A comparison of regression methods for model selection in individual-based landscape genetic analysis

Area of Science:

• Ecology
• Conservation Biology
• Population Genetics

Background:

• Human-caused barriers fragment species populations, hindering their ability to adapt to climate-driven habitat changes.
• Effective conservation requires maintaining habitat connectivity to create viable metapopulations and allow species to track shifting climate envelopes.
• Landscape genetics offers empirical methods to understand gene flow influenced by landscape features, informing conservation strategies.

Purpose of the Study:

• To evaluate the accuracy of seven regression-based model selection methods in landscape genetics.
• To determine which methods best identify landscape factors influencing gene flow under diverse scenarios.
• To provide guidance for selecting accurate models to inform conservation actions.

Main Methods:

• Conducted population genetic simulations across varied landscapes.
• Varied the number, type, magnitude, and cohesion of resistance variables.
• Assessed the impact of transformations on the relationship between genetic and landscape distances.

Main Results:

• Linear mixed effects models demonstrated superior accuracy in identifying the true landscape model influencing gene flow.
• Other methods performed well under conditions of high landscape resistance and low correlation between hypotheses.
• The study assessed model performance across various landscape complexities and resistance characteristics.

Conclusions:

• Linear mixed effects models are recommended for accurate landscape genetic analysis and informing conservation connectivity.
• The findings offer practical guidance for researchers and conservationists in selecting appropriate model selection methods.
• Accurate landscape genetic inferences are vital for effective conservation planning and mitigating fragmentation impacts.