Exploring North Atlantic decadal variability using machine learning
Restricted (Penn State Only)
- Author:
- Gu, Qinxue
- Graduate Program:
- Meteorology and Atmospheric Science
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- May 25, 2023
- Committee Members:
- David Stensrud, Program Head/Chair
Andrew Carleton, Outside Unit & Field Member
Melissa Gervais, Chair & Dissertation Advisor
Laifang Li, Major Field Member
Raymond Najjar, Major Field Member - Keywords:
- Decadal variability
Machine learning
Self-organizing map
Atlantic Multidecadal Variability
North Atlantic Oscillation
Internal variability
Climate change
Atmosphere-ocean-sea ice interactions
North Atlantic Ocean
Decadal variability
North Atlantic Ocean
Machine learning
Self-organizing map
Atlantic Multidecadal Variability
North Atlantic Oscillation
Internal variability
Climate change
Atmosphere-ocean-sea ice interactions
Decadal variabilityachine learning×Self-organizing map×Atlantic Multidecadal Variability×North Atlantic Oscillation×Internal variability×Climate change×Atmosphere-ocean-sea ice interactions×North Atlantic OceanDecadal variability
Machine learning
Self-organizing map
Atlantic Multidecadal Variability
North Atlantic Oscillation
Internal variability
Climate change
Atmosphere-ocean-sea ice interactions
North Atlantic OceanDecadal variabilityNorth Atlantic OceanMachine learningSelf-organizing mapAtlantic Multidecadal Variability - Abstract:
- Decadal variability in the North Atlantic extensively impacts regional and global climates. However, due to the complex interactions within the climate system, significant gaps still remain in our understanding of the mechanisms associated with North Atlantic decadal variability. This dissertation investigates the mechanisms associated with North Atlantic sea surface temperature (SST) internal variability and its changes under global warming on decadal time scales. Employing a novel evolution self-organizing map method, we identified dominant spatiotemporal SST evolutions over 10 years in the Community Earth System Model version 1 (CESM 1) preindustrial control simulation. These evolutions include but are not limited to those associated with the North Atlantic Oscillation (NAO) and Atlantic Multidecadal Variability (AMV). The key mechanisms responsible for the well-known interactions between the NAO and AMV are firstly identified, encompassing buoyancy-driven and wind-driven ocean circulation as well as ocean--atmosphere transient-eddy forcing. Diagnosing the remaining evolutions, it is found that the contributions of resolved ocean advection and surface heat fluxes to the upper ocean temperature tendency seldom counteract each other over 10-year periods in the subpolar North Atlantic. Additionally, we detect ocean--atmosphere transient-eddy feedbacks in almost all the evolutions, with feedback strength tied to the specifics of SST patterns. Finally, we discover that SST evolutions with comparable initial states but divergent trajectories are linked to abrupt shifts in atmospheric variability and are likely unpredictable. Utilizing the CESM2 large ensemble and preindustrial control simulation, we identify a divergence of northern North Atlantic SST among ensemble members around the mid-21st century, which is associated with divergences in large-scale ocean circulation and Labrador Sea deep convection. Further investigation indicates that this divergence can be triggered by stochastic atmospheric variability and amplified by two positive feedback loops: the first involving vertical mixing and surface salinity and the second Arctic sea ice and wind stress. We propose that global warming activates these feedbacks by intensifying the vertical salinity gradient. In summary, this dissertation provides a comprehensive understanding of North Atlantic decadal variability and emphasizes the importance of coupled atmosphere--ocean--sea ice dynamics in generating the North Atlantic decadal variability and its changes due to anthropogenic forcing.