Spatial autocorrelation in stacked, single-season occupancy model (unmarked or ubms)

[Christopher Dennison]

Good afternoon everyone,

I hope this message finds you all well.

As part of my graduate thesis, I have created a stacked, single-season occupancy model in unmarked to determine the potential effect of several environmental variables on bird occupancy.

The model results are positive, but I have recently been attempting to account for potential spatial auto correlation in my data (our acoustic recording sites were part of a collaboration, and most are quite close together).

I had some success with ubms to create a model for single years, but found out that I could not include both spatial random effects and regular random effects in the same model. Of course, I could simply run the spatial occupancy models with ubms on each year (3 years total) to create separate models, but I am wondering if there is an approach to accounting for spatial auto correlation in my stacked single-season model in ubms or unmarked?

This humble graduate student thanks you for your time and support.

Warm regards,

- Chris

[Ken Kellner]

Hi Chris,

I'd suggest looking at the spOccupancy package. I believe it's possible to handle both spatial autocorrelation and an additional random site intercept with the spPGOcc() function. It's a fantastic package with an interface in many ways similar to unmarked.

At some point I'd like to allow for additional random effects in ubms spatial models, but it's not likely to happen in the near future.

Ken