Time-dependent detection probability in spOccupancy

[CARLOS ALBERTO GÓMEZ]

Hello Jeff and everyone,

I am a PhD student working with spatial multi-species occupancy models in spOccupancy for pollinating bees and wasps sampled across livestock production systems.

I would like to ask for advice regarding a potential violation of the closed population assumption in my sampling design, and how this could be incorporated into the detection component of the model.

My sampling consisted of coloured pan traps deployed during three consecutive sampling occasions at each site. Because individuals are collected and removed from the population during each occasion, I am concerned that the probability of detection may decrease through time due to removal effects, especially for solitary bee species with potentially low local abundances.

An ecologist suggested that a relatively simple way to address this issue could be to allow time-dependent detection probability, under the assumption that removal may reduce detection probability across occasions while the site itself remains occupied.

I have a few related questions:

1. Would modelling detection probability as time-dependent across occasions be an appropriate way to address this type of issue in spOccupancy?
2. If so, what would be the best way to implement this in spOccupancy?
3. Would you recommend any alternative approach for dealing with possible removal effects in occupancy models when individuals are collected during sampling?

Additionally, for some solitary bee species, I was considering modifying the detection histories after the first detection (e.g., converting subsequent occasions to NA), but I am worried this could create inconsistencies within a multi-species occupancy framework.

Any thoughts or suggestions would be greatly appreciated.

Thanks a lot,
Carlos

[Jeffrey Doser]

Hi Carlos, 

Thanks for the note. Here are some thoughts: 

• You could in theory incorporate a very basic form of time-dependent detection probability in spOccupancy by including a simple random effect of time in the detection model. However, in an ideal world you want a bit more complexity in the detection model, beyond which spOccupancy can handle. The basic occupancy model assumes conditional independence between the detection probability of a species during one visit and the detection probability during a second visit. This assumption is violated in your case, which would likely lead to overly precise uncertainty estimates and perhaps some bias in the estimates (depending on the rarity of your species). This could be explored with simulations. There may be some work out there that explores how to fit an occupancy model in that case, so I would encourage you to do some searching of the literature to see if that is the case. In general, I would probably not recommend using this approach for your case, as if there are rare species then this dependence in the detection probability could be quite substantial and lead to inaccurate conclusions.
• The only way I am familiar with when using a removal design in occupancy models is to do what you suggested at the end: to use all visits prior to the first detection, and then after that set values to NA for the remainder of the detection history at the given site for the given species. When that is the case, however, each species will have a different detection history (i.e., each will have a different pattern of NA vs. non-NA visits at each site), which will then cause issues in the spOccupancy multi-species model-fitting functions. Unfortunately, such a design cannot currently be handled by spOccupancy, so if you wanted to do this, you would need to fit the model in NIMBLE or some other general-purpose Bayesian software. 

In short, I'm not sure spOccupancy is particularly well-suited to handle this specific removal design, so you might look into some other alternatives. 

Jeff

--
You received this message because you are subscribed to the Google Groups "spOccupancy and spAbundance users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to spocc-spabund-u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/spocc-spabund-users/1a026eb3-5f8d-46e2-9fdc-fa511449b4c6n%40googlegroups.com.

--

Jeffrey W. Doser, Ph.D.

Assistant Professor

Department of Forestry and Environmental Resources

North Carolina State University

Statistical Ecology and Forest Science Lab

Pronouns: he/him/his