Simulation-Based Spatially Explicit Close-Kin Mark-Recapture.

Area of Science:

• Ecology
• Conservation Biology
• Computational Biology

Background:

• Estimating wild population size is vital for conservation but traditional methods are costly and labor-intensive.
• Genetic methods, like close-kin mark-recapture (CKMR), offer a less invasive and potentially cheaper alternative.
• Existing CKMR models struggle with spatially structured populations and heterogeneous sampling.

Purpose of the Study:

• To develop a spatially explicit close-kin mark-recapture (CKMR) method for population size estimation.
• To address limitations of current CKMR models in handling spatial structure and sampling heterogeneity.
• To create a more accurate and robust tool for ecological and conservation assessments.

Main Methods:

• Developed CKMRnn, a spatially explicit approach combining individual-based simulation with deep convolutional neural networks.
• Utilized simulations to test the accuracy and robustness of CKMRnn under various spatial conditions.
• Applied CKMRnn to empirical data from a Ugandan elephant population.

Main Results:

• CKMRnn demonstrated high accuracy in population size estimation, even with spatial heterogeneity and population structure.
• The method effectively accounted for potential confounding factors like unknown population histories.
• Empirical application to Ugandan elephants yielded point estimates consistent with traditional methods but with a ~30% narrower confidence interval.

Conclusions:

• CKMRnn provides a powerful, accurate, and robust method for estimating population sizes in spatially structured populations.
• This novel approach enhances the utility of genetic data for wildlife population monitoring and conservation.
• The improved precision of CKMRnn can lead to more informed conservation and management decisions.