Dispersion model input appendices and what to check

The dispersion model inputs appendix is the part of an air quality or odour assessment a reviewer should read first and the part most often skipped. The modelled concentrations at the front of a report are a function of perhaps a dozen input choices, and several of them can move a result by a factor of two without any change to the scheme. Meteorological year, surface roughness, minimum Monin-Obukhov length, building downwash geometry and the emission rate derivation do most of that work. If the appendix does not record them, the result is unverifiable.

Why the appendix matters more than the result

A dispersion model is deterministic. Give it the same inputs and it returns the same answer, which means the only questions worth asking a modeller are about the inputs.

So the review sequence runs backwards through the document. Establish the model and its version, then the input parameters, then the emission rates, then the verification and adjustment, and only then look at the predicted concentrations. A reviewer who starts at the conclusions ends up arguing about judgement when the disagreement is about a number in an appendix.

Three red flags justify going no further until they are resolved. An appendix that gives no model version. An appendix that lists parameters without saying where they came from. And a report whose body quotes a single headline concentration with no indication of which meteorological year produced it.

The appendix should be capable of letting a competent third party rebuild the model run. That is the standard, and it is rarely met in full.

Meteorological data years and representativeness

Five years of hourly sequential data from a representative station is the usual expectation, and each year should be run separately rather than averaged.

Check four things. Which station, and its distance and direction from the site. Why that station is said to be representative, with reference to terrain, coastal influence and urban setting rather than just proximity. Which years were used, and whether any were rejected for data capture below the usual threshold. And whether the year used for model verification matches the year of the local monitoring data the verification is based against.

That last point is the one most often wrong. Verifying a 2023 model run against 2019 monitoring data embeds a mismatch that no amount of adjustment factor arithmetic fixes.

Report the result for each year and identify the worst case. A report that gives only the five year mean has concealed the variability that matters, because a planning decision is made against an annual mean objective that a single bad year can breach.

Where a site sits in a valley or close to the coast and the nearest station does not, say so and treat the uncertainty openly. An honest statement that met representativeness is the largest uncertainty in the assessment is better evidence than a confident number.

Surface roughness and terrain

Surface roughness length describes how the ground disturbs the wind profile, and the model needs two values: one for the meteorological site and one for the dispersion site. Using a single value for both is a common error.

Typical values run from around 0.1 m for open agricultural land, through intermediate figures for parkland and open suburbia, to around 1 m for large city centres. A reviewer should check that the dispersion site value matches the actual surroundings within a kilometre or so, and that the met site value matches the airfield or monitoring station rather than the development.

Roughness pushes results in a predictable direction. A higher value increases turbulence and dispersion, which lowers ground level concentrations near the source, so an overstated dispersion site roughness flatters the scheme. Ask for a sensitivity run at the adjacent value.

The minimum Monin-Obukhov length deserves its own question. In urban areas a minimum value is applied to stop the model generating unrealistically stable night-time conditions that never occur over a heat island, and the value chosen should reflect the urban scale. An urban site run with a rural minimum will produce pessimistic results, and a rural site run with a large urban minimum will produce optimistic ones.

Terrain matters where gradients are steep. Where slopes exceed roughly one in ten within the modelled domain, a complex terrain treatment should be used and the terrain file identified.

Building downwash and stack height

Where a stack discharges close to a building, the building's wake pulls the plume down and ground level concentrations near the source rise sharply. Models handle this through a simplified building geometry, usually a single effective building, so the choice of which building to represent is a judgement with consequences.

Check which buildings were included, their dimensions and the orientation used, and whether the modeller tested more than one representation. A downwash run that selects a small ancillary building over the larger adjacent block is understating the effect.

As a rule of thumb, a stack shorter than about two and a half times the height of the adjacent building will usually be affected by downwash, and a stack height assessment should demonstrate the height at which the effect becomes acceptable rather than asserting that the proposed height is adequate.

Two further inputs belong here. Efflux velocity and exit temperature drive plume rise, and both should be given at the operating condition rather than at the design maximum, because a plant running at part load has a lower velocity and a lower plume. And where the stack serves intermittent plant, the operating hours assumed should match the application description.

Emission rates and their source

The emission rate is the single input with the most direct effect on the answer, and it is the one most often taken on trust.

A reviewer should establish whether the rate came from source testing, from manufacturer data at a stated reference condition, or from a published emission factor. Each is legitimate and each needs different scrutiny. Manufacturer figures should be normalised to dry gas at the stated temperature and oxygen reference, and a figure quoted without those conditions cannot be compared with a limit or used in a model.

For odour work, the source term should come from sampling to the European standard for dynamic olfactometry rather than from a literature value for a notionally similar process, and where a literature value is unavoidable the report should say so and run a higher case. Where a literature value has to be used, the sensitivity case is what carries the conclusion.

For nitrogen dioxide, check the method used to convert modelled oxides of nitrogen to nitrogen dioxide, because the available approaches give materially different answers at the same emission rate. Then check the background concentrations: the mapped background year, whether the source has been removed from the background to avoid double counting, and the size of any verification adjustment factor. A large adjustment factor indicates a model that is not reproducing local conditions, and the report should explain it rather than apply it silently.

Sensitivity runs a reviewer should ask for

Ask for these as a set rather than one at a time, because a modeller can rerun them in a morning.

1. Each meteorological year reported separately, with the worst case identified.
2. The adjacent surface roughness value for the dispersion site.
3. With and without building downwash, and with an alternative building representation.
4. The emission rate increased by a stated margin, commonly half again, to test headroom.
5. Alternative nitrogen dioxide conversion methods where that is the controlling pollutant.
6. Unadjusted alongside verified results, so the magnitude of the adjustment is visible.
7. Receptor heights at upper floors as well as ground level where residential receptors are multi-storey.

Where the conclusion survives all seven, it is a conclusion worth relying on. Where it survives only the central case, say so in the review rather than accepting it, because an objector will run the same sensitivities.

An Air Quality Assessment or an Odour Assessment that includes a sensitivity annex is easier to defend than one that does not, and for an Environmental Impact Assessment chapter it is close to expected. Note that an Environmental Statement must be prepared by competent experts under the Town and Country Planning (Environmental Impact Assessment) Regulations 2017, and an AI-generated draft is a starting point for those experts rather than a substitute for their work.

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, and the professional directory is the route to a review request. This is general guidance, not legal advice. Check the position with your local planning authority or a qualified professional before relying on it.