[_yb2s]

So I asked- and they said (1) when the facility is large enough, which many are, you can randomly sample locations and if they died from injuries elsewhere these will still end up in the sample; (2) they do a lot of separate experiments to measure things like the rate of carcass removal by scavengers which go into the calculations; (3) depending on the type of bird and turbine, in most cases injuries are immediately fatal, and they can quantify this for each case with direct observation.

I’ll myself add that presumably the data collected on larger farms can be used as a prior to make accurate inferences on smaller farms.