Automatic variable selection in ecological niche modeling: A case study using Cassin's Sparrow (Peucaea cassinii)

Area of Science:

• Ecological modeling
• Climate science
• Computational biology

Background:

• Ecological niche modeling (ENM) often struggles to directly incorporate complex global climate model (GCM) outputs.
• Feature selection is crucial for reducing dimensionality and identifying relevant environmental variables for ENM.

Purpose of the Study:

• To introduce MERRA/Max, a novel feature selection approach for dimensionality reduction in ENM.
• To enable the direct use of GCM outputs in ecological niche modeling.
• To streamline the process of identifying key environmental predictors for species distribution modeling.

Main Methods:

• MERRA/Max employs a Monte Carlo optimization with multiple MaxEnt runs on selected variables.
• The algorithm operates on filesystem-stored data, ensuring scalability for large datasets.
• Software components are designed for parallel processing in cloud environments for high performance.

Main Results:

• MERRA/Max successfully identified key bioclimatic predictors for Cassin's Sparrow (Peucaea cassinii) from standard datasets.
• It also selected ecologically plausible predictors from a comprehensive set of 86 variables derived from NASA's Modern-Era Retrospective Analysis for Research and Applications Version 2 (MERRA-2).
• The method demonstrated efficiency and scalability in selecting relevant environmental variables.

Conclusions:

• MERRA/Max offers a technological solution to expand the use of GCM outputs in ENM.
• It facilitates exploratory analysis of climate data and streamlines the bioclimatic modeling process.
• The approach has the potential to enable automated, accessible, and low-cost bioclimatic modeling services.