Local-manifold-distance-based regression: an estimation method for quantifying dynamic biological interactions with...

Area of Science:

• Ecology
• Complex Systems Analysis
• Time Series Analysis

Background:

• Quantifying species interactions is vital for ecological understanding.
• Nonlinear time-series analyses like S-maps show promise for ecosystem studies.
• Existing methods face limitations in Jacobian estimation accuracy with noisy, non-equilibrium data.

Purpose of the Study:

• To introduce a novel, robust analytical method for ecological interaction quantification.
• To address limitations in Jacobian estimation accuracy in nonlinear time-series analyses.
• To improve prediction and interaction effect estimation in complex ecosystems.

Main Methods:

• Introduced the local manifold distance (LMD) concept, a non-equidistant measure.
• Developed LMD-based regression (LMDr) by integrating LMD with advanced computation.
• Validated LMDr using synthetic time series and an experimental protozoan predator-prey system.

Main Results:

• LMDr demonstrated superior robustness to noise compared to existing methods.
• The method showed improved accuracy in Jacobian estimation for ecological interactions.
• Application to a protozoan system revealed better correspondence with known predator-prey dynamics.

Conclusions:

• LMDr offers a robust and efficient approach for analyzing complex ecological networks.
• The method enhances the understanding of species interactions under dynamic and noisy conditions.
• LMDr advances ecological time-series analysis for prediction and interaction quantification.