Strehz, Alexander (2009) Twentieth century temperature variability: externally forced and internally generated. (Diploma thesis), Christian-Albrechts-Universität, Kiel, Germany, 97 pp.

Text 765_Strehz_2009_TwentiethCenturyTemperatureVariabilityExternally_Hschri_pubid12301.pdf - Reprinted Version Restricted to Registered users only Available under License German copyright act UrhG. Download (3MB) | Contact

Abstract

Major climate change was observed during the 20th century. In particular the rise in the global mean temperature has received a lot of attention. This diploma thesis focuses on the temperature change during the last century and its possible causes. Undoubtedly the major part of the observed increase in global mean temperature is of anthropogenic origin, but as climate is known to vary on all timescales the question arises can part of the warming be attributed to internal climate variability? In this diploma thesis a simple statistical model is used to separate the anthropogenic part of the temperature increase from those attributed to natural causes, such as internal climate variability. For this it is assumed that anthropogenic climate change can be described by a time-invariant pattern whose amplitude is proportional to the logarithm of the lagged carbon dioxide concentration. As the best estimate of the pattern, the first Empirical Orthogonal Function (EOF) of the multimodel mean of the CMIP3 20th century experiment is chosen. Both assumptions are tested using the CMIP3 20th century experiments and they prove to be justified. In fact almost all of the global mean temperature increase in the ultimodel mean can be explained by this pattern and the proposed time evolution. An estimate of the anthropogenic part of the observed temperature increase is obtained by fitting this statistical model to the observations via linear regression. In addition to the anthropogenic signal a low frequency variation in global mean temperature with a period of about 60 years is detected. This result is independent of the pattern used for the fit, the timespan the fit is applied to and the observational dataset used, although the amplitudes of the residual show some dependence on the two latter factors. The linear trend from the Hadcrut3 observational dataset between 1978 and 2007 is estimated to be 0.484K/30 y. The largest part of this trend (82% ± 13%) is attributed to anthropogenic climate changes, leaving 18% of the trend to be caused by natural variability. The result is largely independent of the pattern used. However, applying the fit to a different time span and to a different data set does have a considerable effect on the trends. The effect is that the trend inferred from an observational dataset is at the upper bound of the the error estimate of the trend in the anthropogenic temperature increase. The residual resulting from subtraction of the fit from the observations is analyzed in terms of well known climate indices and by comparing the residual of cold and warm periods. The Atlantic multidecadal variability (AMV), the Pacific decadal oscillation (PDO) and an El-Ni˜no Southern Oscillation index (NINO3) derived from the residual are shown to be in good agreement with the respective indices computed from linearly detrended observations. Especially the North Atlantic shows significant temperature changes between warm and cold periods in the global mean of the residual.

Actions (login required)

View Item