Causal interaction in high frequency turbulence at the biosphere-atmosphere interface: Structural behavior.

Area of Science:

• Atmospheric Science
• Environmental Physics
• Data Science

Background:

• High-frequency eddy covariance (EC) measurements capture land-atmosphere interactions at 10 Hz.
• Traditional spectral analyses focus on timescales of 15-60 minutes, often overlooking high-frequency interdependencies.
• Multivariate EC data contain rich information on variable interactions like wind, humidity, temperature, and CO2.

Purpose of the Study:

• To move beyond traditional spectral analyses of EC data.
• To explore and quantify causal dependencies among interacting variables at high frequencies.
• To develop a methodological framework for understanding causal structure evolution in turbulent systems.

Main Methods:

• Utilized 10 Hz EC data from an agricultural site in the Midwestern US.
• Quantified high-frequency interdependencies using directed acyclic graphs (DAGs).
• Applied distance-based classification and k-means clustering to analyze DAG structural evolution over time.

Main Results:

• Identified well-defined clusters representing similar causal dependency structures.
• Demonstrated the ability to characterize structural similarities and differences in causal interactions.
• Observed dynamic changes in causal structure during a clear sky day and a solar eclipse event.
• The method effectively selects an unbiased number of clusters.

Conclusions:

• The developed DAG-based approach provides a robust framework for analyzing high-frequency causal relationships in turbulent systems.
• This methodology enhances the understanding of how causal dependence manifests in complex environmental data.
• The study highlights the potential of high-frequency EC data for uncovering intricate atmospheric dynamics.