Photo tagging approach for SMR of group living species using remote cameras

[Joel Ruprecht]

I’m trying to estimate densities using a SMR framework for group living species (elk) using remote cameras. A portion of the population has uniquely marked collars. I have a question regarding photo tagging.

 I am getting lots of photos of multiple individuals on a single photograph, and many photos of the same group of individuals over a short period of time. Imagine a group of elk milling around the camera location for roughly half hour at a time with individuals entering and leaving the frame within the half hour period. I am wondering which photo tagging approach would be better:

1) Tag number of individuals on each individual photo

2) Tag number of individuals on each sequence of photos of the same group of animals during the same time period. This would require treating detections of a group of elk as a single “event”, and counting the numbers of individuals in each “event”

Does anyone have any insight on which approach would be better in a SMR framework? Thanks very much for the assistance.

[Ben Augustine]

This seems like a difficult problem.  There is a severe lack of independence between the observations of the different animals because they move around in herds.  Further, the activity centers may be spatially clustered due to the herding behavior, but maybe not if they switch herds often.  I haven't seen anyone tackle these two problems but would be interested if someone has. 

I suspect the lack of independence is even more of a problem in an SMR model because the underestimated uncertainty in the detection function parameters coming from the lack of independence will propagate less uncertainty into the estimated number of unmarked individuals that produced the unmarked counts you observed.  Thus the variance of the density estimator will be further underestimated.  And the lack of independence should lead to negative bias if it behaves like the lack of independence between multiple marking methods (http://onlinelibrary.wiley.com/doi/10.1002/eap.1587/full).

I think you need to model the observation process in a manner that accounts for the lack of independence, but I don't have any ideas about how to do that.

On Thu, Jun 29, 2017 at 3:02 PM, Joel Ruprecht <ruprec...@gmail.com> wrote:

I’m trying to estimate densities using a SMR framework for group living species (elk) using remote cameras. A portion of the population has uniquely marked collars. I have a question regarding photo tagging.

I am getting lots of photos of multiple individuals on a single photograph, and many photos of the same group of individuals over a short period of time. Imagine a group of elk milling around the camera location for roughly half hour at a time with individuals entering and leaving the frame within the half hour period. I am wondering which photo tagging approach would be better:

1)     1) Tag number of individuals on each individual photo

2) 

2)    2) Tag number of individuals on each sequence of photos of the same group of animals during the same time period. This would require treating detections of a group of elk as a single “event”, and counting the numbers of individuals in each “event”

Does anyone have any insight on which approach would be better in a SMR framework? Thanks very much for the assistance.

--
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+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Ben Stevenson]

Hi,

The lack of independence due to clustered activity centres is equivalent to the problem in acoustic SCR whereby calls from the same animal have the same source location. This issue does not appear to affect point estimates of call density, but results in underestimated variance estimates (e.g., confidence intervals have less than their nominal coverage). This is mentioned in this paper, which also describes how you can obtain appropriate variance estimates via simulation, but this requires that you have external information about herd size (for clustered activity centres due to herds) or call rates (for clustered sound sources in acoustic SCR). This solution is not entirely satisfactory given the requirement for this external information. However, I have semi-developed methods for things that work if you can recognise which herd an animal comes from (for clustered activity centres due to herds), or if you can recognise which calls come from the same animal (for clustered sound sources in acoustic SCR).

Anyway, just thought I'd mention that there are probably solutions to this problem, but none are currently available, as far as I am aware!

Cheers,

Ben

[Ben Augustine]

Interesting comparison.  I was thinking along those lines--an estimate of herd size and herd-level detection events, but then realized the herd membership likely changes through time and it would be difficult to put an identity to each herd, a point you also recognize.  There might be some mega model with herd dynamics here.  GPS data would help greatly in estimating the spatial dynamics.