Grid-scale energy sites more often on provisional BMV farmland
Intersection with Provisional Best & Most Versatile (BMV) farmland
Grid-scale energy sites sit on provisional best-and-most-versatile farmland more often than infill sites
Grid-scale energy sites sit on provisional Best and Most Versatile (BMV) farmland far more often than urban infill does.
| Development type | Sites | On provisional BMV |
|---|---|---|
| Grid-scale energy sites | 55 | 72.7% (40) |
| Large residential sites | 347 | 54.2% (188) |
| Logistics and freight | 18 | 50.0% (9) |
| Urban infill and estate redevelopment | 232 | 46.6% (108) |
| Whole cohort | 652 | 52.9% (345) |
Pearson’s χ² = 12.72, df = 3, p = 0.0053.
Four limits
The soil data is provisional. Natural England’s provisional classification doesn’t split Grade 3 into 3a (BMV) and 3b (not BMV). Applicants routinely commission auger surveys to downgrade 3a to 3b. So the data show infrastructure proposed on legally contestable soil, not proven Grade 1–2 land.
It is less robust than the other findings. It survives whole-thesis false-discovery correction (adjusted p = 0.0095) but not whole-thesis Bonferroni (adjusted p = 0.095).
The pattern belongs to energy. Logistics is 9 sites out of 18 (small numbers) and sits close to infill.
The categories are ours. The development types are our own rule, not the campaigners’ or the planning system’s. Energy and logistics are keyword matches on the campaign name and its topic tags. “Urban infill and estate redevelopment” is mostly (201 of 232 sites) a land-cover rule: outside the Green Belt with low tree canopy, with no keyword at all. “Large residential sites” is whatever is left. We have not yet tested these labels against a second coder, so differences between types may partly reflect land cover.
Revised 12 November 2026: Title and labels changed to name the site type rather than “capital”. Added a note that the site types come from our own rule.
Corrected again 12 November 2026 after a check of the data pipeline: the description of how the development types are assigned is corrected (most “infill” labels come from tree canopy, not keywords) and marked as untested.
Evidence type: Association. Source: Manuscript ch. 8.1–8.2, appendix A.2. No claim in this series is causal.