Home range area estimation with camera traps via secr

59 views
Skip to first unread message

Benjamin Debetencourt

unread,
Sep 21, 2022, 4:59:58 AM9/21/22
to secr

Dear all secr users,

I am using camera traps data to perform secr models. I have 4 different chimpanzee communities, and I wanted to see if I could estimate the home range of each group thanks to the results of the secr model (and the estimation of sigma).

From what I gather through reading, home range area is possible to obtain using the value of sigma, if gaussian encounter probability model is used. The best AIC model was the MBk one, I used the half normal key function, and computed the 95% territory usage via the calculus pi*(sigma*sqrt(5.99))^2. (Chapter 5 of the secr book from 2014).

In this chapter, on the wolverine example, the authors though advise against biological interpretations of home range area based on the sigma estimate from SECR. From what I understand it could be misleading if the traps affect the individuals use patterns, which I assume is not the case in my study. I will still treat those home range area estimates with cautions but thought this would be more relevant than MCP that is fully constrained by my camera array. I wonder if anyone has any thoughts on the matter.

Furthermore, I would like to be able to map those home range areas. I used the function fxi.contour on all my individuals to get a “space use” for each of my individuals in one group. I checked what happened if I set the fxi.contour percentage at 100%. I am obtaining a very big area by overlaying all the polygon from each individual extracted from the fxi.contour results.

Bakoun_100perc_all_adults.jpg

The blue dots are the cameras, and they cover an area of 10 km x 10 km.

I then decided to try to understand what was happening and had a look at each individual space use estimated via fxi.contour.

On this second figure (only a subsample of all individuals), blue points are still camera, green points are the camera where the specific individual has been captured and on the title is of each graph the name of the individual and the number of times we captured them. The big red dot is the home range center.

 Space_use_individuals_bakoun.png

 I directly noticed that the more captured an individual has been, the more tighten around the home range center the results of the fxi.contour was. For the individuals with more than 2 captures (top middle and right, Cybele and Phine) the results are circle very closed to the home range center with even some cameras where they have been captured not included in that area.

I assume I should not take the 100% data, but I was wondering as the number of captures is affecting the size of the “space use” if there was a recommendation on having only a certain number of captures to be able to perform fxi.contour function on individuals (obviously 1 capture individuals should not be used but 2 times capture individuals also give unrealistic results). And I also wondered if the results on fxi.contour with some restrictions on capture frequency could be used the way I showed above to determine an home range area for my different groups. If not, would there be another way to access the territory of my different groups via the results of the secr model?

Any advice on the matter would be greatly appreciated!

Best regards,

Benjamin Debetencourt

Murray Efford

unread,
Sep 21, 2022, 6:06:28 AM9/21/22
to Benjamin Debetencourt, secr
Hello Benjamin
One basic point: fxi.contour and related functions do not describe space use at all. The contours are for the location of an animal's activity centre, given the data and the model. They always become tighter with more sampling (maybe not quite 'always' but that is how it should work).
The relationship between (halfnormal) sigma and 95% home range size is obviously coarse and subject to all sorts of hedging - useful as a ballpark figure but not much else. It doesn't let you plot each home range on a map. There has been some unpublished work combining sigma (as an indication of space use) and fxi-type statistics (a talk at the vNCSE conference by Rishika Chopara, from memory), but I'm not sure how far that has gone.
Murray

--
You received this message because you are subscribed to the Google Groups "secr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to secrgroup+...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/secrgroup/26519da7-6fcf-4fc8-848e-162a34278883n%40googlegroups.com.

Benjamin Debetencourt

unread,
Sep 21, 2022, 7:56:46 AM9/21/22
to Murray Efford, secr
Hi Murray,

Thanks a lot for the quick answer and the clarification ! I will check to see if I can find some of the work you mentioned. 

Best,

Benjamin
Reply all
Reply to author
Forward
0 new messages