Issue with sporadic sampling

questions concerning analysis/theory using program PRESENCE

Issue with sporadic sampling

Postby cgoodma2 » Fri Mar 06, 2020 3:37 pm

Hello all,

I am trying to run an analysis for a summer's worth of data. However, sites were sampled relatively sporadically throughout the period of time. This is to say that some sites were sampled up to 5 times within a week, whereas other sites were sampled twice, and still other sites sampled about once per month.

As a result, I have a dataset like this Image (https://ibb.co/zF9V18L), with a great number of missing values.

I have a few questions:

1. I have the capacity to sample in a more dedicated fashion this spring (i.e. sample all sites 3 times within a 2-week period), so are these data even salvageable?

2. Given how sparsely populated the data are, does it make sense to split these up into closed windows, and then use a multi-season model? Is there a way to intuit results given that both occupancy status as well as detection may very well vary as a result of the specific ponds being sampled, rather than the time-step (i.e. there would be some primary periods where there are a greater proportion of known occupied ponds)?

3. If I go that route, does it make sense to keep sites where I only sampled once in a given primary period?

Any feedback is appreciated, even the obvious 'why didn't you think of these things prior to last summer'.
cgoodma2
 
Posts: 1
Joined: Fri Mar 06, 2020 3:03 pm

Re: Issue with sporadic sampling

Postby jhines » Fri Mar 13, 2020 10:56 am

1. Yes, although you'll be limited in the models that can be run. If that's all of the data, only the most simple model (psi(.)p(.)) will probably work.

2. The key to deciding on multi-season models is whether occupancy changes over the course of the surveys. Optimally, you'd like the surveys within each season to occur in a relatively short period of time and a longer period of time between seasons (where change in occupancy can occur). The key to estimation is replication, so you need replicates for each pond if you want to estimate occupancy/detection at each pond. If occupancy at each pond is constant over all surveys, you don't need a multi-season model, just a single-season model with pond as a categorical covariate.

3. Sites which were only visited once still provide some information, so yes, it makes sense to keep them.

If you plan to add to these data in future seasons, I suggest thinking about a study design which will give you the best chance at estimating the things you're interested in. A tool to help with this is program Genpres (included with Presence). It allow you to get an idea of how many sites/surveys you need to get a specified level of precision of occupancy and/or detection.
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA


Return to analysis help

Who is online

Users browsing this forum: Bing [Bot] and 9 guests

cron