Skip to content
Air quality

What a diffusion tube network can and can't tell you

What NO2 diffusion tubes measure, what LAQM requires, and what a network of fixed points can and cannot tell you about the ground between them.

Almost every local authority in the UK runs its NO2 monitoring on passive diffusion tubes. They are cheap, they need no power, and they can go almost anywhere there is a lamppost. They are also the reason most authorities can say anything at all about NO2 outside the handful of sites with a reference analyser.

This page sets out what that network is doing under the statute, where its uncertainty actually sits, and — the part that gets least attention — what it can and cannot tell you about the ground between the tubes.

What the statute asks of the network

Under Part IV of the Environment Act 1995, section 82 requires every local authority to review air quality in its area, present and likely future. Section 83 requires designation of an Air Quality Management Area where the objectives are not being achieved or are unlikely to be. In England, section 83A then requires an Air Quality Action Plan. Section 85 gives the Secretary of State reserve powers, and LAQM Policy Guidance (revised May 2025) confirms that a ministerial direction under section 85(3) is available where an Annual Status Report is six months overdue. ASRs are due through the LAQM Portal by 30 June each year.

The NO2 objectives you are testing against are 40 µg/m³ as an annual mean, and 200 µg/m³ as a 1-hour mean not to be exceeded more than 18 times a year (LAQM.TG(22), Table 1-1; current version May 2025, v2.1).

A tube network can only speak to the first of those. Tubes integrate over roughly a month, so TG(22) para 7.202 is explicit that they are useful for the annual mean but "cannot be used to assess the number of hours greater than 200 µg/m³". The working substitute is para 7.97: exceedances of the 1-hour mean are unlikely where the annual mean is below 60 µg/m³ — a relationship derived predominantly at roadside and kerbside sites, and one TG(22) says should not be applied where industrial emissions dominate.

What has to happen before a tube reading means anything

Three processing steps stand between the laboratory result and a number you can put in an ASR.

Data capture and annualisation. Below 75% capture — about nine months on the Defra calendar — results must be annualised, and a minimum of three months is needed to do it at all (TG(22) 7.202, 7.216).

Bias adjustment. Tubes systematically over- or under-read against the chemiluminescence reference method (BS EN 14211). You correct with either a local factor from a co-location study — triplicate tubes within 1 m of the analyser inlet — or the national factor for your laboratory and preparation method, never both (TG(22) 7.220–7.227, Box 7-13). TG(22) 7.229 warns against deriving a factor at a background site and applying it to roadside results, or the reverse.

Precision. Bias can be corrected. Precision cannot. TG(22) 7.221 sets the threshold: good precision means the coefficient of variation across triplicates is under 20% for eight or more periods and averages under 10%. Defra FAQ 144 makes the corollary plain — with a single tube per site, you cannot identify an outlier at all, and one contaminated tube is a lost month.

Where the residual uncertainty comes from is now well characterised. NPL's 2025 co-location work found unsheltered Palmes tubes carried an overall measurement bias of 13–24% against reference, falling to 0–8% when the same tubes were mounted in a shelter, and to 1.5–7% for the wind-protected UK Urban NO2 Network design. Their uncertainty model found the single largest contributor was wind bias (Medland et al., Atmospheric Environment 352, 2025). TG(22) 7.206 now recommends authorities consider wind caps for this reason.

For scale: the data quality objective for indicative measurement under Directive 2008/50/EC Annex I is 25% uncertainty, against 15% for fixed reference measurement.

The spatial question

Everything above concerns the number at the tube. The harder question is what happens ten metres away.

NO2 near roads falls off steeply. Defra's own evidence base — NO2 Concentrations and Distance from Roads (Air Quality Consultants for Defra, July 2008) — found the road contribution declines essentially linearly against the natural log of distance from the kerb, from 0.1 m out to 140 m, and that the DMRB screening model and Caline-4 both understate that decline. TG(22) 7.494 makes the modelling consequence explicit: coarse spatial resolution "is likely to miss hot spots", which is why 7.493 asks for 5–10 m grid spacing within 30–50 m of roads.

So when you use the fall-off calculator to move a kerbside tube back to a facade, you are not measuring the facade. You are applying a national empirical curve. The Defra report says so directly: results derived this way "will have a greater uncertainty than the measured data". TG(22) 7.84 reflects the same caution, recommending distance correction for any site above 36 µg/m³ to account for the inherent uncertainty in tube data.

Interpolating between tubes carries a heavier set of assumptions: that the concentration field is smooth between the points; that no unmeasured source sits in the gap; that each tube is representative of its immediate surroundings rather than of its own microenvironment; and that the form of the surface between points resembles the model you fitted. The first and third are the ones that fail. Mobile measurement in Oakland found NO2 and black carbon varying persistently by 5–8× within a single city block (Apte et al., Environ. Sci. Technol. 51(12), 2017).

A tube network is a set of points. Nothing in it observes the surface between them. That is not a flaw in the method — it is what the method is.

Where mobile measurement fits, honestly

Street-level measurement from a moving vehicle is one way to add spatial detail between fixed points. Make Sense is building toward this. Our first vehicle system is in build. It is not MCERTS certified, and it is not a reference method.

That matters concretely. Indicative data does not replace bias-adjusted tube data or reference analyser data for anything statutory: not an AQMA declaration, not an ASR return, not a revocation case. What it can legitimately support is the work upstream of that — screening a corridor before you commit tubes to it, seeing the shape of exposure along a street rather than at one point on it, and deciding where the next tube or the next reference monitor should go.

And it has its own real limits, which do not go away with better hardware:

  • Dwell time. A vehicle is at any given point for seconds. A tube is there for a month.
  • Meteorology. A single pass captures one wind direction, one temperature, one traffic state. Wind speed is already the dominant uncertainty in a month-long passive measurement; it is far more consequential in a two-second one.
  • The annual mean problem. Comparing passing measurements to a 40 µg/m³ annual mean requires many repeat visits, not one. The Oakland work reached estimates precise to within roughly 10–20% only after sampling each block on an average of about 31 different days across a year.
  • Standards coverage. CEN/TS 17660-1:2021 classifies sensor system performance for gaseous pollutants at fixed sites, and PAS 4023 is a code of practice for sensor systems in outdoor ambient air. Neither is written for mobile deployment. TG(22) discusses low-cost sensors at 7.257–7.262 and does not address vehicle-mounted measurement at all.

Anyone telling you mobile measurement replaces a tube network is selling you something. The honest claim is narrower: it observes a different thing — the gaps — and the two are complementary only if the provenance and confidence of each reading travel with it.

If you want to talk through where indicative spatial data would and would not help in your area, get in touch. We would rather have that conversation now, while the system is being built, than after.

Sources

Last reviewed 12 August 2026

Related
Talk to us about this