Overestimating Abundance Estimates - N-Mix Models

[Mike Boyd]

Hello all!

I have a question regarding the overestimating of abundance in my N-mixture models.

Background: point count survey with 142 sites and 3 visits. Models for 38 species after I eliminated flocking species, those with low number of detections and/or sites, or the species’ global model c-hat >2. Those with c-hat <>1 were adjusted accordingly.

Some total abundance estimates look ‘good’ (e.g., pewee estimate=12.57 vs 9 summed of max detections/sites) and others don’t (e.g., >3,500 b-c chickadees). I suspect it’s due to low detection estimates (<0.1) and/or its standard error overlaps with zero.

I’m concerned about estimates that fall in between (e.g., 61.68 song sparrow vs 26 survey total, 67.65 ruby-crowned kinglet vs 43 survey total). Is there a suggested cutoff for detectability or its SE where I can accept/discard the model (i.e. ,<0.1 detectability or det-SE=0)? Or other suggestions for when to accept/discard a model?

Many thanks!

Mike Boyd

[ismae...@hotmail.com]

Hi Mike,

There is this paper of Dennis et al. (2015) about the infinite estimates produced by N-mixture models "particularly when working with a limited number of sampling occasions and low detection probability":

[Dennis, E. B., Morgan, B. J., & Ridout, M. S. (2015). Computational aspects of N‐mixture models. Biometrics, 71(1), 237-246.]

It may be your case..

Another thing is that if you are using a negative binomial distribution for the local abundance parameter, that can produce unreliable huge estimates.

Kéry (2018)  discusses some diagnostics to identify cases of infinite abundance estimates and also the problem with the negative binomial distribution.

[Kéry, M. (2018). Identifiability in N‐mixture models: A large‐scale screening test with bird data. Ecology, 99(2), 281-288.]

An alternative to be able to estimate the parameters for those "poor-data" species would be fitting multi-species N-mixture models.. but I think they are only available under a Bayesian analysis.

Chandler, R. B., King, D. I., Raudales, R., Trubey, R., Chandler, C., & Arce Chávez, V. J. (2013). A small-scale land-sparing approach to conserving biological diversity in tropical agricultural landscapes. Conservation Biology, 27, 785–795.

Yamaura, Y., Kéry, M., & Royle, J. A. (2016). Study of biological communities subject to imperfect detection: Bias and precision of community n-mixture abundance models in small-sample situations. Ecological Research, 31, 289–305.

Regards

Ismael Brack