Revealing recurrent regimes of mid-latitude atmospheric variability using novel machine learning method

Area of Science:

• Atmospheric Science
• Climate Dynamics
• Nonlinear Data Analysis

Background:

• Low-frequency atmospheric variability involves persistent, hemispheric-scale states (teleconnection patterns/regimes).
• Identifying and dynamically representing these states is crucial for improving seasonal and longer-term weather predictability.
• Current methods face challenges in reliably detecting and characterizing these atmospheric regimes.

Purpose of the Study:

• To introduce a novel data-driven method for detecting recurring atmospheric variability regimes.
• To obtain dynamical variables that effectively represent these identified regimes.
• To enhance the modeling and forecasting of large-scale atmospheric circulation.

Main Methods:

• Combines recurrence quantification analysis and kernel principal component analysis from nonlinear dynamics.
• Applies the method to a quasi-geostrophical atmospheric circulation model.
• Utilizes reanalysis data of geopotential height anomalies for Northern Hemisphere winter seasons (1981-present).

Main Results:

• Successfully detected recurring regimes of atmospheric variability.
• Identified a set of dynamical variables that serve as an embedding for these regimes.
• Demonstrated that the detected regimes explain large-scale weather patterns, including severe winters in Eurasia and North America.

Conclusions:

• The new method effectively identifies atmospheric regimes and their dynamical variables.
• This approach offers significant prospects for improving empirical modeling of atmospheric circulation.
• The findings pave the way for enhanced long-term forecasting of climate variability and extreme weather events.