Why isn't the nearest meteorological station always the best choice?

Geographic proximity is only one of several factors that determine whether one location's met data is representative of another. Two stations 20 km apart can have very different wind regimes if one sits in a sheltered valley and the other on an exposed coast. Conversely, a station 60 km away on similar terrain at a similar elevation, with a similar fetch to the prevailing wind, may give a much better estimate of conditions at the modelling site than a nominally closer station behind a ridge.

What factors should I consider when selecting a meteorological station?

A defensible station selection considers at least: distance from the modelling site, similarity of terrain and surface roughness, similarity of elevation, exposure to the prevailing wind, distance to and orientation of the coast, station observation type (synoptic versus automated), and the length and completeness of the available record. For sensitive assessments, considering more than one station and presenting the choice transparently to the local authority is good practice.

Where do UK meteorological data for air quality modelling come from?

The Met Office's MIDAS archive (via the CEDA Archive) is the most common source of historical hourly meteorological data for the UK. Met Office MOEDAS and commercial dispersion-modelling data services repackage MIDAS data into the formats required by ADMS and AERMOD. Some sites also use NWP-derived data such as MET-NWP files where surface stations are sparse — for example, in upland or coastal locations.

Choosing a Representative Meteorological Station for UK Dispersion Modelling

Introduction

Few decisions on an air quality assessment affect the modelled result as much as the choice of meteorological dataset. Wind speed, wind direction, atmospheric stability and boundary-layer depth all drive the dispersion of pollutants from a road, a stack or a kitchen extract. Get those wrong and the predicted ground-level concentrations can be off by enough to flip a planning recommendation from approval to refusal — or, just as bad, to wave through a development that should have triggered mitigation.

And yet the most common approach in practice is also the most defensible-sounding and the most wrong: pick the nearest meteorological station and use whatever data come out. This article sets out why “nearest” is rarely a good enough answer, what a representative station actually looks like, and how to structure the choice so that it stands up to local-authority scrutiny.

The nearest met station is rarely the most representative. Distance is one criterion among many — not the only one.

Why Met Data Matters in Air Quality Assessments

Dispersion models such as ADMS-Roads, ADMS-Urban and AERMOD use hourly meteorological data to drive their dispersion algorithms. The model needs to know, for each hour of the year, the wind speed, wind direction, temperature, cloud cover, precipitation and other variables that determine how a plume behaves once it leaves the source.

From those hourly inputs the model derives the atmospheric stability class (or in the case of ADMS, the boundary-layer parameters directly), which governs how readily a plume mixes vertically and horizontally. A change of stability class can change the predicted ground-level concentration by a factor of two or more for the same source. The choice of input data is therefore not a peripheral decision; it is the foundation on which every result rests.

The “Nearest Station” Mistake

The intuition that the nearest met station must be the most representative is rooted in a reasonable starting assumption: weather is, broadly, a regional phenomenon. Over very flat, very homogeneous terrain that assumption holds tolerably well. Over much of the UK it does not.

Consider three real situations any UK practitioner will recognise:

The coastal-to-inland transition. A coastal airfield will record a wind regime dominated by sea breezes and a constrained range of wind directions tied to the orientation of the coastline. Inland, only 30–40 km away, the wind rose can be very different.
The upland-to-lowland transition. A station at 200 m above sea level on the flanks of the Pennines will record higher mean wind speeds, different stability distributions and a different prevailing direction than a site 25 km away in a sheltered river valley.
The urban heat island. Met stations in major cities record reduced overnight stability and elevated minimum temperatures relative to rural stations only a short distance away. Using rural data to model an urban site (or vice versa) will systematically bias the predicted concentrations.

In each of these cases the nearest station may be the worst choice. A station further away, but on terrain that resembles the modelling site, will produce more reliable results — and is more defensible if challenged.

What Makes One Station Representative of Another Location?

A representative station is one whose meteorological conditions can reasonably be assumed to resemble those at the modelling site, across the variables that matter for dispersion. In practice that judgement rests on several independent criteria, each of which can move the answer in or out of acceptability:

Distance. Closer is generally better, all else being equal — but rarely the deciding factor on its own.
Terrain and elevation. A modelling site at 50 m above sea level is poorly represented by a station at 250 m, and vice versa. Surface roughness (urban, suburban, rural, open water) should also match if possible.
Coastal influence. Sites within around 5 km of the coast experience meteorology dominated by land–sea temperature differences. Coastal-to-coastal or inland-to-inland comparisons are usually preferable to mixing the two.
Exposure to the prevailing wind. A station sheltered behind a ridge in the prevailing wind direction will under-represent typical wind speeds. The fetch upwind of both station and modelling site matters.
Station observation type. Synoptic stations with trained observers historically produce higher-quality cloud-cover and precipitation observations than fully automated stations. Modern automated stations are perfectly serviceable for most dispersion work but the distinction is worth keeping in mind for sensitive assessments.
Record length and completeness. A short or gap-filled record reduces statistical reliability. Five years of complete hourly data is the conventional minimum; some authorities and some applications expect longer.

A Practical Selection Procedure

The most defensible approach is to make the selection visible. Rather than naming one station in passing and moving on, the assessment should record:

The candidate stations considered (typically the three to five most plausible options within a reasonable radius)
The basis on which each was retained or excluded — usually a short table covering the criteria above
The years of data used and the source (Met Office MIDAS archive via the CEDA Archive is the standard reference)
Where appropriate, a sensitivity test using a second station to demonstrate that the result is not unduly dependent on the choice

This level of documentation rarely adds more than half a page to the assessment but transforms how it is received. A planning officer reading a met-data section that names one station with no justification has to take the choice on trust. A section that walks through the alternatives, classifies them on multiple axes and explains the decision is materially harder to challenge.

UK Meteorological Data Sources

For a UK assessment, the practical options are:

Met Office MIDAS archive (CEDA). Hourly surface observations from the historic UK surface network. The reference source for academic and consultancy dispersion modelling. Free for academic use; commercial use is licensed.
Met Office MOEDAS and commercial repackagers. The same underlying data prepared in ADMS-ready or AERMOD-ready formats, with quality control and gap-filling applied.
Numerical weather prediction (NWP) data. For sites with no nearby surface station, gridded NWP data can be used to derive a synthetic met dataset. NWP data avoids the representativeness problem geographically — the cell can be centred on the modelling site — but loses the local roughness signal that a real surface station captures.

Common Errors Reviewers Spot

From reviewing other consultancies' assessments on behalf of clients, the recurring problems are predictable:

One station named with no justification, often the nearest by straight-line distance
Coastal data used for inland sites or vice versa, with no acknowledgement
Three years of data used where five is the convention — sometimes the only three years for which the consultant happens to have a licence
A station with a long-term mean wind speed obviously incompatible with the modelling site (e.g. exposed coastal station used for a sheltered urban courtyard)
No windrose presented at all, leaving the reviewer with no way to sanity-check the data

Each of these is fixable with a small amount of additional work at the scoping stage. None of them is fixable cheaply once the model has been run and the report drafted.

How Air Dust Odour Approaches Met Station Selection

At Air Dust Odour we maintain an internal reference library of windroses and summary statistics for the principal UK airfield meteorological stations. Each station is classified against a small number of independent axes — distance, terrain, elevation, coastal influence, prevailing-wind fetch, observation type — and selection for any given assessment runs through a structured procedure rather than a default to the nearest record. The result is a met-data section that names a specific station, explains why, and tabulates the alternatives that were considered and rejected. Reviewing officers consistently respond well to this format.

For more detail on how we structure air quality assessments end-to-end, see our pages on air quality assessments and dispersion modelling, or get in touch via the form below.

Conclusion

Meteorological station selection is one of the few decisions on an air quality assessment that can change the answer materially without changing anything about the source or the receptors. Reaching for the nearest station is rarely wrong on its face but often wrong in fact. A defensible selection considers distance, terrain, elevation, coastal influence, exposure and observation type; presents the choice transparently; and where the answer is finely balanced, demonstrates robustness with a second-station sensitivity test. The additional effort is modest. The reduction in risk to the planning recommendation is not.

Get a Free Quote →