Community-level density estimate

[Stefanny Sirleg Villagomez Palma]

Hi, 

I have been working on calculating the density of a terrestrail bird community and want to see how it changes during different times of the year (spring and fall). First, I separated the birds into groups and assessed whether detection probabilities were similar across the groups I assigned (e.g., sparrows, warblers), and then I used those groups as covariates. However, for some species, detections were very few, so I decided to keep only those species with a sufficient number of detections that together represented more than 50% of the total detections. The models improve significantly when species are included as covariates, since detection probabilities vary among species, but I would like to know whether there is a better way to estimate community-level densities.

Another question is that within my study area, there are points with zero detections. I think it's important to consider these points when estimating true density. I created the sample table, including all points, even those with no detections, and an observation table with only the detected individuals and their distances. Is there a different way to account for this?

Thank you very much in advance for your help. 

Best

[Laura Marshall]

Hello,

Thanks for your query.

That is correct to include all points in the analyses, recording sites with zero detections are essential for unbiased density estimation. Your current structure for the sample and observation tables sounds appropriate.

For modelling, you have a couple of sensible options: (1) model all data together using species groups or species as covariates, or (2) fit separate detection functions for groups, again allowing species as a covariate where indicated via a model selection. Using AIC to compare these competing approaches is appropriate, as long as models are fitted to the same dataset. AIC can also be used to decide whether to fit a single model or multiple models: just sum the AIC values across all models, provided the multiple models represent the same data as the single model.

Regarding your filtering step, it’s a bit unclear how the “>50% of total detections” rule was applied. In general, species with very few detections won’t support reliable estimation of a detection function, even if that is via including species as a covariate in the model. For those, as you have done, it can help to pool them with similar species (based on ecology or detectability) or perhaps refer to previous studies to guide grouping. However, without sufficient data to support the detection functions you are relying on your assumptions being correct.

Here is a reference to a paper describing grouping bird detections: Alldredge, M. W., Pollock, K. H., Simons, T. R., & Shriner, S. A. (2007). Multiple-species analysis of point count data: A more parsimonious modeling framework. Journal of Applied Ecology, 44, 281–290. https://doi.org/10.1111/j.1365-2664.2006.01271.x

Best wishes,
Laura