Check if edge window is mostly black.

[bw4sz]

Data quality drops off at the edges of the flightline. I think its 100% reasonable to delete tiles if there is significant black area. @MarconiS there isn't any reason we need to predict all the way to the tile AOP boundary? I was just looking at our low scoring winows, and i'm seeing things like

[MarconiS]

absolutely I agree with you. Tiles like these shall probably be dropped, but there may be some others that are only marginally bad (let's say 70% good and 30% no meaningful data). My suggestion is to pass a filter and/or eventually QC the amount of unusable data in each tile (dropping tiles that have more than NN% of corrupted pixels)

[bw4sz]

I also think I want to try to contact tristan to get a sense for the relationship between the flight data and the actual site footprint. We have plenty of data, might as well restrict it to the site?

On Sat, Feb 1, 2020 at 10:54 AM Sergio Marconi notifications@github.com wrote:

absolutely I agree with you. Tiles like these shall probably be dropped, but there may be some others that are only marginally bad (let's say 70% good and 30% no meaningful data). My suggestion is to pass a filter and/or eventually QC the amount of unusable data in each tile (dropping tiles that have more than NN% of corrupted pixels)

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/weecology/NEON_crown_maps/issues/2?email_source=notifications&email_token=AAJHBLCTP4TNC5CNIHDB2PTRAXANTA5CNFSM4KOPN6UKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKRECFA#issuecomment-581058836, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJHBLAWB2YLHA5HI5P3L4LRAXANTANCNFSM4KOPN6UA .

-- Ben Weinstein, Ph.D. Postdoctoral Fellow University of Florida http://benweinstein.weebly.com/

[MarconiS]

That, I kind of rather not restrict: NEON sites are generally well conserved areas, while the AOP footprint can, sometimes, catch both the NEON site and some less conserved/peri-urban areas. I think it is worth to keep the extra 50% of the space outside NEON sites borders to get a better view of landscape patterns in the ROI. That said, there are the boundaries of each NEON site in one of the NEON metadata shape files. Wanna me to dig it up?

[bw4sz]

I think this is a reasonable point of debate, @Ethan White ethan@weecology.org can we bring this up at a meeting.

On Sat, Feb 1, 2020 at 7:25 PM Sergio Marconi notifications@github.com wrote:

That, I kind of rather not restrict: NEON sites are generally well conserved areas, while the AOP footprint can, sometimes, catch both the NEON site and some less conserved/peri-urban areas. I think it is worth to keep the extra 50% of the space outside NEON sites borders to get a better view of landscape patterns in the ROI. That said, there are the boundaries of each NEON site in one of the NEON metadata shape files. Wanna me to dig it up?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/weecology/NEON_crown_maps/issues/2?email_source=notifications&email_token=AAJHBLDHRXMLKQ6JNGEVX6LRAY4KBA5CNFSM4KOPN6UKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKRMXLI#issuecomment-581094317, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJHBLEQ5FLFJHPZ6GQSGEDRAY4KBANCNFSM4KOPN6UA .

-- Ben Weinstein, Ph.D. Postdoctoral Fellow University of Florida http://benweinstein.weebly.com/

[ethanwhite]

This is an interesting question. In an in-person meeting with @bw4sz I said the same thing as @MarconiS, but after a few hours for the idea of percolate I'm having second thoughts. Interestingly the second thoughts are based on an ethics consideration (and maybe one worth asking NEON about). Clearly the NEON site itself has signed off on this imagery being collected and used for this sort of purpose, but I'm wondering if that's definitely true of the property outside of the borders. It seems possible that some of this could be privately owned or tribal land (there's at least one site that's affiliated with tribal land) and so I think it's an open question whether or not we should be inferring trees in these areas without checking on permissions related to the imagery. If that's a reasonable concern then the easy thing to do for know is restrict to the research site boundaries.