| Summary: | <p> This thesis explores the concepts and techniques of observation-model comparisons of the natural variability of the near-surface ocean on three different time scales. The emphasis on natural variability includes removing the all-time trend and seasonality of the data. All analyses used model outputs of the Community Climate System Model version 3.5 (CCSM3.5). </p><p> The first work, <u>An Ensemble Observing System Simulation Experiment of Global Ocean Heat Content Variability,</u> introduces the use of ensemble of model time series to study how a set of observations and how they are processed can capture the statistics of the system being observed. The technique is applied to global ocean heat content (OHC) down to 700m as observed and processed by the In-Situ Analysis System 2013. This study found that before the implementation of the global Argo program (1990-2005), the observed variability is too significantly biased by the low spatial resolution of the observations to return any meaningful estimates of global OHC variability with a median correlation score of 60% and a signal to noise ratio (SNR) of 1.9. The Argo era (2005-2013) is found to do a much better job at estimating global OHC variability to a median correlation score to 95% and an SNR of 14.7. However, this is only true for annual running means and longer; sub-annual variability is still unreliably resolved. </p><p> The second work, <u>Probability Angular Momenta of Multidecadal Oscillations of the North Atlantic,</u> explores concepts in non-equilibrium statistical mechanics, specifically probability angular momentum, as new tools in observation-model comparisons. The indices analyzed include an index related to the Atlantic Multidecadal Oscillation (AMO) and indices related to other oscillations thought to influence the observed variability in the AMO; the atmospheric North Atlantic Oscillation (NAO), the subsurface-ocean Atlantic Meridional Overturning Circulation (AMOC), and outflow from the Labrador Sea (LSO). The PAM analysis was found to detect cycles of the same magnitude and sign as traditional analyses for the simulated indices; for example, the -NAO leads +AMOC by 2 years, +AMO leads -NAO by 10-20 years, and PAMV leads +AMOC by 2-20 years, although the PAM results typically had too low of confidence to support any conclusions from the observed data. The PAM technique also returned a novel insight; a staistically-significant oscillation in the simulated LSO and AMO on the order of 400-1000 years. Since the model output has a time span of only 720 years, this indicates that the PAM technique may be able to detect modes of oscillation with periods on the order of or longer than the time span of the data analyzed, something that cannot be done to any statistical significance via traditional correlation and spectral techniques. </p><p> The final work, <u>PhaseMap: Comparison of Late Pleistocene Surface Temperature Proxies to an Accelerated CCSM3 Simulation,</u> compares the simulated ocean surface in a CCSM3 model run forced using the last 300,000 years of climate forcings to 50 paleotemperature proxies from deep ocean cores around the world. The accelerated model, which was accelerated 100x to simulate 300,000 years of climate in 3,000 model years, was found to agree poorly with the core proxies. While the core proxies correlate strongly with greenhouse gas, ice volume, and sea level forcings, the model results primarily follow the local insolation. It is unclear from this analysis whether this disagreement results from the model being too sensitive to insolation forcing, not sensitive enough to other forcings, or from the fact that the model's subsurface ocean doesn't respond quickly enough to the accelerated forcings. </p><p> These three different fields of ocean study are also inter-compared to explore their individual strengths and weaknesses, and where the techniques of one field may be useful in another. The modern subsurface ocean observations are plagued with uncertainties, and applying the observing system properties to a model was shown to help interpret the uncertainties associated with the spatio-temporal variability in the number and frequency of observations as well as the methodology used to create global maps from these observations. Paleoceanographers often have to work with proxy data that are unevenly sampled in time, and techniques commonly used to mitigate this issue (e.g. Lomb-Scargle method of periodogram estimation) can be used in modern studies where observational data is not available for short periods of time. </p><p> These works explore and propose techniques and concepts regarding surface and near-surface ocean variability on different temporal scales. They also highlight the importance of establishing connections across disciplines working on these different temporal scales. Together, these results improve our understanding of the role of the ocean on the climate system we depend on, and how different disciplines in ocean science can work together to improve our understanding even further.</p><p>
|
|---|
