[bmsan]

Dang, hits home. When I was a senior in high school, I was lucky to able to volunteer under Dr. Eric Brown De Colstoun at NASA Goddard, checking error rates for tree cover estimates using Landsat data^. Many hours that fall spent trudging around parks and forests, looking at the sky through a PVC pipe. It still kind of blows my mind at how much is able to be gained from images where each pixel is 15mx15m of ground-level area (and, I believe, with an important component of Landsat 7's imaging system broken for most of its lifespan).

I also wasn't aware that Landsat program imagery had been made free to access a few years later. Nice.

^(A massive thank you to him, since I wouldn't have graduated without being able to participate in that project. And a massive apology for going on to get a fine arts degree.)

[is_true]

What did you do with the pvc pipe?

[MostlyStable]

if I had to guess, they were going to randomly selected locations, looking at the sky through the PVC pipe (presumably straight up), and seeing if it was obstructed by tree canopy or not, and then comparing to whether or not the satellite said there was tree cover in that location

[bmsan]

Exactly this. Specifically, how much error rates increased as you moved towards the treeline. The "am I looking straight up?" check was a washer hanging from the pipe by a piece of string.

[bargle0]

Yeah, the scan line corrector broke. Landsat 7 images had these “whiskers” of missing data running perpendicular to the path. For a while there it was just old Landsat 5 and broken Landsat 7.