Paleoclimate Model-Data Comparisons of Hydroclimate over North America with a Focus on Megadroughts

For the first time in the history of the Intergovernmental Panel on Climate Change, the Paleoclimate and Coupled Model Intercomparison Projects Phases 3 and 5 produced an ensemble of forced transient simulations of the last millennium. This wealth of model output, when combined with a growing collection of high spatial and temporal resolution pa- leoclimate estimates of past climate variability, represents an important and unprecedented source of information on climate variability over decades to centuries. This dissertation thus combines paleoclimate evidence with climate modeling to define a physical and statistical paradigm through which to analyze these combined sources of information and subsequently to characterize the features of the North American climate system that cannot be sufficiently understood using instrumental data alone. This includes features that have long timescales of variability or that are rare, and by consequence have few degrees of freedom over the short instrumental interval (1850 C.E. to Present), as well as interannual dynamical relationships that, while potentially well characterized by observations, are non-stationary. An integrative approach to analyzing these features or relationships serves two fundamental purposes: 1) It provides a more comprehensive characterization of past climate variability, albeit with the caveat of model bias, to clarify understanding of the dynamics that produce these features or relationships in the real world; and 2) it assesses whether coupled general circulation models (CGCMs) are able to simulate these features or relationships, which is necessary to determine that state-of-the-art CGCMs can accurately constrain the risk of future climate change. The focus herein will be on multidecadal hydroclimate change, or megadroughts, in the paleoclimate record of the American Southwest to better inform our understanding of the risk of future hydroclimate change over the region. Two fundamental understandings are derived from this work. Firstly, CGCMs are successful at simulating discrete periods of mul- tidecadal hydroclimate change that are characteristic in length, magnitude, and frequency of occurrence of megadroughts in the paleoclimate record. The simulated megadroughts are not tied in any coherent way to exogenous forcing, however, suggesting that CGCMs simulate large-magnitude internal variability on multidecadal timescales. Secondly, the dynamical characteristics of CGCMs are important in determining the atmosphere-ocean variability that drives multidecadal hydroclimate change. The dynamical characteristics of relevance include teleconnection realism and stationarity, the magnitude of ocean variability, and the relative magnitudes of different modes of atmosphere-ocean and purely atmospheric vari- ability. Additionally, a new understanding of real-world megadrought dynamics is derived herein, with the characteristics of some CGGMs providing a better representation of these dynamics