Verifying a NOx model against local monitoring data

Model verification for nitrogen dioxide is the step that decides whether the rest of an air quality chapter means anything. A dispersion model predicts road NOx from traffic data and meteorology, those predictions carry systematic error, and verification measures that error against local monitoring and corrects for it. Skip it, or do it badly, and the predicted concentrations at your receptors are arithmetic rather than evidence. Reviewers reading an assessment for a local planning authority should treat the verification section as the first place to look.

Why the correction is needed at all

Road source models underpredict in most United Kingdom applications, and the reasons are structural rather than careless. Emission factors are derived from test cycles that do not match a congested urban junction. Traffic counts are projected. Street canyon effects, roadside turbulence and queue behaviour are simplified. Background concentrations come from mapped estimates at 1 km resolution.

None of that is fixable by better model inputs alone, so Defra's local air quality management technical guidance sets out a verification procedure: compare modelled with measured road contributions at monitoring locations, derive an adjustment factor, and apply it to modelled road NOx before recombining with background.

The correction is applied to the road NOx contribution, not to total NO2. Getting that order wrong is the single most common error in submitted assessments, and it is detectable from the reported numbers alone.

Selecting monitoring locations you can defend

Verification is only as good as the monitoring data behind it. The sites used have to be comparable with the modelled receptors in the ways that matter.

• Same calendar year as the model base year, with full data capture or a documented annualisation.
• Diffusion tube results bias adjusted, stating whether the factor came from a local co-location study with a reference analyser or from the national bias adjustment spreadsheet, and which version.
• Distance from the kerb measured rather than estimated, and modelled at the same distance and height as the tube.
• A spread of road types, since one factor derived entirely from open rural roads will not serve a canyon.
• Enough sites to see a pattern. Verification against a single tube is not verification, it is a coincidence.

Where the host authority has few tubes, look at the neighbouring authority's network and at automatic monitoring stations within a comparable setting. An assessment that says no local monitoring exists, in a borough with a published annual status report, invites the reviewer to check.

Separating the road contribution from background

The arithmetic is unforgiving about double counting, and this is where careful work shows.

Take the bias adjusted measured NO2 at each site. Convert to measured total NOx, then subtract the background NOx for that grid square to leave the measured road NOx. Model the road NOx at the same point from your source inventory. The two road NOx figures are the comparison.

The step practitioners miss is sector removal. Defra's background maps include a road traffic component, so where your model already represents those roads explicitly, the corresponding sector has to be stripped out of the mapped background before subtraction. Leave it in and you inflate the measured road contribution, which flatters the model and produces an adjustment factor that is too low.

Conversion between NOx and NO2 uses Defra's calculator, and the version, region and year selected all change the answer. State them. A reviewer who cannot tell which calculator version was used cannot reproduce the figure, and reproducibility is the point of the exercise.

Deriving and applying the adjustment factor

With paired measured and modelled road NOx values, the factor is the ratio between them, usually taken as the gradient of a regression through the origin across all sites rather than an average of individual ratios.

Apply the resulting factor to modelled road NOx at every receptor, add the background, and convert the total to NO2. Then report the same table twice: modelled against measured before adjustment, and after. The second table shows whether the correction has actually improved agreement or merely moved the error around.

Where sites split into clear groups, a single factor may be the wrong answer. Separate factors for canyon and open road sections are defensible if the grouping is explained and each group has enough sites to support a gradient. Separate factors chosen to make individual receptors comply are not defensible, and a reviewer should ask why the grouping follows the receptors rather than the geography.

How much adjustment is too much

A factor close to one suggests the model is performing well. A large factor is a signal, and the proper response is to interrogate the inputs rather than to apply the number and move on.

Defra's guidance also asks for a measure of agreement rather than a single ratio. Root mean square error between modelled and measured concentrations, compared against the annual mean nitrogen dioxide objective of 40 micrograms per cubic metre, gives the reviewer a sense of scale: an error of a few micrograms is a different proposition from an error approaching a quarter of the objective, and the guidance expects justification as the error grows.

Honest reporting of a poor fit is better than a suspiciously tidy one. I have more confidence in an assessment that says agreement was imperfect at two of five sites and explains why than in one reporting near-perfect agreement at every tube.

Reporting verification so it can be checked

The verification section should be reproducible by a competent reader with the same inputs. That means a table of monitoring sites with grid references, site type, distance from kerb, raw and bias adjusted concentrations, data capture and annualisation. It also means the background values used and the sectors removed, the calculator version, the modelled and measured road NOx, the derived factor, and the before and after comparison.

Put it in the main report rather than an appendix nobody opens. The same discipline applies to measurement based assessments generally, as with the duration and position questions discussed in our note on unattended noise monitoring.

Where the assessment supports a planning application, the verification evidence belongs with the submitted Air Quality Assessment rather than being held back for the consultation response.

What a reviewer should challenge

Working through a submitted chapter, these are the questions that most often change the conclusion.

1. Was the factor applied to road NOx, or wrongly to total NO2?
2. Was the road sector removed from the mapped background before deriving measured road NOx?
3. Which bias adjustment factor was used, from which source, and for which year?
4. Do the modelled verification points match the tube positions in distance and height?
5. Does the base year of the traffic data match the monitoring year?
6. How many sites support the factor, and does the scatter justify a single gradient?
7. Is the calculator version stated, with the region and year?
8. If no verification was carried out, what is the stated justification, and does the authority's own monitoring contradict it?

Those questions apply equally to assessments of combustion plant, where stack height and dispersion assumptions carry the same burden of proof, as covered in our note on chimney heights and biomass boilers, and to small commercial schemes where traffic generation drives the result, as in our piece on air quality assessment for a drive-through unit.

A reviewer is entitled to ask for the model files. Competence in this field is judged on method rather than conclusion, and the directory of professional reviewers exists for exactly this sort of sign-off.

Draft the surrounding assessment on PlanScope, then have the modelling and verification prepared and checked by a suitably qualified air quality specialist. PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval. For specialist technical reports, qualified professional input may be required.