Date of Award

May 2024

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Atmospheric Science

First Advisor

Sergey V Kravtsov

Abstract

Superimposed on the linear upward trend of observed global surface air temperature anomalies since the late nineteenth century is, what appears to be, a multidecadal undulation. However, this undulation is either muted or virtually absent in both the previous and current generations of the state-of-the-art climate models used to not only simulate past climates but also predict future climates. One possibility is that this signal is due to a series of complex responses to the global climate forcing; an alternative is that this signal is contained within the internal variability and teleconnected via atmospheric channels. Either way, the existential threat of anthropogenic climate change fuels the need to better simulate both the natural and forced climate variability.

We utilized many runs (ensemble simulations) from two generations of climate models, paired with linear pattern recognition methods to optimally isolate the forced signal, and further identify the dominant multidecadal variability in the, presumably, unforced residual; this variability in the observed data represents the so-called “global stadium wave.” The methods underlying the forced signal identification are based on the notion that the ensemble averaging of multiple historical simulation tends to retain the forced signal and average out the internal variability. The signal-to-noise-maximizing pattern filtering allows one to accurately estimate the forced signal using a smaller number of ensemble members compared to the direct ensemble averaging. The main statistical tool we used in extracting the stadium wave from the estimated internal variability is the Multichannel Singular Spectrum Analysis (M-SSA).

We compared the magnitudes, spatial patterns, and periods of the dominant, M-SSA derived, multidecadal signal contained within two different reanalysis datasets and the model simulations. We first performed the stadium-wave identification using a limited set of observed and model simulated Northern Hemisphere’s climate indices and developed a new technique to compare the observed and model simulated stadium waves. This was followed by the global analysis, which confirmed the results of the index-based analysis. We further tested the robustness of the observed stadium wave in the global analysis by: 1) using restricted regional datasets from around the globe to attempt to replicate a reference pattern (i.e., the signal obtained from using the full global dataset); 2) introducing variants of the original data gaps that plague the earlier parts of the timeseries and (particularly) the Southern Hemisphere, filling them in, and then filtering the data with the same M-SSA filtering method; and 3) extending the beginning and end of the timeseries to test the end effects, given the analysis of a multidecadal signal over a relatively short instrumental period.

We found that regardless of the inhibiting factors, the same observed signal is replicated in all versions of the regional or global analyses. We also confirmed that found that neither of analyzed climate models is able to accurately simulate the observed signal. Some newer-generation models are able to simulate magnitudes equal or even above observations, yet produce vastly different spatial patterns. We used these finding to direct our future analysis by using how the simulations differ from observations to answer why the models differ.

Share

COinS