Recurrence measure of conditional dependence and applications.

Area of Science:

• Complex Systems Science
• Data-Driven Causality Inference
• Information Theory

Background:

• Causal relation identification from observational data is challenging.
• Transfer entropy is a common information theory tool but relies on probability estimation.
• Existing methods struggle with accurate causality inference in complex systems.

Purpose of the Study:

• To propose a novel data-driven causality inference approach.
• To integrate recurrence plot features with information theory.
• To quantify causal dependence using the Recurrence Measure of Conditional Dependence (RMCD).

Main Methods:

• Developed the Recurrence Measure of Conditional Dependence (RMCD).
• Incorporated recurrence plot features into an information theory framework.
• Quantified causal dependence by analyzing joint recurrence patterns between past and present states.

Main Results:

• The RMCD effectively quantifies causal dependence between processes.
• The method accounts for the influence of a potential driver on a driven variable.
• Successfully identified the influence of Pacific Ocean sea-surface temperature on Amazonian precipitation during droughts.

Conclusions:

• The RMCD offers a robust data-driven approach for causality inference.
• This method overcomes limitations of traditional entropy-based techniques.
• RMCD provides insights into time-scaled causal influences in complex environmental systems.