Time-removal for detection with an occupancy model..

310 views
Skip to first unread message

Sarah Thompson

unread,
Jun 9, 2016, 2:29:04 PM6/9/16
to unma...@googlegroups.com
Hi unmarked users -- 

Does anyone know if any occupancy models in unmarked allow time-removal methods for assessing detection, or could be altered to accommodate that? Possibly by treating 10 equal time intervals as repeated visits in some fashion (maybe by altering data to presume that once a species is detected, it is detected in all subsequent time intervals).

I also have repeated samples, but I would like to compare results of the 2 methods, with the goal of trying to isolate availability from presence (somewhat).

Conversely, would it be inappropriate to use multinomPois() on presence/absence data, if I don't use parameters in the abundance side of the model & only look at detection? I think that is the only model that specifically accommodates removal sampling.

Any advice or relevant papers on the topic would be appreciated!

Thanks,

Sarah Thompson

Kery Marc

unread,
Jun 9, 2016, 2:33:32 PM6/9/16
to unma...@googlegroups.com
Dear Sarah,

I think you can use any traditional occupancy modeling software to fit occupancy models with a removal design by simply setting all observations after first detection to missing. At least, that's what we wrote in the AHM1 book (see page 621), after observing that this yielded exactly the same answers as the more classical formulation of the model using a multinomial distribution for the detection history frequencies.

Best regards  --- Marc



From: unma...@googlegroups.com [unma...@googlegroups.com] on behalf of Sarah Thompson [thom...@umn.edu]
Sent: 09 June 2016 20:29
To: unma...@googlegroups.com
Subject: [unmarked] Time-removal for detection with an occupancy model..

--
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Sarah Thompson

unread,
Jun 9, 2016, 2:38:10 PM6/9/16
to unma...@googlegroups.com
Thanks Marc, I have the book right in front of me. I guess I should have looked there first :)

Sarah

Kery Marc

unread,
Jun 9, 2016, 2:44:02 PM6/9/16
to unma...@googlegroups.com
well, I admit it's kind of hidden...
Best regards  -- Marc


Sent: 09 June 2016 20:38
To: unma...@googlegroups.com
Subject: Re: [unmarked] Time-removal for detection with an occupancy model..

Carlos Henrique Russi

unread,
Sep 30, 2022, 11:46:29 PM9/30/22
to unmarked
Hi!

I got very interesting results with my removal sampling data using the single species approach that Kery describes here and which is described on pg. 621 of AHM1.
However, the next step I would like to take is to use this approach in the multispecies context of Rota et al. 2016 (unmarked function occuMulti). In this case, observations with NA for some species would not be NA for others, and vice versa. I made some attempts in unmarked, but the function "unmarkedFrameOccuMulti" turns into NA all cells with NA in at least one of the species, which is obviously a problem. Does anyone know how I could proceed in this case?

Ken Kellner

unread,
Oct 1, 2022, 9:11:35 AM10/1/22
to unmarked
This is not possible with occuMulti. It might be possible in BUGS -  but I haven't seen an example of anyone doing this for multispecies data.

Ken

Carlos Henrique Russi

unread,
Nov 2, 2022, 4:34:41 PM11/2/22
to unmarked
Thanks, Ken

In a multispecies scenario without interspecific interactions, this could be done in ubms r-package, right? Simply stacking the data for all species into a single matrix, where each row represents a combination of site and species, and each column a sampling event (visit). Then, in the model formula, species identities would be included as random effects: ~(1|sp). What do you think?
If I'm correct, even N-mixture multispecies models could be fitted this way in ubms.

Ken Kellner

unread,
Nov 4, 2022, 8:58:45 AM11/4/22
to unmarked
Yes, you're correct. I've experimented with this a bit. One caveat is that you can't at the moment do data augmentation so you can't fit a community model where you have hypothetical unobserved species, estimate richness, etc. I'd like to someday add a proper version of this model to ubms.

You might want to look at the camtrapR package - there's a neat tool for fitting community occupancy models with more flexibility using a JAGS/nimble backend.

Marc Kery

unread,
Nov 4, 2022, 9:26:20 AM11/4/22
to unmarked
Dear Ken, Carlos,

that's excellent news. A colleague and me, we have distance sampling data for 8 species in the same area and thought about fitting the model to all of them at once as a community distance sampling model. I see that ubms has the Stan version of distsamp(), so we appear to be good for a model without temporary emigration (this would require gdistsamp).

Best regards  --- Marc


From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of Ken Kellner <con...@kenkellner.com>
Sent: Friday, November 4, 2022 13:58
To: unmarked <unma...@googlegroups.com>
--
*** Three hierarchical modeling email lists ***
(1) unmarked (this list): for questions specific to the R package unmarked
(2) SCR: for design and Bayesian or non-bayesian analysis of spatial capture-recapture
(3) HMecology: for everything else, especially material covered in the books by Royle & Dorazio (2008), Kéry & Schaub (2012), Kéry & Royle (2016, 2021) and Schaub & Kéry (2022)
---
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages