Comment by bnjmn | Hacker News Reader

Comment by bnjmn | Hacker News Reader

Parent Full thread

A potentially less labor-intensive way to estimate animal population is to capture a sample, tag them, release them back into the wild, and then later take a second sample and see how many of those animals are tagged: https://en.wikipedia.org/wiki/Mark_and_recapture

I was hoping this article would say something about improving the margin of error for this kind of statistical estimate, or at least provide a critique of mark-and-recapture statistics for counting squirrels (maybe they are just too... squirrely), but alas it sounds like they threw a bunch of volunteers at the problem:

> The trick is to divide and conquer. They drew a grid of 350 hectares—plots of land measuring 10,000 square meters—over Central Park. Think of them as something like Census tracts. Volunteers then fanned out and conducted two counts, one in the morning and another at night. The Squirrel Sighters, as they were called, spent 20 minutes per count searching for furry subjects, looking up in the trees and down in the bushes, and listening to the clawing and clucking sounds they make. Allen likens it to an Easter egg hunt; some volunteers found many squirrels, others saw none.

I was talking to a bird person who used to participate in various population studies with the national park service. There exist species for which tag-and-release doesn't work at all: jays and corvids in particular are verrrry good at avoiding the places and types of traps where they've been captured before.

300-ish man-hours from (probably) untrained volunteers, seems OK for an exhaustive census like this. I'd expect mark-and-recapture to be more expensive, just because it's tough to get people skilled in catching and tagging squirrels. Maybe you can get grad students to do it.

If they couldn't get enough volunteers, the natural next step would seem to be counting a sample of the 350 plots rather than all of them.

A friend of mine worked on this project.

He said the methodology of dividing up into hectares had to do with squirrel behavior. They tend to mostly live within about a 100x100m area, so it was an easy way to be relatively accurate.

I do not mean this as a joke comment, but actually participating in a squirrel census would be a lot of fun. Sometimes the fun is in the journey, not the destination. And in this case, fun journey and the destination is reached.

I wouldn't trust an amateur citizen hobbyist to tag an animal population in a humane manner. There's no accounting for how the tagging is performed, or when the tags get removed and by whom.

If one person does it, others will justify their own efforts to apply tags to wildlife, and then we get into the legalities of animal cruelty, poaching on conservancy land, and on and on.

It's one thing to crowd source passive eyes spotting for squirrels. It's entirely another to invite people to try and handle them, and even interfere with instruments applied to their body.

What methodology could possibly ensure that something with so many error factors can give a reasonably accurate estimate?

Volunteers can vary in their abilities to detect squirrels

Squirrels can vary in their propensity in showing themselves based on weather and more

Squirrels can move between these areas, leading to double-counting

Baby and old squirrels might not even be visible

Are they territorial? That might skew the sampling. Maybe tag one per hectare, then later capture one per hectare, the fraction that are tagged yields some kind of estimate of the total.

Reply on news.ycombinator.com