Dynamic multispecies occupancy with interactions

[Marc Kery]

Dear Ken,

I am asking this question here because I hope it may be of general interest to some.

We have occupancy data for two bird species with very directed interactions; one is a predator of the other. It's multi-year and we may or may not distinguish more than one kind of occupied site (such as the typical 'occupied without repro' and 'occupied with repro').

My favourite approach would be to have a multistate, dynamic occupancy model with a Waddle-type of species interaction, where we use the presence/absence of the predator as a predictor in the model for the prey. We can and eventually probably will fit such a model in JAGS, but as a learning journey for the student, would like to fit as much as possible also in unmarked. What are our options ?

• We will likely explore the data with colext() for single-species, single-state dynamic occupancy models.
• Then, we can use function occuMS() to fit single-species, multistate dynamic occupancy models
• But what about species interactions ? Here, we 'only' have the Rota et al models, with function occuMulti(), which does not support data from multiple years, right ?
• And then, what about the Waddle-et al type of species interactions ? I think I heard rumours that a beta version for fitting this model might be just around the corner ?

Thanks for advice and sorry in case I overlooked something important because I didn't do my homework.

Best regards   ---- Marc

[Ken Kellner]

Hi Marc,

At the moment, yes, the only type of interaction model supported is the Rota model via occuMulti.

I have working beta code for the single-season Waddle model. The model fitting works, I just never finished implementing all the necessary methods and tests and then got distracted with other projects. I can upload this to a branch on Github in the meantime if there's interest so you and others can try it.

I will also note that Mike's 'wiqid' package supports the very similar Richmond version of the model (I believe also single-season only).

It is very likely I will eventually finish the Waddle model and get it into the released version of unmarked (probably not in the super-near future). It is less likely, but possible, I'd add support for the dynamic version of the model (it's probably more likely I'd add support for random effects and thus stacked models).

I think it's pretty unlikely I'd add support for a dynamic, multi-state version. I think that's too specific of a model structure to justify all the development/maintenenance that would be required. Of course if there are others out there that want to work on implementing this in unmarked I'd welcome that!

Ken

On Wed, Jan 15, 2025 at 05:20:48PM +0000, Marc Kery wrote:
> Dear Ken,
>
> I am asking this question here because I hope it may be of general interest to some.
>
> We have occupancy data for two bird species with very directed interactions; one is a predator of the other. It's multi-year and we may or may not distinguish more than one kind of occupied site (such as the typical 'occupied without repro' and 'occupied with repro').
>
> My favourite approach would be to have a multistate, dynamic occupancy model with a Waddle-type of species interaction, where we use the presence/absence of the predator as a predictor in the model for the prey. We can and eventually probably will fit such a model in JAGS, but as a learning journey for the student, would like to fit as much as possible also in unmarked. What are our options ?
>

> *

> We will likely explore the data with colext() for single-species, single-state dynamic occupancy models.

> *

> Then, we can use function occuMS() to fit single-species, multistate dynamic occupancy models

> *

> But what about species interactions ? Here, we 'only' have the Rota et al models, with function occuMulti(), which does not support data from multiple years, right ?

> *

> And then, what about the Waddle-et al type of species interactions ? I think I heard rumours that a beta version for fitting this model might be just around the corner ?
>
> Thanks for advice and sorry in case I overlooked something important because I didn't do my homework.
>
> Best regards ---- Marc
>
>
>

> --
> *** 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.
> To view this discussion visit https://groups.google.com/d/msgid/unmarked/ZR0P278MB0869C232F18D552316A8BD90EB192%40ZR0P278MB0869.CHEP278.PROD.OUTLOOK.COM.

[Marc Kery]

Dear Ken,

thanks for the quick and helpful reply. We really want multi-season and probably best with dynamics; so neither your Waddle beta version nor the Richmond version in wiqid would do for us. occuMulti() only does single season, right ?

Thanks and best regards  —Marc

---

From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of Ken Kellner <con...@kenkellner.com>
Sent: Wednesday, January 15, 2025 19:07
To: unma...@googlegroups.com <unma...@googlegroups.com>
Subject: Re: [unmarked] Dynamic multispecies occupancy with interactions

 

To view this discussion visit https://groups.google.com/d/msgid/unmarked/Z4f5YSwwov1GazBi%40BADGER.

[Ken Kellner]

Yes, occuMulti is single-season only. Looks like JAGS is probably the best option.

Ken

> To view this discussion visit https://groups.google.com/d/msgid/unmarked/ZR0P278MB08699412CB702FFBE0F3768EEB192%40ZR0P278MB0869.CHEP278.PROD.OUTLOOK.COM.

[Marc Kery]

Thanks, Ken !

Marc

---

From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of Ken Kellner <con...@kenkellner.com>

Sent: Wednesday, January 15, 2025 19:55

To view this discussion visit https://groups.google.com/d/msgid/unmarked/Z4gEpkWNRb0ZOap3%40BADGER.

[Jennifer Moore]

Hi Marc,

I know this is an unmarked list-serv... however, we used a dynamic (multi-year) 2-species co-occurrence occupancy model in this paper (https://esajournals.onlinelibrary.wiley.com/doi/abs/10.1002/eap.2397) using RPresence. Not sure this is exactly what you are looking for but could be a good starting place - happy to discuss more if this is helpful.

Jennifer

To view this discussion visit https://groups.google.com/d/msgid/unmarked/ZR0P278MB0869F191F74CE3FA681B5DA5EB192%40ZR0P278MB0869.CHEP278.PROD.OUTLOOK.COM.

[Marc Kery]

Dear Jennifer,

thank you very much for this interesting and relevant information. We will gladly check this out. We are not religiously tied to unmarked as opposed to PRESENCE or even MARK.

Best regards  ---- Marc

---

From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of Jennifer Moore <jenn.f...@gmail.com>
Sent: Wednesday, January 15, 2025 20:17

To view this discussion visit https://groups.google.com/d/msgid/unmarked/CALePw%3DycVoY0BfMnD2HTZw_AKpxO-i%3DsiuSK_CCF72dRfnF1fQ%40mail.gmail.com.

[Brittany Mosher]

Hi Marc et al.,

Similarly, the dynamic 2-species conditional binomial formulation (the Richmond formulation) is available in RMark. We used it here for a host-pathogen dynamics question: https://esajournals.onlinelibrary.wiley.com/doi/full/10.1002/eap.1699.

Best,
Brittany

To view this discussion visit https://groups.google.com/d/msgid/unmarked/ZR0P278MB08690D2236493AACC583637EEB192%40ZR0P278MB0869.CHEP278.PROD.OUTLOOK.COM.

[Marc Kery]

Also very welcome, thank you very much !

Marc

---

From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of Brittany Mosher <mosher....@gmail.com>
Sent: Wednesday, January 15, 2025 20:48

To view this discussion visit https://groups.google.com/d/msgid/unmarked/CAHe9k%3DsOgsNmO5oKvF4MvpfMaKCH1wrVtmPXPgqFcPjTu919sA%40mail.gmail.com.