12. Can We Trust the Models?
Updated: Oct 13
Dispersion modelling is the accepted way to predict odour, ammonia and dust before a farm is built, and this report does not suggest it should be thrown out. A model, however, is only as good as what goes into it. The applicant’s models rely on inputs and settings that matter most for the conditions Brayford is likely to experience: a site on a col above deep valleys, still damp evenings and hot summer days. The applicant’s own consultants explain some of these choices, but do not test what difference they make.
How the modelling works
The model (ADMS 6†) takes three inputs [1]:
What comes out of the sheds. Emission rates are calculated from bird numbers and standard emission factors per bird.
The weather. Hour-by-hour wind speed, direction and air stability over several years.
The ground. Terrain and buildings that change how air moves.
It then calculates concentrations at each assessed home for every hour, and reports a summary statistic, such as the 98th percentile for odour or the annual mean for dust. The result can be technically correct and still miss real-world conditions that the inputs do not capture.
The applicant's assumptions
Assumption | What the applicant did | Why it matters to you |
Weather data | Computer-modelled weather from a global forecast system (GFS) for 2022 to 2025, at roughly 7 km resolution, adjusted for local terrain. No on-site weather station [1] | A 7 km grid cannot directly capture how air drains down or pools around a particular site. The terrain adjustment estimates it |
Terrain setting | Terrain was modelled, but the consultant changed a standard model setting so it behaves more like flat terrain. The reports explain that the default can over-predict for raised sources such as roof fans, and cause “gross under prediction” for low-level sources, “particularly important overnight” [1] | By the consultant’s own explanation, the change lowers predictions from the roof fans, the farm’s main source. No results using the default setting are shown for comparison |
Calm conditions | The detailed ammonia run left out calm periods, because they cannot be combined with terrain in the model. An earlier run without terrain was said to show calms made little difference [2] | Calm, still air is when pollutants build up near the ground |
Today’s emissions | The ammonia model charges today’s free-range farm with both the full indoor and the full outdoor emission factors, about 4,320 kg a year. Environment Agency guidance gives about 1,200 kg [2][3] | Today’s levels are overstated, so the claimed improvement at the ancient woodland is overstated too (see Article 3) |
Bird numbers | 176,000, while other documents give 173,926 and 172,728 [4] | The differences are small (1 to 2%), but they show the documents were not checked against each other |
Crops per year | Modelling reports: birds kept to about 38 days, 7.5 crops a year. Rest of the application: 7 crops a year on a 49-day cycle [1][4] | The bird age is the same. The number of crops differs by about 7%, which changes every yearly total |
Steady emissions | The dust model assumes constant output. The consultant warns short-term results “should only be used comparatively” [5] | Real emissions rise sharply late in each crop and at clean-out |
Noise survey | Background noise measured 17 to 20 March 2026, in cool weather: 35 dB by day and 25 dB at night [6] | The loudest fans run in hot weather, which the survey did not capture |
None of these choices is tested to show how much the results would change if it were made differently. The one exception, calm conditions, was tested only without terrain. Two choices clearly work in the applicant’s favour. The terrain setting, by the consultant’s own explanation, lowers the predictions from the roof fans, the farm’s main source. The overstated emissions for today’s farm make the claimed improvement look larger. The consultant gives technical reasons for its settings, and this report does not suggest otherwise. When a choice works in the applicant’s favour, though, it seems reasonable to ask for the results to be tested both ways.
Where modelling has been questioned elsewhere
Examples from other farms that have resulted in issues.
Footbridge Farm, Bridgnorth, Shropshire (2023 appeal). A planning inspector dismissed an appeal for a large chicken farm. The inspector found its odour assessment “fails to consider peak odour concentrations at the end of the growing cycle”, gave “limited explanation” of its inputs and lacked “empirical evidence” for its conclusions [7].
Hollins Lane, Shropshire (2021). A resident living about 460 metres from a 235,000-bird unit told the Shropshire Star that he and his neighbours had made more than 80 odour complaints. The Environment Agency said the farm was complying with its permit. The resident said the modelling relied on weather data from Shawbury, more than 10 miles away [8].
The benchmarks themselves. The benchmarks themselves. The odour benchmark of 3.0 comes from the Environment Agency’s H4 guidance, which was withdrawn and replaced on 3 December 2025 by “Odour management: comply with your environmental permit” [9]. The new guidance sets no numerical benchmark. The applicant’s odour report, written five months later, says H4 was withdrawn “with no known replacement” and still measures the scheme against it [1]. The dust report itself describes the guidance it used to choose which homes to assess as “now withdrawn” [5].
What this means
The question is not whether the model is "right" or "wrong". It is whether the margin between the prediction and a level residents would notice is large enough to absorb the known weaknesses. For odour at Braycross Cottages, the proposed 2.15 sits below a benchmark of 3.0 that has no current official standing, using a model the consultant says may under-predict at night.
The council should ask for sensitivity testing: local or better-resolved weather data, terrain left switched on, calm periods included, and end-of-cycle peaks reported separately.
Sources
[1] Odour report (App 2), AS Modelling & Data, 13 May 2026: sections 2 to 5, including notes on meteorological data and terrain.
[2] Ammonia report (App 6), AS Modelling & Data, 13 May 2026: sections 4 and 5.
[3] Environment Agency, Ammonia emission factors for pig and poultry (gov.uk).
[4] Environmental Statement, Acorus, August 2026: paragraphs 3.1.4, 5.1.2, 5.1.8 and 15.3.2.
[5] Dust report (App 4), AS Modelling & Data, 18 May 2026: section 3.7 and Summary and Conclusions.
[6] Noise Impact Assessment (App 3), Matrix Acoustics, 8 June 2026: section 4.2.
[7] Appeal decision APP/L3245/W/21/3289216, Footbridge Farm, Tasley, Bridgnorth, 5 May 2023, paragraph 9.
[8] Shropshire Star, “Dozens of complaints over chicken farm smell”, April 2021.
[9] Environment Agency, Odour management: comply with your environmental permit (gov.uk, 3 December 2025), which replaced H4.