Inferring causation from time series in Earth system sciences.

Area of Science:

• Earth system science
• Complex dynamical systems
• Causal inference

Background:

• Real experiments are infeasible in large-scale complex dynamical systems like Earth's system.
• Traditional correlation techniques are insufficient for understanding causality.
• Vast amounts of observational and simulated data are now available.

Purpose of the Study:

• Provide an overview of causal inference frameworks.
• Identify generic application cases for causal methods in Earth system sciences and beyond.
• Introduce a benchmark platform to bridge the gap between method users and developers.

Main Methods:

• Overview of causal inference frameworks.
• Identification of data-driven causal methods.
• Discussion of challenges in applying causal inference.

Main Results:

• Causal inference offers a powerful alternative to correlation for complex systems.
• Promising applications of causal methods identified across various scientific domains.
• Initiation of the causeme.net benchmark platform.

Conclusions:

• Data-driven causal inference is crucial for advancing Earth system science.
• The causeme.net platform aims to foster collaboration and standardize causal method application.
• Bridging the gap between causal inference methods and their users is essential for scientific progress.