Date of Award




Document Type

Master's Thesis

Degree Name

Master of Science (MS)


Department of Atmospheric and Environmental Sciences

Content Description

1 online resource (ii, iv, 46 pages) : illustrations (some color), color maps.

Dissertation/Thesis Chair

Aiguo Dai

Committee Members

Brian Rose


CanESM2, CESM1, climate, ensemble, trends, variability, Climatic changes, Climatology

Subject Categories

Atmospheric Sciences


Internal climate variability (ICV) can influence trends in observations and model simulations over periods spanning decades, making it difficult to quantify the response of the climate system to external forcing. We analyze two large ensembles of climate simulations from 1950-2100 by two fully coupled climate models, namely the CanESM2 and CESM1, to quantify the effects of ICV on apparent trends in annual surface air temperature (T) and precipitation (P) over different time periods and regions. In comparison with observations, the CanESM2 overestimates the global T and P trends during 1979-2014 while they fall within the ICV-induced range of the CESM1. The ICV induces large trend spreads for global-mean T and P over shorter periods, but these spreads decrease rapidly as the length of the period is extended through the historical period; thereafter, the spreads decrease gradually. In individual realizations, local P trends may still be dominated by ICV during the first half of the 21st century over many regions, while the externally-forced T trends are already significant (compared with ICV) over many low-latitude regions and will become statistically significant everywhere before the mid-21st century. The largest contributors to ICV for T and P are the Interdecadal Pacific Oscillation (IPO) and Arctic sea-ice variability. Furthermore, the evolution of the T response to external forcing is found to be similar everywhere, while the evolution of the P response can be similar (opposite) to the T response in the high latitudes and tropics (subtropics). The ICV in a control run is found to be similar to that from the forced runs for T, but the internal P variability in the forced runs is larger than that in the control run.