Open Population Spatial Capture Recapture: a Gentle Introduction with an Application

Speaker: Richard Glennie (University of St Andrews)

What is this seminar about?

Open population spatial capture-recapture (openSCR): a statistical method to estimate detectability, survival, recruitment, immigration/emigration, density, and spatial distribution of a population from repeated detections (by visual transect, camera trapping, hair snares, or acoustics) of identifiable individuals across multiple detectors.

Who is this seminar for?

This is a seminar of two halves. In the first half, I will give a gentle statistical introduction to these methods for people who are unfamiliar with them. In the second half, I present a case study that I hope will interest both beginners and experienced users of these methods.

What will the gentle introduction cover?

I’ll describe the intuition behind the complex statistical ideas (e.g. latent variable regression, continuous-time hidden Markov models and thinned point processes) that are brought together to make inference from openSCR possible. The methods are presented in Glennie, R, Borchers, DL, Murchie, M, Harmsen, BJ, Foster, RJ. Open population maximum likelihood spatial capture-recapture. Biometrics. 2019; 75: 1345 – 1355. https://doi.org/10.1111/biom.13078

What is the case study?

The case study is a multi-year, photo-ID survey on bottlenose dolphins in Barataria Bay, Louisiana that has been monitored due to the impact of the nearby Deepwater Horizon oil spill. Over 2000 capture histories have been collected between 2010 and 2019. The aim is to estimate how density varies across space and how population dynamics vary over time.

What will I find out during the seminar?

What we will find is that inference using methods can provide a rich and detailed picture of a population. However, we will also find that openSCR presents challenges and shortcomings in all model components: detection, movement, population dynamics, and spatial distribution. I will argue that each issue that arises in the case study points to more general problems in applying openSCR. Calling into question the assumptions and complex models being employed and (I hope) instigating further research.