What GSD do you need to spot a badger sett entrance from satellite imagery?

The short answer
A badger sett entrance hole runs about 20-30cm across. No commercial satellite sensor gets you there. Even the sharpest panchromatic satellite product on the market sits around 0.3-0.5m GSD, and that's before you factor in off-nadir angle, shadow, and the fact that an active hole is often tucked under bramble or a tree root ball. What you actually resolve at satellite scale is the ground disturbance around the hole, never the hole itself.
What you can resolve at 0.5m is the spoil heap: the mound of excavated soil at the entrance, which on an active sett is usually 1-3m across and bare or sparsely vegetated against the surrounding ground cover. You can sometimes pick up the worn path leading to it, and on a big sett with multiple entrances, the overall disturbance pattern across a bank or hedgerow. That's the actual deliverable from high-res satellite imagery for this kind of survey: a disturbance signature, not a hole count.
At 2m GSD, forget it. A spoil heap at 2m is a handful of pixels with no clear edge, indistinguishable from a rabbit scrape, a fallen branch, or just a patch of bare chalk. If your Phase 1 or protected species scoping relies on 2m imagery, you're not going to flag setts reliably, full stop.
0.5m vs 2m: what each actually shows you for ecology work
This split matters for every small-feature ecology job, not just badgers, so it's worth being clear about where the line sits.
At 2m GSD you're working at the level of land cover and broad habitat type. Woodland edge, scrub encroachment, field boundary change, that kind of thing. Fine for a desk study baseline or a habitat condition check across a big site. Not fine for anything that needs you to point at a specific spot and say "this is a feature of interest."
At 0.5m GSD you start seeing individual trees, hedgerow gaps, bare ground patches, and track networks. That's the resolution band where spoil heaps, badger paths, and other small ground-disturbance signatures become visible enough to flag for ground-truthing. It's still not a substitute for a walkover survey. Natural England's licensing position and most consultancy methodologies still expect eyes on the ground to confirm occupancy status before you write anything in a report. What 0.5m imagery buys you is a way to prioritise where those eyes go, especially across a site too large to walk entirely within a single survey window.
Sub-0.5m imagery exists from some providers but isn't widely tasked commercially at scale yet, and even where it is, you're still looking at a proxy signature rather than the hole itself. So the honest framing is: 0.5m is roughly the floor for this work, and anything coarser than about 1m stops being useful for sett-scale detection.
Why this isn't a classifier you can buy off the shelf
Here's the practical problem. Spoil heap signatures vary by soil type, substrate colour, and how recently the sett's been worked. A badger spoil heap on chalk downland looks nothing like one on heavy clay in a hedgebank. No general-purpose land cover classifier has a "badger disturbance" class, because it's not a category anyone's trained a model on at scale. It's too site-specific and too rare an object for the big vendors to bother with.
Which is usually where a consultancy ends up either paying for a bespoke classification build they'll only use once, or just accepting that satellite imagery can't help and going straight to a full walkover. There's a middle path: label a handful of confirmed spoil heaps from your own site or a comparable one, and run that against your imagery rather than hoping a generic model happens to fit. That's the approach behind Few-Shot Mapping, built for exactly this kind of one-off feature class that doesn't exist in any standard product.
If you've got confirmed sett locations from a previous survey and want to see whether the same disturbance pattern shows up elsewhere on a larger consent area, that's worth a conversation before your next survey season.