Multidecadal Variability in Climate Models and Observations
Climate change attribution and prediction using state-of-the-art models continue to garner an ever-growing focus amongst both the scientific community and public alike. Recent analyses showing discrepancies in the structure of modeled and observed decadal climate variability (DCV), therefore, have e...
| Main Author: | |
|---|---|
| Format: | Text |
| Language: | unknown |
| Published: |
UWM Digital Commons
2018
|
| Subjects: | |
| Online Access: | https://dc.uwm.edu/etd/2004 https://dc.uwm.edu/context/etd/article/3009/viewcontent/Oser_uwm_0263m_12295.pdf |
| Summary: | Climate change attribution and prediction using state-of-the-art models continue to garner an ever-growing focus amongst both the scientific community and public alike. Recent analyses showing discrepancies in the structure of modeled and observed decadal climate variability (DCV), therefore, have engendered efforts to not only diagnose the dynamics underpinning observed DCV, but also to characterize the behavior of DCV within climate models. In this thesis, we employ Multichannel Singular Spectrum Analysis (M-SSA) to show that while the DCV signal in observations is best described as a coherent oscillation with complex propagation across the globe, modeled DCV lacks this structure altogether. Specifically, the modeled DCV has a considerably smaller magnitude than its observed counterpart, and tends to exhibit simpler spatiotemporal behaviors. In particular, within the vast majority of models, the DCV structure is best characterized either by globally synchronous, quasi-oscillatory patterns lacking propagation, or, secular trends punctuated with weak, oscillatory-like signals. Both observed and simulated DCV has the largest magnitude in the polar regions. However, the observed anomaly propagation suggests Atlantic control, whereas it is the Arctic that appears to be setting the tone for globally averaged variability in most model runs. Broadly, these results confirm contrasting DCV structure within models and observations, while identifying some qualitative commonalities between the observed and simulated quasi-oscillatory behavior within a few model simulations, thus providing important clues for further DCV research. |
|---|