unequal sampling interval length in PRESENCE

[qrobinson]

Hello everyone,

I'm wondering about the capacity of PRESENCE to handle sampling intervals of unequal length. I'm sure this is a simple question, but so far I've been unable to find a definitive answer.

I'm working with remote camera data of elusive desert mammals. For the sake of illustrating my situation simply, say that I sample 50 different sites in a particular month. Within this month, I might have several locations that were sampled by a camera for the entire duration, while other locations were only sampled for a portion of the month. Obviously the duration that a camera samples a location could influence the likelihood of detecting an animal, and I'd like to incorporate these unequal sampling intervals into my analysis if possible.

I know that MARK is able to simply deal with such scenarios; can PRESENCE do so as well?

Any guidance would be much appreciated, and again I apologize if this is foolishly simple and I've just overlooked something.

Best,

Quinn Robinson

[jhines]

Hi Quinn,

You're saying 'sampling interval' as if a month is a single sample and some sites are sampled for the entire month and some are only sampled for part of the month. For the occupancy models, you should treat each day as a sample, and some sites will have 30 samples in a month and others will have fewer. The input data will contain 1 row for each site and 1 column for each day. If the species of interest is detected at a site for a particular day, that row/column will contain a '1'. If it is sampled, but not detected, it gets a '0'. If the site was not sampled that day, the row/column gets a '-' to indicate 'not sampled'. So, some rows will have more '-' characters than others. PRESENCE will apply the detection parameter to only the columns which contain a '1' or '0', so it accounts for differing numbers of samples per site.

It would be possible to treat a month as a single sample and account for the number of days sampled using a covariate, but then you would lose the replication at each site needed to estimate the parameters.

Jim

[darryl]

Further to Jim's response, you don't have to break your camera trap data into days for use with PRESENCE, you can use any period you like (eg 30 1-day periods, 10 3-day periods or 3 10-day periods). Provided each period is independent, how you define that period to be shouldn't make much difference to your results.
Cheers
Darryl

[mrodtka]

Hierarchical Multi-Scale Occupancy Estimation for Monitoring Wildlife Populations Pavlacky et al. 2012
Tigers on trails Hines et al. 2010
Multi-scale occupancy estimation and modelling using multiple detection methods Nichols et al. 2008

All contian discussions on temporal sampling/independence considerations for these types of datasets that I've found useful.

[qrobinson]

Hi everyone,

Thank you for your clear and helpful responses. Much appreciated!

Best,

Quinn