A third of contested sites sit where baseline wildlife records are thin, and it isn’t the poorest places.

What we measured

A low-record area is a site with fewer than 5,000 historical NBN Atlas records within 1 km (National Biodiversity Network Atlas). 223 of the 639 sites with a resolved count (34.9%) qualify. 13 sites whose lookup failed are excluded rather than counted as low-record areas.

What it shows

There is no class gradient: 34.9% of deprived-band sites are low-record areas, 38.5% of the middle band and 29.9% of the affluent band (Fisher exact p = 0.42, deprived against affluent). The rate instead peaks at 47.4% (46 of 97 sites) in decile 6, in the middle of the range that takes in commuter belts and logistics corridors.

The pattern is the same at a threshold of 2,500 records (17% of sites overall) and 10,000 (56%): no gradient with deprivation, and the middle band is highest each time. The thesis reads that peak as the peri-urban fringe of extended urbanisation, neither urban enough to attract civic data mapping nor designated enough to attract formal surveys. That reading is interpretive. The decile 6 peak is a descriptive count, not a tested effect.

Why it matters

A thin baseline lets an applicant assert a low ecological value almost anywhere. Capitalised expertise can commission a private counter-survey to settle the question; a community inside a 21-day window generally can’t.

Revised 12 November 2026: The p-value now refers to the rate of low-record areas (Fisher p = 0.42), not a different score. Added a check at two other thresholds. “Data desert” is now “low-record area”.