Storm Overflow Spill Reduction: The Hydraulics, Storage Sizing and Control That Actually Cut Spills

What causes a storm overflow to spill?

A combined sewer carries foul flow and surface water in one pipe. In dry weather it runs at a small fraction of capacity. In rain, runoff from impermeable surfaces enters at a rate set by rainfall intensity and connected area, and the flow can rise by one to two orders of magnitude within minutes.

Consider a catchment of 1 km² with 60% impermeable cover serving 12,000 people. Dry weather flow at 150 l/head·d is 1,800 m³/d = 75 m³/h. A 6 mm/h storm generates 0.6 × 0.006 × 10⁶ = 3,600 m³/h — 48 times dry weather flow. If the pass-forward is set at six times DWF (450 m³/h), the surplus is 3,225 m³/h, and any storage smaller than that fills in under an hour.

This is why spill frequency is dominated by small, frequent storms rather than by extreme events. The design question is never how to pass the hundred-year storm; it is how many of the ordinary storms in a typical year can be captured.

How the pass-forward flow is set: Formula A

In England and Wales the flow that must be passed forward to full treatment before an overflow may operate is conventionally derived from Formula A, written into the permit for the works or the overflow.

Two consequences follow directly. First, the pass-forward is proportional to population, while the storm inflow is proportional to impermeable area. Urban catchments where area has grown faster than population — car parks, extensions, paved gardens, new estates with high impermeable fraction — drift steadily towards more frequent spilling even with no change to the sewer or the works. Second, raising the pass-forward has a cost at the works: every additional cubic metre per hour of FFT must be screened, settled, treated biologically and settled again, and the secondary stage is what makes it expensive.

The interaction with works capacity is exactly the arithmetic that drove the Beckton expansion described in our article on the UK largest treatment works.

Sizing storage: from rainfall statistics to a tank volume

Storage sizing is a statistical exercise, not a single-storm calculation. The correct method uses a long continuous rainfall series — ideally 10–30 years at 5-minute resolution — run through a hydraulic model of the catchment, counting spills for each candidate storage volume.

The result is always the same shape: a steeply diminishing return curve. For a typical UK urban catchment, indicative behaviour is:

Figures are illustrative and catchment-specific; the shape, not the values, is the transferable lesson. Halving spills from 40 to 20 needs modest storage; halving again from 4 to 2 needs as much storage as the entire first reduction. Cost per spill avoided rises by roughly an order of magnitude between the two ends of that curve, which is why credible programmes sequence cheap catchments first and reserve very large tunnels for the few places with no alternative.

Why emptying time matters as much as volume

A storage tank is only available if it is empty when the next storm arrives. The controlling parameter is therefore the ratio of the inter-event dry period to the emptying time.

Storage empties at the surplus of pass-forward capacity over the current inflow — typically a small number. A 2,400 m³ tank emptied at 100 m³/h takes 24 hours. UK rainfall in winter frequently produces successive frontal systems 12–24 hours apart, so a tank sized on a single-storm basis is routinely still half full when the second storm lands, and it spills on an event it was nominally designed to contain.

Three design responses follow: size the emptying pumps against the inter-event statistics rather than an arbitrary drain-down period; return stored flow to the works during the recession, when the works has spare capacity, rather than after it; and, where the works is the constraint, provide the storage at the works rather than in the network so that return is instantaneous. Sizing the works-side buffer uses the same logic as any industrial flow equalisation design.

Separation: the only lever that removes volume permanently

Storage defers a spill; separation prevents the water from ever entering the combined system. Removing 1 hectare of impermeable area from a combined sewer removes, in a 6 mm/h storm, 0.9 × 0.006 × 10,000 = 54 m³/h of inflow permanently, in every storm, for the life of the asset, with no pumping energy and no operational attention.

• Highway and roof disconnection to soakaways, swales or separate surface water sewers.
• Permeable paving and infiltration where ground conditions and contamination risk permit.
• Attenuation at source — green roofs, rain gardens, storage crates — which reduces peak rather than volume, but peak is what triggers the spill.
• Infiltration reduction in the sewer itself: groundwater entering through defective joints can be a third or more of wet-weather flow in some catchments, and is often cheaper to seal than to treat.

The barrier to separation is rarely engineering; it is that the assets belong to highway authorities, developers and householders rather than to the water company, so delivery is a coordination problem. Where that coordination is achievable, cost per spill avoided is typically far below the tank equivalent.

Real-time control: using storage that already exists

Most sewer networks contain substantial in-pipe storage that is never used, because local weir levels cause one part of the catchment to spill while another remains half empty. Real-time control (RTC) uses actuated penstocks, controlled pumping and flow and level instrumentation to redistribute flow across the network, filling every available volume before any overflow operates.

The formal statement is a constrained optimisation solved on a rolling horizon:

Reported reductions from network-wide RTC are commonly in the range 10–30% of spill volume at a small fraction of the capital cost of equivalent storage, because the storage is already built and paid for. The limits are honest ones: RTC cannot create capacity that does not exist, it depends on instrumentation that must be maintained in a hostile environment, and it introduces a control system whose failure mode must default safely to the passive hydraulic configuration. The same predictive control ideas applied to treatment plants are covered in our guide to digital twins and model predictive control.

What event duration monitoring actually measures

Event duration monitors record when an overflow is discharging, usually by detecting water level over the spill weir. The published statistic is a count of events and hours, not a volume or a load. That distinction has real engineering consequences.

• A 12/24 counting convention is typically applied: discharges separated by less than 12 hours are counted as one event, and an event of more than 12 hours counts as an additional event for each subsequent 24-hour period. Two schemes with identical volume can therefore report very different counts.
• Frequency and impact are not the same thing. Twenty short, highly dilute spills into a large river may have less ecological effect than one long spill into a chalk stream at low flow — yet the count penalises the former.
• Measurement artefacts matter. Sensor drift, weir fouling and backing-up from the receiving water all produce recorded events with no real discharge, so data quality assurance is part of the engineering.

Designing to reduce a count pushes towards capturing the many small events; designing to reduce load pushes towards capturing the first flush and the large events. Where a scheme is justified on ecology rather than on statistics, say so explicitly in the design basis, because the two objectives produce different optimum storage volumes.

Worked example: comparing four options on one catchment

Take the 1 km², 12,000-population catchment used above, currently spilling 38 times a year, with a target of 10.

No single option reaches the target economically, and that is the general finding. The efficient solution is almost always a combination: RTC first because it is cheapest and fastest, separation next because it is permanent, then the smallest tank that closes the remaining gap. Sizing the tank last, after the cheaper levers are counted, typically halves the volume required — and it is the tank volume that dominates both cost and embodied carbon. A properly staged scheme of this kind is the kind of work that process and hydraulic design services exist to structure.

Screening, aesthetics and what spills actually carry

Where a spill cannot be eliminated, its character can be improved. Overflow screening at 6 mm removes the gross solids — sanitary items, plastics, wipes — that dominate visual and amenity impact, and is a standard permit requirement on overflows discharging to sensitive or amenity waters.

Screening does not reduce dissolved or fine particulate load. A typical CSO discharge is dilute by the standards of crude sewage — often 50–150 mg/L BOD against 250–350 mg/L in dry weather — but arrives in a large volume, at high velocity, with a first flush that mobilises sewer sediment and can carry solids concentrations several times the dry weather value in the opening minutes. That first flush is precisely what modest storage captures best, which is a further argument for the small-tank-plus-control combination over doing nothing until a large tank is affordable.

For the equipment side of this, see our guides to screening and grit removal, both of which must be sized for storm rather than average flow.

How to size a storm overflow reduction scheme

1. Establish the baseline. Verify EDM data quality, then build or update a calibrated hydraulic model of the catchment against measured flow and level.
2. Define the objective. State whether the target is a spill count, a spill volume or an ecological outcome — they give different optimum designs.
3. Run continuous simulation. Use 10–30 years of rainfall at 5-minute resolution, not a design storm, and count spills for each candidate intervention.
4. Take the free capacity first. Quantify unused in-network storage and the reduction achievable with real-time control before sizing any new asset.
5. Quantify separation potential. Identify disconnectable impermeable area and infiltration hot spots; convert each hectare to m³/h of inflow removed.
6. Size the residual storage. Size the tank to close only the remaining gap, and check the emptying time against inter-event dry period statistics.
7. Check the works. Confirm the treatment works can accept the returned volume and any increased FFT without breaching its own consent.