Population & density estimates - 'detect', 'unmarked', Distance?

[dyeany]

I am looking for some guidance on producing density and population estimates from a point count dataset for about a dozen grassland bird species while linking relationships to a suite of habitat data. I had intended to use package 'detect', adapting previously used code, but in light of a recent paper by Solymos et al. 2018, it does not seem appropriate for these data, as all species have fewer than the 1000 detection sample size suggested by Solymos.

So I have been struggling to determine whether using a function in the package 'unmarked' is appropriate or if I should be looking to program Distance. The purpose of my project is practical grassland bird habitat management and conservation. With unmarked, I am unsure which functions should be utilized and how to get the process going. I have reviewed various literature on the package, but would greatly appreciate any help here.

Here is the structure for my data:

37 sites, 521 points, 2 survey visits in one year, 3 distance bands (0-50m, 50-100m, >100m), 5 - 1min time intervals-singing & non-singing detections recorded separately

survey covariates: site_name, survey_yr, time since local sunrise, sky, wind, temp, surveyor

habitat covariates: 29 variables: 12 different grassland community types (proportion), 10 disturbance types (proportion), cover: herb, shrub, bare (integer), height: herb, bare, shrub (integer)

Thank you in advance for any help you can provide!

David

[Chris Merkord]

I would suggest deciding what kind of model you want to use before deciding which software to implement it with. You have distance data, so you could use distance sampling. But you also have repeated surveys within a season, so you might consider an N-mixture model. With your time intervals you could use a removal model. And there are ways to combine those data, for example by using a time-removal plus distance sampling model. Part of your choice should be based on your main question. If you need a density estimate for an area, you probably want a distance sampling model of some sort. If you are content to know N, the number of birds observed (or expected to be observed) on a given point, then an N-mixture model of some sort may suffice.

To learn more about the various types of models (many of which are implemented in unmarked), I suggest you get a copy of this book:

• Kéry, Marc, J. Andrew Royle. 2015. Applied hierarchical modeling in ecology: analysis of distribution, abundance and species richness in R and BUGS: volume 1: prelude and static models. Academic Press. London, UK.

I do not consider myself an expert on any of these types of models, and given that nobody else has answered in the last couple of days, I suggest you re-post your question on the unmarked Google group: https://groups.google.com/forum/#!forum/unmarked. Or if you're interested in going Bayesian, the hmecology: Hierarchical Modeling in Ecology group https://groups.google.com/forum/#!forum/hmecology

Let us know what you find out.