How decisions about fitting species distribution models affect conservation outcomes

Area of Science:

• Ecology
• Conservation Biology
• Biodiversity Informatics

Background:

• Species distribution models (SDMs) are crucial tools for conservation and land-use planning, informing biodiversity patterns.
• The construction of SDMs involves numerous choices (data prep, variable selection, fitting, evaluation) that impact predictive outcomes.
• A significant challenge is the lack of true species distribution data, leading to uncertainty in model selection and application.

Purpose of the Study:

• To analyze the impact of 11 routine decisions in SDM construction on conservation priority patterns.
• To quantify the relative influence of different modeling choices on conservation planning outcomes across multiple species.

Main Methods:

• Employed MaxEnt for building SDMs and Zonation for deriving conservation priority ranks for 25 species.
• Systematically varied parameters including model complexity, predictor variables, bias treatment, and threshold settings.
• Evaluated changes in conservation priority rankings based on different SDM configurations.

Main Results:

• All tested SDM variants demonstrated good performance (AUC > 0.7) but yielded spatially divergent predictions and conservation priorities.
• Decisions regarding bias treatment and binary threshold application had the most substantial impact, altering rankings in 40% and 35% of cells, respectively.
• Model complexity, data quantity, and bias correction methods also influenced conservation solutions, with varying degrees of impact on priority area overlap.

Conclusions:

• Model-based uncertainty is inherent in SDM applications for conservation planning.
• Understanding and addressing uncertainty, particularly concerning decisions with high impact, is critical for robust conservation strategies.
• Planners should consider alternative modeling approaches and uncertainty quantification when making critical land-use and conservation decisions.