Quantifying the spatial scales of animal clusters using density surfaces.

Area of Science:

• Ecology
• Spatial Statistics
• Animal Behavior

Background:

• Existing methods for quantifying animal clustering are often scale-independent or fail to represent animal perception.
• Anisotropic and hierarchical clustering patterns are not well-captured by current proximity models.

Purpose of the Study:

• To develop a novel, parameter-free method for quantifying animal clustering at multiple spatial scales.
• To automatically detect cluster diameters that reflect animal perception neighborhoods.
• To apply the method to both artificial and real-world animal datasets.

Main Methods:

• Utilized kernel density estimation to create a density function from point-location data.
• Employed smoothing kernels to automatically detect cluster diameters, mimicking animal perception.
• Tested the method on diverse artificial datasets and a dataset of African bush elephants.

Main Results:

• The method accurately assigns higher clustering values to scales with greater clustering.
• Successfully identified spatial scales corresponding to actual cluster diameters.
• Demonstrated robustness, with accuracy insensitive to the choice of kernel function.

Conclusions:

• The developed method offers a robust, parameter-free approach for analyzing animal clustering.
• It is particularly effective for anisotropic and hierarchically clustered systems.
• Facilitates the detection of relevant spatial scales in animal group dynamics.