Storm overflow performance: what a 35% fall in spills actually measures

TL;DR. Monitored storm overflow spills in England fell 35% in 2025, and the Environment Agency says much of that reflects unusually dry conditions after a wet 2024. A spill count is largely a rainfall measurement, so it is a poor guide to storm overflow performance. The more honest figure in the same dataset is the 20% fall in average duration per spill, because how fast an asset clears is far less weather-dependent than how often it starts.

Key takeaways

• England recorded 291,492 monitored storm overflow spills in 2025, a 35% fall on 2024, with total duration down 48%.
• The Environment Agency states that much of the improvement reflects unusually dry conditions in 2025 following a particularly wet 2024.
• Spill activity is strongly linked to rainfall, so a spill count without a rainfall denominator does not measure storm overflow performance.
• Average spills per overflow fell from 31.8 in 2024 to 20.5 in 2025.
• Average duration per spill fell 20%, a figure much less sensitive to how much it rained.
• Every storm overflow in England now has an event duration monitor fitted, giving the most complete national picture to date.

Last updated: 18 August 2026

What did the 2025 storm overflow data actually show?

It showed a large fall on every headline measure. Environment Agency monitoring data recorded 291,492 spill events in England in 2025, a 35% reduction on the previous year, with total monitored spill duration down 48% and individual company reductions ranging between 40% and 70%.

The coverage behind it is now complete. Every storm overflow in England has an event duration monitor fitted, which the Environment Agency describes as the most complete national picture to date. That matters, because the dataset is no longer limited by which assets happen to be instrumented. Coverage is solved. Interpretation, and therefore any honest read of storm overflow performance, is not.

Why is a falling spill count a poor measure of storm overflow performance?

Because a storm overflow operates when it rains, so counting operations largely counts rainfall. The Environment Agency says this plainly in the same release: much of this improvement reflects unusually dry conditions in 2025 following a particularly wet 2024, and that spill activity is strongly linked to rainfall levels.

That is not a criticism of the data. It is a statement about what the number is. A spill count is an output of two inputs, how much rain fell and how the network handled it, and the published figure does not separate them. Read as storm overflow performance it flatters a dry year and punishes a wet one, regardless of what anyone did to the asset.

The practical consequence is that year-on-year comparisons of spill counts tell you very little unless the two years had similar rainfall. Comparing 2025 to 2024 compares a dry year to a wet one, which is why the fall is large and why it should not be read as a change in storm overflow performance. Statutory obligations continue regardless: section 82 of the Environment Act 2021 requires continuous water quality monitoring upstream and downstream of storm overflows and treatment works, and the Storm Overflows Discharge Reduction Plan sets targets that a dry year does not satisfy.

Which number in the dataset is closer to real performance?

Average duration per spill. It fell 20% in 2025, and it is the figure in that release least distorted by the weather. For storm overflow performance, rainfall largely determines whether an overflow starts. What happens next, how long it runs before the network recovers enough to stop it, depends far more on capacity, blockages, pump availability and how quickly anyone reacted.

The distinction is worth holding onto. Count is a question about frequency and is dominated by rainfall. Duration is a question about recovery and is dominated by the asset. Both are in the published data, but only one of them is a reasonable proxy for storm overflow performance, and it is not the one in the headlines.

Spills per overflow behaves similarly. The average fell from 31.8 in 2024 to 20.5 in 2025, which normalises for how many assets exist but still not for how much it rained. It is a better denominator than a national total and still not a measure of storm overflow performance.

How do you measure storm overflow performance on your own assets?

By recording level and rainfall together, so every event has its own denominator. Storm overflow performance is a local question before it is a national one. An event duration monitor answers whether a spill happened and for how long. It does not tell you how much rain caused it, which means a single site’s year-on-year trend has the same interpretation problem as the national one until rainfall sits alongside it.

Level is where this starts. An AQUAIOT Radar smart water level sensor reads to within ±2 mm using 60 GHz mmWave, non-contact, through the foam, vapour and condensation that make an enclosed chamber hostile to other techniques. Because nothing sits in the flow there is nothing to foul, which matters on an asset that only operates in the worst conditions. ATEX variants are available where the chamber classification requires them.

What that gives you is the shape of each event rather than a binary. Time to rise, peak level, and time to recover after the rain stops. Recovery time is the local version of duration per spill, and it is the storm overflow performance signal that moves when a screen blinds, a pump degrades or a downstream restriction develops, which is precisely what a spill count hides.

Two events with comparable rainfall and different recovery times is evidence of something changing on the asset. Two events with different rainfall and different durations is just weather. Continuous level data is what lets you tell them apart, and it is the same data that underpins sewer and overflow monitoring generally.

For background on what the national monitors do and do not capture, our explainer on event duration monitoring covers the mechanics, and pumping station monitoring covers the assets whose recovery time this argument depends on.

Frequently asked questions

Did storm overflow performance really improve in 2025?

Partly, but the published fall overstates it. The Environment Agency attributes much of the 35% reduction in spills to unusually dry conditions in 2025 after a wet 2024. The 20% fall in average duration per spill is the stronger evidence of genuine improvement, because duration depends more on the asset than on the rainfall.

What is the difference between spill count and spill duration?

Count is how many times an overflow operated, which is largely driven by how often rainfall exceeded the network’s capacity. Duration is how long each operation lasted, which reflects how quickly the network recovered. Count answers how often it rained hard. Duration answers how well the asset coped.

Does every storm overflow in England have a monitor?

Yes. Event duration monitors are now fitted to every storm overflow in England, which the Environment Agency describes as the most complete national picture to date. Coverage is no longer the limitation on the dataset, but an event duration monitor still records only whether and for how long a spill occurred, not the rainfall that caused it.

What should a site record alongside spill data?

Rainfall and continuous level. Rainfall gives each event a denominator so events can be compared fairly, and continuous level gives the shape of the event, including time to peak and time to recover. Recovery time under comparable rainfall is the clearest local signal of storm overflow performance, and the clearest sign that something on the asset has changed.

Reading next year’s number

2026 has been exceptionally dry across much of England, so the next set of national figures will very likely fall again. That will be reported as progress, and some of it will be genuine storm overflow performance. The way to know which part is to look past the count to duration, and on your own assets to recovery time under comparable rainfall.

If you want your own storm overflow performance data to survive that question, it needs level and rainfall in the same record. Speak to an AQUAIOT engineer about instrumenting the overflows that matter on your network.

By GP, Digital and IoT lead at AQUAIOT.