Eco-ISEA3H, a machine learning ready spatial database for ecometric and species distribution modeling

Area of Science:

• Ecology
• Geoinformatics
• Computational Biology

Background:

• Ecological modeling, particularly species distribution modeling, requires comprehensive global datasets.
• Existing datasets often lack the integration and consistent spatial referencing needed for large-scale machine learning applications.
• Accurate environmental and geographical data are crucial for predicting species ranges under changing conditions.

Purpose of the Study:

• To introduce the Eco-ISEA3H database, a novel compilation of global spatial data.
• To provide data suitable for machine learning-based ecological and species distribution modeling.
• To facilitate continental- to global-scale ecometric analyses and predictions of species ranges.

Main Methods:

• Compilation of global spatial data on climate, geology, land cover, geography, and mammalian species ranges.
• Integration of data from 17 sources, totaling 3,033 variables.
• Organization of data using the Icosahedral Snyder Equal Area (ISEA) aperture 3 hexagonal (3H) discrete global grid system (DGGS) at six nested resolutions.
• Development and release of scripts for data incorporation.

Main Results:

• The Eco-ISEA3H database successfully integrates diverse global datasets into a unified hexagonal grid system.
• The database supports machine learning applications by providing consistent observational units for species occurrence and environmental data.
• A case study demonstrated the database's utility in analyzing bioclimatic envelopes for ten large mammalian species.

Conclusions:

• The Eco-ISEA3H database is a valuable resource for large-scale ecological modeling and species distribution studies.
• The DGGS approach ensures equal-area representation, crucial for accurate spatial analyses.
• This database enables robust predictions of species' responses to past and future environmental changes.