The North Atlantic Oscillation (NAO) is the dominant pattern of northern wintertime atmospheric circulation variability and is increasingly being simulated realistically. The NAO is closely related to the Arctic Oscillation (AO), which has an additional annular component around the Arctic. There is strong evidence that the NAO arises mainly from internal atmospheric processes involving the entire troposphere-stratosphere system. Fluctuations in Atlantic Sea Surface Temperatures (SSTs) are related to the strength of the NAO, and a modest two-way interaction between the NAO and the Atlantic Ocean, leading to decadal variability, is emerging as important in projecting climate change.
Climate change may manifest itself both as shifting means, as well as changing preference of specific climate regimes, as evidenced by the observed trend toward positive values for the last 30 years in the NAO index and the climate “shift” in the tropical Pacific about 1976. While coupled models simulate features of observed natural climate variability, such as the NAO and ENSO, which suggests that many of the relevant processes are included in the models, further progress is needed to depict these natural modes accurately. Moreover, because ENSO and NAO are key determinants of regional climate change and can possibly result in abrupt and counter intuitive changes, there has been an increase in uncertainty in those aspects of climate change that critically depend on regional changes.
The possibility for rapid and irreversible changes in the climate system
exists, but there is a large degree of uncertainty about the mechanisms involved
and hence also about the likelihood or time-scales of such transitions.
The climate system involves many processes and feedbacks that interact in complex
non-linear ways. This interaction can give rise to thresholds in the climate
system that can be crossed if the system is perturbed sufficiently. There is
evidence from polar ice cores suggesting that atmospheric regimes can change
within a few years and that large-scale hemispheric changes can evolve as fast
as a few decades. For example, the possibility of a threshold for a rapid transition
of the Atlantic THC to a collapsed state has been demonstrated with a hierarchy
of models. It is not yet clear what this threshold is and how likely it is that
human activity would lead it to being exceeded (see Section
F.6). Atmospheric circulation can be characterised by different preferred
patterns; e.g., arising from ENSO and the NAO/AO, and changes in their phase
can occur rapidly. Basic theory and models suggest that climate change may be
first expressed in changes in the frequency of occurrence of these patterns.
Changes in vegetation, through either direct anthropogenic deforestation or
those caused by global warming, could occur rapidly and could induce further
climate change. It is supposed that the rapid creation of the Sahara about 5,500
years ago represents an example of such a non-linear change in land cover.
Coarse resolution AOGCMs simulate atmospheric general circulation features well in general. At the regional scale, the models display area-average biases that are highly variable from region to region and among models, with sub-continental area averaged seasonal temperature biases typically ±4ºC and precipitation biases between -40 and +80%. These represent an important improvement compared to AOGCMs evaluated in the SAR.
The development of high resolution/variable resolution Atmospheric General Circulation Models (AGCMs) since the SAR generally shows that the dynamics and large-scale flow in the models improves as resolution increases. In some cases, however, systematic errors are worsened compared to coarser resolution models, although only very few results have been documented.
High resolution RCMs have matured considerably since the SAR. Regional models consistently improve the spatial detail of simulated climate compared to AGCMs. RCMs driven by observed boundary conditions evidence area-averaged temperature biases (regional scales of 105 to 106 km2) generally below 2ºC, while precipitation biases are below 50%. Regionalisation work indicates at finer scales that the changes can be substantially different in magnitude or sign from the large area-average results. A relatively large spread exists among models, although attribution of the cause of these differences is unclear.
Other reports in this collection |