Air quality

What a heatwave does to air quality

During the August 2026 heatwave Defra declared a High ozone episode. Why ozone peaks in the countryside, and why a tube network never saw it.

Between 9 and 15 August 2026 England had a heatwave. It has ended: the Met Office has no heat warnings in force, and all nine UKHSA regions returned to green on 18 August.

Something else happened in the same week that got almost no coverage. On 13 and 14 August, Defra issued air pollution bulletins for ozone. Concentrations passed the public information threshold of 180 µg/m³, reaching 197 µg/m³ at London Honor Oak Park on the 13th and 204 µg/m³ at St Osyth in Essex on the 14th — a Daily Air Quality Index of 8, in the High band (Defra UK-AIR, 13 August, 14 August).

That episode is over too. Defra currently reports no alerts and forecasts Low across the UK. But it is worth understanding what happened, because it is the clearest recent illustration of a monitoring problem most authorities have.

Ozone is not emitted. It is made.

Nothing has an ozone exhaust pipe. Ground-level ozone forms photochemically, from nitrogen oxides and volatile organic compounds reacting under strong sunlight. Hot, still, anticyclonic weather — exactly the conditions that produce a heatwave — gives that reaction long sunny days and no wind to disperse the product. Heatwave and ozone episode are not two coincidences. They are the same weather.

The counterintuitive part: ozone is lowest where the traffic is

This is the bit that catches people out, and it matters for where you put monitors.

Fresh vehicle exhaust contains nitric oxide, and nitric oxide destroys ozone on contact:

O₃ + NO → NO₂ + O₂

So a busy roadside — full of fresh NO — actively scavenges ozone. Concentrations there are suppressed. Move away from the traffic and there is nothing left to consume it, so ozone accumulates. During the August episode the exceedances were recorded at London sites and at Chilbolton, Ladybower, Wicken Fen, Yarner Wood and Cwmbran. Rural sites. The chemistry predicts precisely that pattern.

Two consequences follow, and both are uncomfortable.

That reaction does not remove pollution — it converts it. The NO becomes NO₂. What is conserved across the reaction is the sum, sometimes written as oxidant, OX = O₃ + NO₂. A roadside monitor showing low ozone is not showing clean air.

Reducing urban NOx tends to raise urban ozone. Less NO means less ozone scavenging. This is well established and it is an awkward result for an authority whose air quality action plan is built, correctly, around cutting NOx. It does not make that work wrong. It does mean the ozone number can move the other way, and someone will eventually ask why.

What else changes in the heat

PM2.5 rises, and the composition shifts: in hot conditions the increase is almost entirely organic matter — secondary organic aerosol formed in the same photochemistry — with contributions from dry soil dust and, in some years, wildfire smoke.

One correction worth making, because it circulates every summer: nitrate-driven particulate episodes are not heat events. Ammonium nitrate is semi-volatile and evaporates as temperature rises. Those episodes are a cool, still, early-spring phenomenon, usually agricultural. Attributing them to a heatwave gets the season and the source wrong.

Why a diffusion tube network saw none of this

A passive NO₂ diffusion tube integrates over roughly a month. The August ozone episode lasted about 48 hours.

Even setting aside that tubes measure NO₂ rather than ozone, the time resolution makes the episode structurally invisible. Defra's own guidance is explicit that tubes are useful for the annual mean and cannot be used to assess short-period exceedances — the point we cover in more detail in what a diffusion tube network can and can't tell you.

So an authority relying on tubes alone had a two-day High air quality episode across its area and no instrument capable of registering it. Not through any failure of process. The method simply does not have the temporal resolution, and it was never claimed to.

What would have seen it: a reference or demonstrated-equivalent analyser, of which there are few, or Defra's national network — which is why the episode was declared nationally rather than locally.

What this is, and is not, an argument for

It would be easy to finish this page by suggesting that more sensors are the answer. That is not quite the argument, and it is worth being precise about why.

An episode like this is regional and meteorological. It is driven by a synoptic weather pattern across hundreds of kilometres, not by a hotspot on one street. Dense spatial sampling is the right tool for spatial variation — where concentrations differ between one road and the next. It is the wrong tool for an episode where the whole region is elevated at once and the national network already detects it well.

What denser measurement genuinely adds during an episode is exposure pattern: which streets, schools and routes sat at the high end while the region was elevated, and how that maps onto where people actually are. That is a different question from "was there an episode", and it is the one a local authority is usually being asked.

Where any of this data can legitimately be used is set out in what your air quality data is admissible for. Short version: indicative data can tell you where to look. It cannot be your compliance number.

Where Make Sense sits

Make Sense is building a multi-sensor pod that mounts to vehicles already making their rounds, intended to record air quality alongside road surface condition and building heat loss. The first vehicle system is in build. Nothing is deployed, we hold no certification — we are not MCERTS certified — and we have no accuracy figures to publish, because nothing has been validated.

We did not measure the August episode. We are writing about it because the chemistry is genuinely counterintuitive, the roadside-versus-rural inversion is not widely understood, and the resolution gap it exposed is the sort of thing worth knowing before the next one.

If you are thinking about what your monitoring did and did not see this month, we would be interested in the answer regardless of whether it leads anywhere with us. Get in touch.

Sources

Last reviewed 19 August 2026