Bias adjustment and annualisation of NO2 diffusion tube data
Nitrogen dioxide diffusion tube results need bias adjustment, annualisation and sometimes distance correction before they can carry an air quality baseline. Here is the order and the evidence to show.
Raw nitrogen dioxide diffusion tube results are not measurements of ambient concentration. They are laboratory results from a passive sampler that reads high or low depending on the tube preparation method, the laboratory, the exposure period and the site. Diffusion tube bias adjustment converts that raw number into something comparable with the annual mean objective, and annualisation converts a partial year into a full one. A baseline that skips either step is not defensible, and a reviewer should send it back.
Why raw tube data cannot go straight into a baseline
A diffusion tube collects nitrogen dioxide by molecular diffusion onto a triethanolamine-coated grid over a set exposure period, usually a calendar month. The result depends on the diffusion path length, the preparation method, the handling and the analysis. Comparison studies against chemiluminescence reference analysers consistently show tubes departing from the reference, and the direction and size of that departure vary by laboratory and preparation.
That is bias, and it is systematic rather than random. Averaging more tubes does not remove it.
Three separate corrections then apply, and they are often confused with each other.
- Bias adjustment corrects the systematic departure of the tube method from the reference method.
- Annualisation corrects a monitoring period shorter than a full calendar year, or one with insufficient data capture, to an estimated annual mean.
- Distance correction moves a concentration from the monitoring location to the relevant receptor facade, where the two are not in the same place.
A fourth step, model verification, uses monitoring data to adjust a dispersion model's output. It is not bias adjustment and the adjusted monitoring data is the input to it, never the other way round.
Bias adjustment factors and where they come from
Two sources of factor exist, and the hierarchy between them matters.
A local factor comes from a co-location study at a site in the study area where triplicate tubes are exposed alongside a continuous reference analyser over the monitoring period. The factor is the ratio of the analyser mean to the tube mean across the co-located periods. It reflects the actual laboratory, preparation method, handling and local conditions of that specific survey, which is why the technical guidance prefers it where a valid study exists.
A national factor comes from the spreadsheet Defra maintains from co-location studies submitted by authorities across the country, grouped by laboratory and preparation method. Use it where no local study is available, and state which version of the spreadsheet was used, because the values are revised as new studies are added.
An assessment that applies a factor without saying which laboratory and preparation method it relates to has not shown its working. State the laboratory, the preparation method, the factor, the source of the factor, and the number of co-location studies or periods behind it.
| Situation | Factor to use | What to state in the report |
|---|---|---|
| Local co-location study with sufficient valid periods | Local factor | Analyser site, number of valid periods, triplicate precision, derived factor |
| No local study, tubes from a known laboratory and method | National factor | Spreadsheet version, laboratory, preparation method, number of studies |
| Mixed laboratories across the survey | Separate factors per laboratory | Which tubes fall under which factor, and why the survey was split |
| Tubes from an unlisted laboratory or method | Treat as indicative only | The limitation, and why the data is still presented |
Triplicate precision and data capture
Precision is separate from bias. Triplicate tubes exposed at the same location should agree with each other, and the spread between them, expressed as a coefficient of variation, is the measure. Poor precision points at handling, preparation or exposure problems, and it undermines the co-location factor derived from that set. The technical guidance sets the thresholds at which precision is judged good or poor, and a reviewer should look for the coefficient of variation to be reported rather than asserted.
Data capture is the other validity test. An annual mean calculated from a partial data set is only reliable above a capture rate of about three quarters of the year, and below that the guidance calls for annualisation instead. Lost tubes, vandalised tubes, tubes exposed for the wrong duration and tubes returned outside their holding time all reduce capture, and the reasons should be tabulated rather than hidden in a total.
Watch for exposure periods that drift. A tube nominally exposed for a month but actually out for six weeks is not invalid, but the concentration must be weighted by the actual exposure duration when the annual mean is calculated. Simple arithmetic means of monthly results with unequal exposure lengths are a common error in baseline chapters.
Annualising a short monitoring period
Where monitoring runs for part of a year, the measured mean reflects the seasons it covered. Winter periods run higher for nitrogen dioxide. A survey from October to February and a survey from April to August will produce different answers for the same location.
Annualisation uses the ratio between the annual mean and the period mean at nearby long-term continuous background sites, applied to the short-period result. The guidance sets out how many reference sites to use, how far away they may sit, and how to combine the ratios, and the usual practice is to use several background sites rather than one so that a single unrepresentative station cannot distort the answer.
Two things to check in a submitted baseline. That the reference sites are true background locations rather than roadside, because a roadside station's seasonal profile is not transferable. And that the annualisation was applied to the bias-adjusted result rather than the raw one, since applying it in the wrong order produces a different number.
State the annualisation factor, the reference sites used, and the annualised result alongside the measured one. A reviewer who can see all three can accept the conclusion. A single annualised figure with no working cannot be checked.
Presenting uncertainty honestly
Every step adds uncertainty, and the honest response is to say where the result sits relative to the objective rather than to present a single number as fact.
The annual mean objective for nitrogen dioxide is 40 micrograms per cubic metre, with a separate one-hour objective. A bias-adjusted, annualised result of 38 at a facade is not a comfortable pass. It is a result within the uncertainty band of the objective, and the assessment should say so and consider what the scheme adds on top. A result of 22 at the same facade needs no such treatment.
Present the bias adjustment factor with the confidence interval where a local study provides one. Present the range produced by applying the upper and lower bounds where the conclusion is marginal. And identify which receptors change category under those bounds, because that is the information a decision-maker needs.
The same discipline applies to the monitoring evidence behind other environmental topics. Duration, position and validity decide whether a data set can carry an argument, as set out in our note on unattended noise monitoring.
Choosing between local and national data
Three practical judgements decide which data a baseline should rest on.
- Is there a local authority tube network with published, bias-adjusted results near the site, and are the tube locations relevant to the receptors of concern. Published network data with several years of history usually beats a three-month bespoke survey.
- Is there a continuous analyser within a reasonable distance with high data capture. If so, it anchors both the annualisation and the model verification.
- Does the site sit inside an air quality management area, in which case the declaring authority's own review and assessment reports carry the local trend evidence and should be cited directly.
Where a bespoke survey is commissioned anyway, run it long enough to be annualised properly, use a single laboratory throughout, include a triplicate location, and co-locate with an analyser if one is within reach. Those four decisions cost very little at the start and determine whether the data survives review.
An Air Quality Assessment that sets all of this out explicitly is far harder to challenge than one that presents adjusted numbers with no provenance, which is the difference between a baseline that supports a conclusion and one that merely asserts it. The same structure applies to smaller schemes such as a drive-through unit, where the monitoring evidence is thinner and the working therefore matters more.
PlanScope drafts the assessment structure and flags missing site data rather than inventing it, and for specialist technical reports, qualified professional input may be required. Find a reviewer in the professional directory to check the adjusted data before submission.
PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval.
Frequently asked questions
What is a diffusion tube bias adjustment factor?
It is the ratio used to correct the systematic departure of the passive tube method from a chemiluminescence reference analyser. It varies by laboratory and preparation method, so a factor quoted without naming both is meaningless. Factors come either from a local co-location study or from the national spreadsheet Defra maintains from submitted studies.
Should a local or national bias factor be used?
A local factor is preferred where a valid co-location study exists in the study area, because it reflects the actual laboratory, preparation and handling of that survey. Use the national spreadsheet where no local study is available, and state which version was used, since the values are revised as new studies are added.
When does monitoring data need annualising?
Where the survey covers less than a full calendar year, or where data capture falls below roughly three quarters of the year. Annualisation applies the ratio between annual and period means at nearby long-term background monitoring stations. Apply it to the bias-adjusted result rather than the raw one, because the order changes the answer.
Is model verification the same as bias adjustment?
No. Bias adjustment corrects the monitoring data. Model verification adjusts a dispersion model's predictions using that corrected monitoring data. The corrected monitoring is the input to verification, never the output of it. Reports that conflate the two usually end up double-counting an adjustment.
What should a reviewer look for in a tube baseline?
The laboratory and preparation method, the factor and its source, the number of co-location periods, triplicate precision expressed as a coefficient of variation, data capture with reasons for losses, exposure-weighted means, the annualisation factor with its reference sites, and the result presented against the objective with its uncertainty stated.
Sources and further reading
This article is general guidance about the planning system in England and, where stated, the rest of the UK. It is not legal or professional advice. Check your local planning authority’s validation requirements and take professional advice for your specific site.