CREAF

TITLE: "The dynamic occupancy (dynocc) model: one of the most under-utilised statistical models in ecology"

DATE: Thursday, 30 November 2023.

TIME & FORMAT: form 12 to 1pm CET - In-person and online.

Seminars will combine in-person and online formats (CREAF, Sala Graus II, Universitat Autònoma de Barcelona, Spain) but in all cases, talks will be always streamed (not recorded), so they can be followed online.

HOW TO CONNECT: direct link to Mark Kéry's conference.

SUMMARY OF THE WORKSHOP:

The dynamics of species distributions and their drivers have always fascinated researchers in ecology and related sciences like biogeography, conservation biology, fisheries and wildlife management. These topics have gained additional importance in the context of land use and climate change, when accurate assessments of change and its drivers become urgent. A number of statistical and machine-learning methods have been applied to the study of changing distributions, but surprisingly, most deal with dynamics by simply applying static, i.e., non-dynamic, models at each time point separately; only rarely do people adopt models that describe change explicitly. However, such a dynamic model of change was developed exactly 20 years ago: the dynamic occupancy model of MacKenzie et al. (Ecology, 2003). It describes dynamics of presence/absence between times t and t+1 by colonization and extinction rates and enables both to be modelled with covariates, to sharpen predictions and identify drivers of change. The "dynocc" is a conceptually simple but very powerful model. It can easily be fit to presence/absence data collected at two or more discrete points in time. However, I would claim that it is not nearly as widely known and used as it should. I will present the model, give some examples of its many possible uses as well as describe some of its many extensions.

In the afternoon of that same day, together with Oriol Solà I will give a mini-course on the dynamic occupancy model. We will show implementations in various R packages (unmarked, ubms) as well as custom implementations in Bayesian modeling engines (JAGS, NIMBLE). The duration will be about 3h. For the most part, this course will be a demo, where we walk people through a code while we execute it and give plenty of comments on what we are doing and why, also inviting people to actively participate and taking any questions they might have.

ABOUT THE SPEAKER:

Marc Kéry works as a senior scientist at the Swiss Ornithological Institute, Sempach, Switzerland. This is a non-profit NGO with about 180 employees dedicated primarily to bird research, monitoring and conservation. Marc was trained as a plant population ecologist at the Swiss Universities of Basel and Zurich. After a 2-year postdoc at the (then) USGS Patuxent Wildlife Research Center in Laurel, MD, where he learned to understand and how to apply capture-recapture models to plant demographic data, he got hooked on statistical ecology, especially in the context of estimating and modeling distribution, abundance, species richness, and population demography using modern statistical modeling tools. During the last 20 years he has worked at the interface between population ecology, biodiversity monitoring, wildlife management, and statistics. He has published more than 100 peer-reviewed journal articles and five text books on applied statistical modeling.