Nitrogen dioxide modelling and model verification

Most air quality assessments that get sent back are sent back for one reason: the model was never verified against local monitoring, or the verification was done badly and reported in a single sentence. Everything upstream of that step can be competent and the conclusions still carry no weight, because an unverified road model tells you what the equations produce rather than what the street does.

Verification is where a reviewer starts. It should be where the author starts too, in the sense of knowing before any modelling begins whether usable local monitoring data exists.

When modelling is needed rather than screening

Screening comes first. The planning guidance published by the Institute of Air Quality Management with Environmental Protection UK sets indicative criteria based on the change in traffic the development causes: a change of more than 100 light duty vehicle movements a day inside or adjacent to an air quality management area, or more than 500 elsewhere, with much lower figures for heavy duty vehicles. A change in speed band or a new car park of any size can also bring a site into scope.

Below those figures, a short screening note is the proportionate answer. Set out the traffic change from the transport evidence, compare it against the criteria, name the nearest receptors and conclude. Padding a screening exercise into a modelled assessment wastes money and irritates officers.

Above them, or where the authority asks regardless, a dispersion model is the tool. Say in the scoping stage which model you propose and why, and agree it with environmental health before running anything.

Selecting receptors and their heights

Receptors go where relevant exposure occurs, judged against the averaging period of the objective. Annual mean objectives apply at facades of residential property, schools, hospitals and similar locations where people are regularly present over a year. They do not apply in the middle of a carriageway or on a pavement where nobody lives.

Height matters and is often handled carelessly. Ground floor receptors are normally modelled at about 1.5 m, with upper floors at successive storey heights where residential accommodation exists above. On a scheme introducing flats above a busy road, the upper floor results are the interesting ones, and omitting them is a common way to make a result look better than it is.

Number and position need thought too. Modelling three receptors on the most exposed facade and none on the flank produces a picture that is easy to attack. Include the existing worst case receptors, the proposed new exposure, and enough intermediate points to show the gradient away from the road.

Traffic data the model depends on

The model is a traffic model with chemistry attached. Annual average daily traffic by link, heavy duty percentage, average speed and queueing behaviour drive the emissions, and those figures should come from the same transport evidence supporting the application rather than from a separate exercise.

Three inputs cause most of the trouble. Speed, because a link modelled at free flow speed when it queues every evening understates emissions. Fleet composition, since the proportion of heavy vehicles and the assumed fleet year change the emission factors substantially. And the future year, where projected fleet improvement can do more work in the model than any mitigation the scheme offers.

State the emission factor tool and version used, the fleet year assumed for each scenario, and whether fleet projections were applied. A reviewer who cannot reproduce the emission inputs cannot accept the outputs.

Background maps and how they are applied

Background concentrations come from the national mapping published by Defra at one kilometre grid square resolution for the relevant year. Applying them raw alongside a modelled road contribution double counts the roads already represented in the background estimate.

The established remedy is sector removal, subtracting the contribution of the modelled roads from the background before adding the model output. Say explicitly whether you did this and for which sources. An assessment silent on the point invites the assumption that it did not.

Background maps are estimates on a coarse grid, and in a complex urban setting they can be some distance from reality. Where local monitoring sits in the same grid square and shows a different picture, discuss the discrepancy rather than ignoring it.

Verification against monitoring data

Verification compares modelled road contribution with measured road contribution at locations where monitoring exists, and derives an adjustment factor applied to the model output. The method is set out in the local air quality management technical guidance, and the mechanics are straightforward. The judgement is in the inputs.

Use monitoring sites that represent the modelled road environment. A diffusion tube in a back garden and a tube on the kerb of a canyon street will not verify the same thing. Diffusion tube data needs bias adjustment before use, and the factor applied should be stated with its source.

Report the number of verification sites, their locations on a plan, the measured and modelled values, the derived factor and the scatter around it. A factor derived from a single tube is weak evidence and should be described as such rather than presented with the confidence of a calibrated instrument.

Where no local monitoring exists, say so plainly and explain what was done instead. A stated limitation is defensible. A silent one is not, and an Air Quality Assessment that hides this is the kind that gets returned.

Presenting uncertainty honestly

Every stage carries uncertainty: traffic inputs, emission factors, background estimates, meteorological data, the model itself and the verification factor. Compounding them into a single confident number and presenting it to two decimal places misleads the reader.

Better practice is a short section stating the principal sources of uncertainty, their likely direction, and sensitivity tests where a result sits close to an objective. If a receptor models at 39 micrograms per cubic metre against a 40 microgram objective, the honest statement is that the result is within the uncertainty of the method and the conclusion depends on assumptions that should be named.

Impact descriptors then apply the percentage change against the objective read together with the absolute concentration, so a small change at a receptor already near or above the objective is treated more seriously than the same change well below it. Set the table out and let it do the work.

The same discipline about stating model choices and their justification applies to dispersion work in other fields, including the modelling behind an Odour Assessment, and the habits transfer directly.

What reviewers check first

In rough order: whether verification was carried out, how many sites and where. Whether background double counting was addressed. The receptor heights, and whether upper floors were included. The traffic inputs and whether they match the transport evidence. The emission tool version and the fleet years assumed. The meteorological station and its suitability for the site. And whether results close to the objective were tested for sensitivity.

After that come the softer checks. Does the conclusion follow from the tables. Are adverse results discussed or buried. Is the mitigation tied to the assessed effect or lifted from a standard list. Does the construction phase assessment sit alongside, as it does with construction noise controls under a section 61 consent.

For specialist technical reports, qualified professional input may be required, and dispersion modelling with verification is firmly in that category. This is general guidance, not legal advice. Check the position with your local planning authority and its environmental health team before relying on it.

Enter your postcode on PlanScope to see which environmental reports a scheme is likely to need. PlanScope provides AI-assisted draft reports for planning applications. Reports should be reviewed before submission. We do not guarantee planning approval.