Energetic and paleoclimatic constraints on spatial patterns of climate variability

Abstract

Understanding the relative contributions of forced and unforced climate variability across time and space is a central challenge in physical climatology. In this thesis, I combine theory, a hierarchy of climate-model simulations, and paleoclimate proxy records to investigate the spatial and frequency-dependent structure of temperature variability. I first develop an energetic framework for climate variability. I run a set of idealized transient and equilibrium experiments in a general circulation model (GCM) that consists of a dynamic atmosphere coupled to a slab ocean, and interpret these results with two simple energy-balance models: a one-dimensional model for global-mean variability, and a two-dimensional model for the spatial pattern of variability. I find that virtually all key features of forced variability in the GCM—including polar amplification of low-frequency temperature variability—can be understood as a consequence of radiative feedbacks and heat transport. On high frequencies, unforced variability is introduced through stochastic tropical clouds, which is not represented in the energy-balance models. This work supports the idea that, in principle, low-frequency forced temperature variability can constrain equilibrium climate feedbacks, provided known radiative forcing and sufficiently long records (multiple centuries of temperature data) are available. However, high-frequency stochastic tropical variability in top-of-atmosphere radiation, combined with long ocean memory, introduces substantial uncertainty into attempts to constrain equilibrium climate feedbacks from sub-decadal variability. I then shift focus to the paleoclimate record, with the goal of determining whether paleoclimate proxies can distinguish between forced and unforced temperature variability. Using ensembles of fully-coupled forced and unforced GCM simulations, I generate predictions for key metrics of temperature variability (spectral scaling and spatial coherence), and compare them with corresponding estimates from marine sediment, tree-ring, and coral proxy records. Across these archives, I find limited evidence for the large-scale coherent signatures of external forcing predicted by models. Marine sediment and tree-ring records exhibit weak spatial coherence and highly variable spectral characteristics, consistent with substantial non-climatic noise or underestimated (local) internal variability in models. Coral records present a more promising picture: they capture tropical variability dominated by internal dynamics on interannual timescales and are situated in regions that may help distinguish forced from unforced variability on multidecadal and longer timescales. They also show reasonable consistency across proxy types. However, coral records are generally much shorter than marine sediment and tree-ring records, limiting their ability to resolve variability at the longest timescales. Together, these results highlight a fundamental tension between modeled and observed variability and underscore the challenges of detecting forced signals in proxy records. Progress will require improved proxy-system models, better-calibrated reconstructions, replicate proxy measurements at proximal sites, and longer climate-model simulations to robustly characterize low-frequency variability