Camera trapping; Mark-recapture; unidentifable marks

Forum for discussion of general questions related to study design and/or analysis of existing data - software neutral.

Camera trapping; Mark-recapture; unidentifable marks

Postby Zullo » Sat Jul 28, 2012 10:37 am

Hello,
I'm using camera traps to estimate the abundance of male white-tailed deer. The plan is to use the branched-antler pattern of male deer as distinct natural marks and analyze the data in a mark-recapture framework. Here is the problem: Some photographs are too poor of quality to use in the analysis: You can tell the animal has branched-antlers (e.g. carries a unique mark); however, the photo is not clear enough to evaluate whether the individual is a new individual or a recapture.

In the literature, it appears that where researchers are relying on natural markings and encounter this problem they exclude those photographs that are ambiguous. This approach seems problematic. You are either throwing away recaptures or worse, initial captures of unique individuals.

Are there analytical solutions for dealing with photographic data where you know an animal is marked, but cannot evaluate it's identity. I know the mark-resight methods allow for recording recaptures of unidentifable marks when estimating abundance. However, in the present case, there are no "unmarked" bucks. They all have antlers. I also know some closed-population models allow you to introduce a parameter to account for misidentification. However, I don't believe this solves the current problem.

In short, I'm looking for a mark-recapture analysis that accounts for records of unidentifiable marks.
Thanks,
mark
Zullo
 
Posts: 1
Joined: Fri Jul 27, 2012 11:42 am

Re: Camera trapping; Mark-recapture; unidentifable marks

Postby murray.efford » Mon Jul 30, 2012 4:10 pm

Mark

I understand your discomfort! Without pretending to be an expert on photo ID:

My interpretation is that the bad photos provide (almost) no information about the parameter of interest (population size or density), and therefore there is no point in trying to model them. In conventional capture-recapture an animal is unmarked and not known to be alive until it is first caught and identified/tagged. An animal cannot join the marked population or register as a recapture by providing a bad photo, so it may as well not have been photographed at all and it makes sense to discard these photos. There should be no bias as long as all individuals are equally likely to yield bad photos.

I can see two exceptions. Photos of intermediate quality do provide some information and there may be a net benefit in modelling this. Others can point you to the relevant literature, but I wouldn't bother unless there are a lot of these photos.

The other context in which bad photos might convey useful information is in a spatial analysis: there the total number of detections at a point (camera) tells you a little about local density (e.g., one such detection is incompatible with zero density). Some work would be needed to construct this model (basically an SECR model with an extra parameter for the probability of a bad photo), and any gain in precision is likely to be small relative to the usual SECR analysis.

Perhaps others with more experience of photo ID will add to this.

Murray
murray.efford
 
Posts: 686
Joined: Mon Sep 29, 2008 7:11 pm
Location: Dunedin, New Zealand

Re: Camera trapping; Mark-recapture; unidentifable marks

Postby brouwern » Tue Jan 14, 2014 3:25 pm

A mostly speculative comment:
An ambitious option might be to develop an "integrated population model" (IPM; not to be confused with an integral population model) approach and jointly model counts (unidentified/-iable) deer and identifiable individuals. Normally the counts and the mark-recapture data in an IPM have to be independent datasets, but I think people are working on this issue. See the book by Kery and Schuab for how to set up IPMs in WinBUGS.
brouwern
 
Posts: 9
Joined: Thu Jan 21, 2010 9:51 am
Location: USA

Re: Camera trapping; Mark-recapture; unidentifable marks

Postby bmcclintock » Tue Jan 14, 2014 3:53 pm

Another option would be to use mark-resight models that allow for incomplete identification of marked individuals. In this case, the "unmarked" count data are zeros (and the "unmarked" population size is fixed to zero), and the goal is to estimate the unknown number of "marked" individuals. For JAGS code, see supporting information from: McClintock, B. T., Hill, J. M., Fritz, L., Chumbley, K., Luxa, K., Diefenbach, D. R. (2013), Mark-resight abundance estimation under incomplete identification of marked individuals. Methods in Ecology and Evolution. doi: 10.1111/2041-210X.12140
bmcclintock
 
Posts: 45
Joined: Mon Feb 12, 2007 6:10 pm
Location: NOAA National Marine Mammal Laboratory


Return to analysis & design questions

Who is online

Users browsing this forum: No registered users and 12 guests

cron