different sites within multi-season occupancy model

[Rebecca Bracken]

Hi all,

Hoping someone has a suggestion for this.  I have avian point count data from a 21 year period within the same forest stands.  But the first 10 years the points were placed in 4 locations, while in the following 11 years the points were placed in 3 locations, with no overlapping points.

Will I be able to fun multi-season occupancy models given the survey points changed across time, even though all data was collected  in the same forest stands?  If so, just fill in the missing point info with NAs?

Or, and I'm not sure if this would be accepted or would be biased and/or a misuse of these models, would it be considered acceptable to just structure the models with the sites being the stands instead of the points?  Certainly some birds would be more likely to be detected within the entirety of the stand versus individual points, but it would remove the site differences.

Thoughts?  

Thanks,

Rebecca

[Marc Kery]

Dear Rebekka,

 

now that’s an interesting design 😊 Seriously: you have 3 or 4 locations per stand, but you have more than just one stand, right ? Because if 3-4 sites is all you have, then I think I’d not fit these models.

 

When you say multi-season, do you mean the dynamic occupancy model with extinction/colonization rates ? Or do you simply want to fit separate parameters of occupancy and detection for each year ? If you have the latter, then I don’t see any problem with the change of sites in your stands, though perhaps you may still want to fit a before and after effect to accommodate any potential systematic differences between the two sets of points (especially given that there are so few of them). But if you want the former, then that one transition without any shared sites will pose a problem potentially. In the latter case, however, if you fit a model with separate extinction/colonization parameters for each annual interval, then you’re fine: you can then simply ignore the estimate from year 10 to 11, since it is probably not meaningful. However, I’d not be so comfortable to fit a model that constrains the estimates across years in some way, e.g., with a constant or with covariates across years. In that case it may be better to use a general modeling engine such as JAGS, because then you could simply treat the transitions between year 10 and 11 as separate and then model across all the remaining years ?

 

You can collapse your data to treat stand as a “site”. But then you lose information.

 

Basically, you have a multi-level design, with points nested within stand. This can be accommodated in an analysis in at least two ways. One is to focus on the point level and treat stand as a random effect. This can be done in unmarked using the TMB engine to fit such random effects. Another way would be to specify dynamics at both the stand and the point level; see this paper: Tingley, M.W., A.N. Stillman, R.L. Wilkerson, C.A. Howell, S.C. Sawyer, & R.B. Siegel. 2018. Cross-scale occupancy dynamics of a postfire specialist in response to variation across a fire regime. Journal of Animal Ecology 87:1484–1496. [link]

For this type of model you’d need JAGS or a similar generic modeling engine.

 

Lots of ways of looking at your design …. What’s best will also depend on your questions.

 

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 on the web visit https://groups.google.com/d/msgid/unmarked/3ab8dc73-b202-4d98-9322-c740c501e45dn%40googlegroups.com.

[Rebecca Bracken]

Hi Marc,

Thanks for the response!  We definitely have data from more than one forest stand, that is not an issue.  I'll take a look at your suggestions, and will discuss with my committee/coauthors to see if we can get a consensus on what to do.  Using JAGS is certainly an option.

Rebecca