Predicting Flow Reversals in a Computational Fluid Dynamics Simulated Thermosyphon Using Data Assimilation.

Area of Science:

• Fluid dynamics
• Chaos theory
• Computational physics

Background:

• Thermal convection loops exhibit chaotic fluid flow, analogous to Earth's weather systems.
• Predicting these chaotic dynamics is challenging due to limited observational data and model uncertainties.

Purpose of the Study:

• To develop and verify data assimilation (DA) methods for predicting chaotic flow reversals.
• To compare the efficacy of adaptive versus static localized covariance in ensemble-based DA.
• To identify predictive modes of chaotic behavior using Dynamic Mode Decomposition (DMD).

Main Methods:

• Built and verified four distinct data assimilation methods.
• Performed twin model experiments using a computational fluid dynamics simulation of a thermal convection loop.
• Employed the Ensemble Transform Kalman Filter (ETKF) for data assimilation.
• Utilized Dynamic Mode Decomposition (DMD) to analyze temperature and velocity fields.

Main Results:

• Adaptive localized covariance outperformed static localized covariance within the ETKF framework.
• Fewer observations were required for accurate prediction of flow reversals when using adaptive covariance.
• DMD successfully recovered the low-dimensional system governing reversals and identified predictive modes.

Conclusions:

• Adaptive data assimilation techniques offer improved prediction of chaotic phenomena in fluid systems.
• DMD can effectively identify key dynamical modes for forecasting complex behaviors like flow reversals.
• These findings have implications for improving weather and climate modeling accuracy.