To foster international participation, this course will be held online

The course will introduce and demonstrate the use of Bayesian Additive Regression Tree (BART) methods for species distribution modeling (SDM) and other ecological applications. We will explain how BART modeling works, and identify how BARTs improve over other commonly used SDM methods. Participants will work through all steps necessary for conducting a SDM study with BARTs in R, using utilities in the embarcadero and dbarts packages to select informative environmental predictors, train and evaluate BART models of species occurrence, and use trained models to predict species presence or absence from new data.

The course is aimed at advanced students, researchers, and practitioners who have some familiarity with species distribution modeling and want to expand their experience and skills. Participants will need a laptop with a webcam and a good internet connection to participate in interactive live sessions. They should be comfortable working in R, particularly the Rstudio environment, and they should be prepared to install specialized packages, edit and write simple scripts, and manage and read in downloaded datasets.

By the end of the course, participants will:

•    Understand the structure of BART machine-learning models and how they compare to similar ML methods, especially for species distribution modeling
•    Use embarcadero and dbarts utilities to select predictors, and train BART models of species occurrence
•    Visualize and interpret predictor effects and interactions in a trained BART model
•    Use trained BART SDM models to project species distributions into new regions or times

Daily schedule:

15:00 - 18:00 (Berlin time): live lectures and introduction to / review of the practicals
4 additional hours: self-guided practicals using annotated R scripts, with online support

Pre course: Self-guided introduction and installation of necessary packages

Day 1: Species Distribution Models and Bayesian stats

      o    SDMs overview/review
      o    Worked demonstration of SDMs with Boosted Regression Trees
      o    Bayesian stats overview/review
      o    Worked demonstration of Bayesian regression with rstan

Day 2: BARTs and embarcadero

     o    BARTs, theory and comparison to other ML methods for SDMs
     o    Worked demonstration of an SDM with BARTs, using embarcadero — predictor selection, model training, and prediction with new data

Day 3: BART SDM workflow: predictor selection, model evaluation, troubleshooting

   o    Inspecting predictor partial effects
   o    Spatial partial effects
   o    Random-intercept BARTs

Cancellation Policy:

> 30  days before the start date = 30% cancellation fee

< 30 days before the start date= No Refund.

Physalia-courses cannot be held responsible for any travel fees, accommodation or other expenses incurred to you as a result of the cancellation.