space, and entry is a permit and a rescue plan, not a ladder and an assistant.
6. Refill through the mains top-up before the pump set is put back into automatic, so the floating suction is submerged and the pump is not asked to prime on air.

What the Tank Quietly Does to the Softener

Harvested rainwater arrives at essentially zero hardness, which is a genuine benefit and also a trap for anyone specifying treatment on the same drawing. The trap is not that the rainwater needs softening — it does not, and a softener downstream of the tank would be an expense with no duty. It is that moving the WCs and the washing machine onto the tank removes a quarter to a third of the volume that used to pass through the softener — wherever the WC branch was softened too, which is more installations than it should be — and a softener is sized on volume.

Two things follow, one obvious and one not. The obvious one is that nothing fed from the tank should ever be plumbed through the softener, and that WCs should not have been softened in the first place — softening a cistern's fill is salt spent on water that goes straight to drain, which is why a properly configured installation leaves the WC branch and the outside taps on hard mains regardless of whether harvesting is involved. The other is that the softener's regeneration interval stretches. A unit sized for a seven-day cycle at the pre-harvesting volume now sees roughly a quarter less water and reaches the same capacity in nearer ten days — and that runs past the maximum interval most metered valves impose as a calendar override, because resin sitting unregenerated for long periods is its own hygiene problem. Set the override deliberately and record it in the O and M file rather than leaving the installer's default to be discovered years later.

One more coupling belongs on the same drawing: whatever supplies the tank's mains top-up should be taken off the hard main, upstream of any softener, and it should not be softened. Sodium-rich water introduced into a store that may end up on planting, or into a vessel whose overflow discharges to a soakaway, is a chemistry nobody designed and nobody wanted. It is a five-minute decision at the schematic stage and an expensive one to unpick after the ground floor is screeded.