adding lion and wild dog covariates

[klg-2016]

I'm not sure if this is exactly the next step, but I believe it's coming up: what are your thoughts on how to add the lion and wild dog data? We'd discussed using a 0/1 system for absence/presence, which I think would be a reasonable way to approach it to start. I don't know that the data we have is of a high enough resolution to do much more. I'm going to take a look back at the lion data to remind myself what format it's in, then I'll add my thinking here on how best to use it. I'm not sure where you said you thought we could get the wild dog data from?

[klg-2016]

OK looking back at the lion data, we have presence/absence based on camera trap data (so in "cammetadata...withlion", I think any nonzero value in the column "lion_camera_dry1617" would correspond to a 1) or we have the scaled raster data (column "lion_latedry"). That's the value from your work with the raster data getting more detail for lion spatial presence, which we could also try using as is. Do you think we should aim for the same metric (both 0/1 or both some value based on raster data) for both species (lion and wild dog)?

[kaitlyngaynor]

I would say 0-1 (either binary or continuous), not sure it matters. Ask Meredith about the wild dog data- she is working with it for another project with the cameras (regarding ungulates) and knows it the best. It’s very coarse.

On Tue, Feb 2, 2021 at 13:15 klg-2016 notifications@github.com wrote:

OK looking back at the lion data, we have presence/absence based on camera trap data (so in "cammetadata...withlion", I think any nonzero value in the column "lion_camera_dry1617" would correspond to a 1) or we have the scaled raster data (column "lion_latedry"). That's the value from your work with the raster data getting more detail for lion spatial presence, which we could also try using as is. Do you think we should aim for the same metric (both 0/1 or both some value based on raster data) for both species (lion and wild dog)?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kaitlyngaynor/gorongosa-mesocarnivores/issues/108#issuecomment-771992534, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHA7WT42HK5FUHWBBXUHHADS5BTOZANCNFSM4W7TCRRQ .

-- Sent from my phone

[klg-2016]

OK--I'll send her an email and go from there

[kaitlyngaynor]

Let me know once you have a look at what Meredith sent. Happy to advise on the best way to deal with it. I know that Paola's manuscript describing the AWD movement data is in revision at PLoS One now, and once it's out, the summarized movement data will be shared with researchers, so we should have a much better sense of where they are going.

[klg-2016]

Will do--if it seems like taking the rough data and converting it to something usable will be too time-consuming, I might also consider waiting for that summary to be available

[klg-2016]

Thinking about this and I have two questions:

The lion data I used in my thesis was from only one year's worth of data per lion (not necessarily all the same years). If lion is going to be used as a yearlysitecov (which isn't necessarily what we're doing, but I think what we have to do if lion presence is going to be tested as something affecting occupancy across years), then I need to derive/assign a value for each site for each year. Do you have advice on how to do that? One way that occurs to me to try is doing the same isopleth scoring technique you used for the data for my thesis and just repeating it for each year. (If that's the route we try, I'd appreciate advice on how to repeat what you did). We'd also discussed using the camera trap data as a rough indicator of lion presence, which may be an option to consider here as well.

For the wild dog data, Meredith has "high", "low" and "no" dog use by site. These have no date associated with them, so I'm assuming they're just an overview covering the period in which dogs have been there? So with the data right now, I think I could have all sites be "no" dog use for 2016 & 2017, and then apply a value to each of the sites based on her categorization that would be the same for 2018 &2019 (the first introduction was April 2018, right?).

She also said this, which I don't follow: "I'm also attaching hexagonal shapefiles that overlay the camera trap hexagons for plotting purposes (*note: I haven't condensed these into a single layer yet, so different parts of the grid are in different layers, apologies if that's confusing!)"

What do you think?

[klg-2016]

yearlysitecovs.xlsx

This is what I'm thinking right now for yearlysitecovs -- following the example from the google group I found and included in the other issue, I think I understand how to take this format and get the relevant variables into the umf. I took Meredith's no/low/high values for each site and made them 0/0.5/1, which may or may not be the route we want to go. I also have empty columns for the lion data until I have a better sense of what we're thinking there too.

[kaitlyngaynor]

Thinking about this and I have two questions:

The lion data I used in my thesis was from only one year's worth of data per lion (not necessarily all the same years). If lion is going to be used as a yearlysitecov (which isn't necessarily what we're doing, but I think what we have to do if lion presence is going to be tested as something affecting occupancy across years), then I need to derive/assign a value for each site for each year. Do you have advice on how to do that? One way that occurs to me to try is doing the same isopleth scoring technique you used for the data for my thesis and just repeating it for each year. (If that's the route we try, I'd appreciate advice on how to repeat what you did). We'd also discussed using the camera trap data as a rough indicator of lion presence, which may be an option to consider here as well.

Hmm, this is tough since we don't have lion data for later years. Also, interannual variation in the isopleths may just be driven by the fact that different animals were collared in different years, rather than some true variation in lion activity in the site. We might think about including this as a static variable that doesn't change by year? Either isopleth-based or camera-based. I think it's okay to be coarse. We might flag this as a question for Paola, though—we could ask her advice on whether lion space use has changed much in the camera grid area from 2016-2019. Maybe we can wait on this, though, since we may have more questions for her as we dig in.

For the wild dog data, Meredith has "high", "low" and "no" dog use by site. These have no date associated with them, so I'm assuming they're just an overview covering the period in which dogs have been there?

I think they are just from the first few months post-reintroduction (you can clarify with Meredith) and again, are really rough, based on Paola's knowledge looking at a map.

So with the data right now, I think I could have all sites be "no" dog use for 2016 & 2017, and then apply a value to each of the sites based on her categorization that would be the same for 2018 &2019

Yep, this sounds good!

(the first introduction was April 2018, right?).

I think the introduction was later than that. They were in a boma for a while before they were released into the park, which I think happened while I was there, which was mid-June to mid-July 2018. Again, I'd check with Meredith!

She also said this, which I don't follow: "I'm also attaching hexagonal shapefiles that overlay the camera trap hexagons for plotting purposes (*note: I haven't condensed these into a single layer yet, so different parts of the grid are in different layers, apologies if that's confusing!)"

Hard for me to know what this means without seeing the shapefiles? Have a look at them and see if you need them, I guess

[klg-2016]

Sounds good -- I just started a doc in the shared meso folder with questions for Paola, which we can pull from when we decide to reach out to her.

I'll follow up with Meredith on the dog questions.