Date of Award

August 2014

Degree Type

Thesis

Degree Name

Master of Science

Department

Mathematics

First Advisor

Anastasios Tsonis

Committee Members

Kyle Swanson, Sergey Kravtsov

Keywords

Climate, Discrete Fourier Transform, Dynamics, Multi-Periodic, Spectral Analysis, Temperature Records

Abstract

Analyzing 26 short-length (less than 3000 years) instrumental and proxy temperature records and five long-length (greater than 3000 years) proxy temperature records using Discrete Fourier Transform has shown that as the length of significant periods increase in the time domain then so does the power at which the period is observed. A t-test verifies that a positive correlation exist between the length of the significant periods and the power with a confidence level of ∝ >0.05. Significant frequencies with period greater than 30 years are confirmed using Monte Carlo simulations, which were created using a nonlinear approach known as fractional Brownian motion (FBM). While completing the spectral analysis it was observed in the spectral analysis is an absence of significant periods between about 1000 years and 20,000 years. Also, wavelet analysis of all short-length temperature records turned up no significant findings.

Share

COinS