Organization of specialized course from the 27th February to 2nd March 2018, at the Institute of Marine Science (ICM-CSIC), Barcelona, Spain.
Teacher: Dr. Maria Gracia Pennino, Dr. David Conesa, Dr. Joaquín Martínez Minaya
Species Distribution Models (SDMs) are now widely used in many research fields for several purposes across terrestrial, freshwater, and marine realms. In all these contexts, the main issue is to link information on the presence/absence or abundance of a species to environmental variables to predict where (and how much of) a species is likely to be present in unsampled locations or time periods. In ecology, SDMs have been implemented in different theoretical and practical cases, including the identification of critical habitats, the study of the risk associated with invasive species, the potential effects of climate change, the design of protected areas and delineation of hot spots of biodiversity and species richness.
Many algorithms of spatial distribution models can be used to predict the distribution of species, however, these algorithms do not always provide accurate results as data can includes a large amount of variability and errors due to multiple factors (errors in the identification of the species, errors in taking the geographical coordinates , etc.). In this context, Bayesian spatial-temporal methods have several advantages over traditional ones, since they provide a more realistic and accurate estimate of uncertainty.
Objective: The general objective of this course is to provide a critical view of the existing techniques of species distribution models with a frequentist and Bayesian approach, discussing their strengths and limitations. At the end of the course, participants will learn about:
- Formal knowledge about the main statistical approaches relating to SDMs;
- Specific examples of SDMs with different model algorithms (GAMs, Random Forest, Boosted Regression Trees) using a frequentist approach;
- An introduction to the Bayesian inference and modeling framework;
- Specific examples of SDMs using a Bayesian approach