Removal Sampling and Multi-Season Models

questions concerning analysis/theory using program PRESENCE

Removal Sampling and Multi-Season Models

Postby JWatson » Fri Nov 04, 2011 11:17 am

Has anyone used removal sampling in multi-season modeling? Are there issues I should be aware of? I have run some test files and don't see anything unusual in the output but want to cover all the bases before the design is implemented.
JWatson
 
Posts: 4
Joined: Thu Nov 03, 2011 1:37 pm

Re: Removal Sampling and Multi-Season Models

Postby bacollier » Sat Nov 26, 2011 2:09 pm

JWatson wrote:Has anyone used removal sampling in multi-season modeling? Are there issues I should be aware of? I have run some test files and don't see anything unusual in the output but want to cover all the bases before the design is implemented.


JWatson,
I will take a quick stab for you, sorry for slow response on this one, we are usually quicker.

I don't think you have anything specific to be aware of relative to your estimates of colonization and extinction. Colonization and extinction parameters are based off of the presence/absence data collected on your between season surveys on your sample sites, so if you are using an appropriate design wherein you repeating your surveys enough time to ensure detection given the critter was there (e.g., MacKenzie and Royle 2005) then you are probably alright on the estimates for those (and for using covariates on them). However, you need to remember that under a removal design, you are likely not going to be able to model variability in detection using a host of covariates as you go til you hear one then your done, so keep that in mind. I did a pretty simple one of these a few years ago (J. Wildlife Management 74: 140-147), but not knowing your species/study design that is about as specific as I can get, hope it helps.

Bret
bacollier
 
Posts: 230
Joined: Fri Nov 26, 2004 10:33 am
Location: Louisiana State University

Re: Removal Sampling and Multi-Season Models

Postby JWatson » Tue Jan 23, 2018 3:57 pm

I'm hoping Brett or someone else that has conducted multi-season sampling with removal may be able to provide advise on my analysis. We've sampled the same 200 sites for ground squirrels using a 2-survey design with removal in two different years (separated by 5 years). Histories for each site for each of the two years are 1., 01, or 00 ("." indicates removal of the site from the second survey because the species was detected during the first survey). The hope is to estimate gamma and eps using a model with constant P and few covariates. I don't see any multi-season models in PRESENCE with standard parameterization that deal with removal and would appreciate any recommendations as to how to analyze this data. The paper Brett referenced used a removal design but two observers during each survey and "observers" were removed after they made a detection, not the "site" and so histories were "00, 01, or 11". Thus, they could be analyzed with a standard design since there were no "1." records.
JWatson
 
Posts: 4
Joined: Thu Nov 03, 2011 1:37 pm

Re: Removal Sampling and Multi-Season Models

Postby darryl » Tue Jan 23, 2018 9:58 pm

Hi There,
You don't need to do anything special, but with the removal design and 2 surveys per season you have to assume that detection is the same in the 2 surveys (but could be different in each season). You can also include covariates on detection if you want, but again, the effect of the covariate has to be the same for the 2 surveys.
Cheers
Darryl
darryl
 
Posts: 495
Joined: Thu Jun 12, 2003 3:04 pm
Location: Dunedin, New Zealand

Re: Removal Sampling and Multi-Season Models

Postby JWatson » Wed Jan 24, 2018 2:59 pm

Thanks for the quick response. To be clear then, if the two surveys within a season resulted in a history of "1.", then the assumed history is "11", correct? Thanks
JWatson
 
Posts: 4
Joined: Thu Nov 03, 2011 1:37 pm

Re: Removal Sampling and Multi-Season Models

Postby jhines » Wed Jan 24, 2018 3:52 pm

No, the history "11" implies that the site was surveyed in surveys 1 and 2 and the species was detected both times. So, this history provides information for estimating p1 and p2. The history "1." implies that the site was surveyed in survey 1, but not surveyed in survey 2. This history contains no information about p2.

With the removal model, the probability of the history, "1." is just p1. The probability of the history, "01" is (1-p1)*p2. With only 2 pieces of information (number of sites with "1." and number of sites with "01"), and 2 parameters, there are infinitely many solutions. That is why you need to constrain p1=p2 as Darryl said. Then, you're only estimating 1 parameter with 2 pieces of data.

So, leave the history as "1." in the input and PRESENCE will handle it appropriately.
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Removal Sampling and Multi-Season Models

Postby JWatson » Wed Jan 24, 2018 5:50 pm

OK, got it. Thanks again for the reply.
JWatson
 
Posts: 4
Joined: Thu Nov 03, 2011 1:37 pm


Return to analysis help

Who is online

Users browsing this forum: No registered users and 8 guests