Authors: Anni Yang*, University of Florida, Juan Pablo Gomez, University of Florida, Jason K Blackburn, University of Florida
Topics: Biogeography, Quantitative Methods, Medical and Health Geography
Keywords: GARP, variable selection, environmental coverage, explanatory power, Bacillus anthracis
Session Type: Paper
Start / End Time: 1:10 PM / 2:50 PM
Room: Cleveland 1, Marriott, Mezzanine Level
Presentation File: No File Uploaded
Variable selection for, and determination of variable importance within, species distribution models (SDMs) remain an important area of research with continuing challenges. Most SDMs provide normally exhaustive searches through variable space, however, selecting variables to include in models is a first challenge. The estimation of explanatory power of variables and the selection of the most appropriate variable set within models can be a second challenge. Although some SDMs incorporate the variable selection rubric inside the algorithms, such as jackknife in MaxEnt and relative influences in boosted regression trees (BRTs), there is no integrated rubric to evaluate the variable importance in the Genetic Algorithm for Ruleset Production (GARP). Here, we design a novel variable selection methodology based on the rulesets generated in a GARP experiment. The importance of the variables in a GARP experiment can be estimated based on the consideration of explanatory power (the prevalence of each environmental variable in the dominant presence rules of the best model subset) and the exophysiological process (its median range across rules). The performance of this variable selection method were test via simulated species with weak and strong correlations between species occurrence and environment. A real-world case study of exploring ecological requirements and predicting the distributions of the A1.a sub-lineage Bacillus anthracis (the bacterium that causes anthrax) in the United States is provided as an application of the new variable selection procedures. Our new method provides an alternative of variable selection for GARP and allows users to explore the ecological requirements of species distributions.