Modelling the distribution of species and communities accounting for detection using R and BUGS/JAGS
Instructors: Gurutzeta Guillera Arroita, José Lahoz-Monfort (University of Melbourne)
& Marc Kéry (Swiss Ornithological Institute)
Date: 11–15 January 2016
Venue: Mediterranean Institute for Advanced Studies (IMEDEA, CSIC-UIB),
Esporles (Mallorca), Spain
Computers: Bring your own laptop with latest R, JAGS and WinBUGS or OpenBUGS
Costs: 500 Euro (includes course materials and snacks/coffee/tea)
Species distribution modelling has become a fundamental tool in ecology and conservation, where it underpins many applications. However, species distribution models are not always fit for purpose. How informative or useful a model is strongly depends on the type of question and also on the type of data used to fit the model. This course gives a critical overview of existing techniques for species distribution modelling for single species as well as for entire communities. We start by presenting methods for presence-only, presence-background and presence-absence data, discussing their strengths and limitations. We then introduce a key class of models for the analysis of species distributions that explicitly accounts for the observation (or measurement error) process that underlies all species distribution data. This framework, the focus of the course, includes site-occupancy models (MacKenzie et al. 2002, 2003; Tyre et al. 2003) and extensions for modelling abundance, range dynamics and communities. The course partly follows the brand-new book “Applied hierarchical modeling in ecology” (Academic Press, 2015) by Kéry & Royle. Model fitting is shown using the Bayesian BUGS software and the R package unmarked. Importantly, the course showcases the practical model fitting of community, or multi-species, occupancy models highlighted in recent papers such as Iknayan et al. (TREE, 2014) and Mihaljevic et al. (Ecology, 2015).
Course modules include the following:
• Introduction to species distribution modelling: questions/applications, theory and data types
• Overview of presence-only, presence-background and presence-absence methods
• Static (= single-season) occupancy-detection models in R and BUGS
• Dynamic (= multi-season) occupancy-detection models in R and BUGS
• Models for communities, including co-occurrence models
• Overview of extensions and related models, e.g. abundance models, spatial autocorrelation, and integrated models (combining data types)
In addition, the course will start with an introduction to Bayesian statistical modelling using BUGS software and likelihood-based estimation using R. In this applied workshop about 80% of the time is spent on lecturing and the remainder on solving exercises. No previous experience with program WinBUGS, or Bayesian statistics, is assumed. However, a good working knowledge in R programming and of modern regression methods (e.g. generalised linear models) is required.
Please send your application to Guru (email@example.com), describing your background and knowledge in statistical modelling, R and WinBUGS/OpenBUGS/JAGS, by 1 December 2015 at the latest. Workshop invitations will be sent out before 4 December 2015.
Travel information: Esporles is a village at about 14 km north of Palma, the capital of Mallorca. Several companies offer low-cost flights to Mallorca from nearly every European capital. There are regular buses between Palma and Esporles (approx. 35min journey, about 3 € return). Information about accommodation options and a detailed course schedule will be sent out with the workshop invitations.