Climate Change 2001:
Working Group I: The Scientific Basis
Other reports in this collection Decadal and longer time-scale variability

A few studies have attempted to look at model-simulated changes in modes of low-frequency variability due to anthropogenic climate change. Particular attention has focused on changes in ENSO as reported in the SAR and in Section, and the AO or NAO and AAO which are prominent features of low-frequency variability in the Northern and Southern Hemispheres, respectively (e.g., Fyfe et al., 1999; Osborn et al., 1999; Paeth et al., 1999; Shindell et al., 1999; Ulbrich and Christoph, 1999; Zorita and González-Rouco, 2000; Monahan et al., 2000). It should be noted that these studies have used a variety of methods for analysing trends in these modes of variability, including indices based on pressure differentials and principal components (PCs) of hemispheric sea level pressure (SLP). In addition, these indices are sensitive to changes in the SLP patterns with time or forcing and so trends must be treated with some caution. Wallace (2000) finds that in both observations and modelling studies of increased greenhouse gases, the trends are larger in the PC of SLP than in the pressure differential indices. Meehl et al. (2000c) show that the changed base climate state in a future warmer climate could affect the period of global ENSO-like decadal (10 to 20-year period) variability such that there could be a shift to longer periods.

Ulbrich and Christoph (1999) find that the NAO index, based on SLP fluctuations over the North Atlantic in the 300-year control run of their model, shows only a moderate increase over the length of a 240-year scenario run with increasing greenhouse gases. The long-term trend exceeds the variability of the control climate only at the end of the simulation in 2100. In contrast, the steadily growing storm track activity over north-west Europe already surpasses the standard deviation defined from the control run after about 160 years. This effect is associated with a change of the NAO pattern. During the length of the scenario experiment, empirical orthogonal functions for sequential 10-year periods show a systematic north-eastward shift of the NAO's northern variability centre from a position close to the east coast of Greenland, where it is also located in the control run, to the Norwegian Sea (Figure 9.27)

Osborn et al. (1999) show an initial small increase followed by a decrease in the NAO index in one model when forced with increases in greenhouse gases or with greenhouse gases and sulphate aerosols. Paeth et al. (1999) have assessed changes in both the mean and variance of the NAO on decadal time-scales at quadrupled CO2-concentrations using an ensemble of four integrations of a single model. They find a statistically significant increase in the mean NAO index (at the 95% confidence level), especially during late summer/autumn and in winter, suggesting more westerly and typically milder weather over Europe during the cold season. However, the increase in the mean NAO index is accompanied by a reduction in the low-frequency variability of the NAO (Figure 9.28) (significant at the 5% significance level after 1910) suggesting that the NAO stabilises in the positive phase. Shindell et al. (1999) found a trend towards more positive values of the AO index with increased greenhouse gases in a model which included a representation of the stratosphere, but not in troposphere-only versions of the same model. They attribute this to the high correlation of the stratospheric circulation with SLP in the Arctic (e.g., Kitoh et al., 1996; Kodera et al., 1996).

Figure 9.27: Locations of NAO centres (taken to be the position of maximum variance as computed from an EOF analysis of sea level pressure fields) of ECHAM4/OPYC (Ulbrich and Christoph, 1999). The average positions from the entire control run (using winter means) are marked by black squares, those of consecutive decades in the scenario run (using all individual months) are marked by open circles before year 2020 and by black dots thereafter.

Figure 9.28: Decadal probability density functions (PDF) of the ECHAM3/LSG transient greenhouse gas ensemble: each PDF (greyscale bars) consists of 160 NAO index realisations including the monthly means of November to February of the four simulations over one decade. The greyscale background indicates the relative frequency of the classified NAO indices based on a kernel function (Matyasovszky, 1998). The PDF's width indicates the decadal variability. At the bottom, the time-series of the decadal mean (solid green line) and the variability (solid red line) of each PDF as well as the corresponding linear trends (dashed lines) are shown (from Paeth et al., 1999).

In an ensemble of 1900 to 2100 transient integrations with greenhouse gas and aerosol forcing changes, Fyfe et al. (1999) find a positive trend in the mean AO and AAO indices. They argue that in their model this is as a result of essentially unchanged AO/AAO patterns superimposed onto a forced climate change. The result of Fyfe et al. (1999) suggests that since the mean AO/AAO increases, it might imply a change to higher-frequency variability, as the positive AO phase has enhanced westerlies and is typically correlated with above-average storminess. In a subsequent non-linear analysis by Monahan et al. (2000) of a 1,000-year control and 500-year stabilisation integration (with greenhouse gas and aerosol forcing fixed at their year 2100 levels) it is found that (1) in the control integration the AO is part of a more general non-linear mode of tropospheric variability which is strongly bimodal and partitions the variability into two distinct regimes, and (2) in the stabilisation integration the occupancy statistics of these regimes change rather than the modes themselves.

In summary, there is not yet a consistent picture emerging from coupled models as to their ability to reproduce trends in climate regimes such as the recently observed upward trend in the NAO/AO index (Chapters 2 and 12). In addition, whilst several models show an increase in the NAO/AO index with increased greenhouse gases, this is not true for all models, and the magnitude and character of the changes vary across models. Such results do not necessarily suggest that the forced climate change is manifest as a change in the occurrence of only one phase of these modes of variability.

Other reports in this collection