Camera-trap long term data

[Fernando Mateos-Gonzalez]

Hi all,

I have about 7 years of data from Eurasian lynx in the Czech Republic.

The cats have been monitored continuously with camera traps, and the location and number of those camera traps have been changing over the years, to maximise observations (without capture recapture in mind).

I'm trying to find the appropriate way (if any) to analyse these data with secr, but the more I read, the more overwhelmed I feel about options and constraints. Proper paralysis by analysis. 

I have lots of random, math-illiterate beginner questions, eg:

Can I do this with openCR? Or do I need to use some bayesian witchcraft?

How do I select periods to create a robust design out of it? Or should I try that continuous-time secr from Borchers et al., 2014?

What do I do with the detectors for which I don't have effort data? And the ones that are still in the field?

I did my first secr course a decade ago, in Canterbury (thanks Murray! it was great!) but nowadays the more I read the less I understand any of this.  

Any basic pointers/starting steps (or internet hugs) will be very appreciated.

Thanks,

Fernando.

[Murray Efford]

Hi Fernando

Yes, these data could be tackled with openCR, but I think you should first try a chain of closed population analyses with 'secr' in which each year is a 'session' - the data format is the same, so you can easily shift later to openCR.

It's statistically scary that you adjusted the location of cameras to maxmize observations, rather than representative sampling a region of interest  - that threatens to invalidate inference for density over the region. Maybe it's not that bad. At the very least you need to know where cameras were and for how long they were operated (I hope you have this despite your comment about effort).

For a coarse but probably adequate analysis you can treat each year (session) as a single occasion, use a 'count' detector type with Poisson model, and specify the number of days per camera as effort ('usage'). This uses all the data while avoiding worries about defining 'occasions' for a 'robust' design. My personal view is that you would gain nothing by going to continuous time models.

As for detectors for which you don't have effort data: that's embarrassing, and you might have to fill in the gaps with a guess. Perhaps not a big worry so long as you declare it (allowing for effort often has little effect on density estimates).

Of course you face big hurdles translating this into code, but once you take the first steps you will be able to ask more specific questions. Go well!

Murray

[Fernando Mateos-González]

Hi Murray,

Thank you, this is really helpful! 

I agree with you, of course, on the statistically scary aspect of the data. Perhaps I should have clarified that this is not my design: Data come from multiple organisations and individuals, with different "approaches" to monitoring. That's also the reason why I don't yet have effort data on some of the detectors; sometimes it's even difficult to convince people that I need the locations! 

As a matter of fact, my main aim with this analysis is to try and show to all the stakeholders involved what we could achieve with a more coordinated approach. Hopefully it will also help me to convey the need for a design allowing for representative sampling in future projects, but that's a hard one for some "overly attached" conservationists. People want photos of their favourite lynx :)

Another funny one is trying to get information about the cameras that failed (deployed cameras running out of battery or not working properly). I'm sure this situation is more common than the literature shows (or at least, what I've been able to find), but I guess such candor would find it hard to overcome peer review. Let's see how I fare when I try to publish my current efforts!

Thank you again for the pointers, and for the open door to my future, hopefully more specific questions.

Fernando

-- 

--

--

Dr. Fernando Mateos-González

======================

Researcher @ ALKA Wildlife o.p.s.

bioblogia.net

Nº Colegiado 2071 (COBE)

LinkedIn || Twitter || Instagram

--
You received this message because you are subscribed to a topic in the Google Groups "secr" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/secrgroup/d8Xo9eYRNJ4/unsubscribe.
To unsubscribe from this group and all its topics, send an email to secrgroup+...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/secrgroup/15a84dd7-22df-4860-a34a-af9f445cd73eo%40googlegroups.com.