Summary and Conclusions

Observational data in combination with theoretical studies and modeling experiments suggest that the surface temperature of the Northern Hemisphere during the past 1000 years could be characterized as follows: First, a slow ongoing cooling is by and large consistent with orbital forcing or the so-called Milankovitch effect. This amounts to an annual cooling of some 0.2 K over 900 years. Second, a rapid temperature increase starting at the beginning of the 20tli century remained more or less unchanged through the whole century. This temperature increase, according to the interpretation given here, must have been of anthropogenic origin. Superimposed on this overall pattern are substantial fluctuations from annual to multidecadal time scales, which in all likelihood are caused by internal, mainly stochastic processes in the climate system. The effects of major volcanic eruptions also contribute to such fluctuations and could have caused notable cooling of the climate system over several years. Low-frequency variations in solar irradiation, if they do exist, can similarly give rise to temperature fluctuations over longer time scales.

However, based on available data and model studies, it appears that we can rule out the possibility that the unparalleled warming which took place in the 20th century was a consequence of any of the natural processes as we know them, since both the amplitude and the time period of sustained warming are too large to be reproduced by any climate model. We also can find no support in the observational records from the last 1000 years that such a massive warming has occurred.

That the warming during the last century is of anthropogenic origin is further supported by model simulation studies. Coupled transient model experiments using observed data for greenhouse gases and anthropogenic aerosols can at least reproduce the long-term trend in the observed warming pattern during the 20th century. Possible trends in other parameters, such as precipitation, are not yet statistically significant but rather are changes in characteristic weather patterns.

Even if available climate models agree in simulating a warming, there are considerable model differences, particularly in the pattern and speed of warming. Results obtained so far are strongly model-dependent, suggesting the importance of dynamical and physical feedbacks in determining regional changes in surface temperature, precipitation, and weather pattern. The result of a climate model calculation is currently determined more by the model than by the details of the particular forcing being used. This strongly suggests that climate models have to be realistic and rather detailed, since any systematic model deficiency could create an erroneous response pattern. At least a necessary condition must be that climate models be able to realistically reproduce the present climate and its characteristic variations. Simple models could in this context be quite misleading.

Present models still have major deficiencies due to insufficient horizontal and vertical resolution, which leads to difficulties in representing orography and coastlines as well as limitations in reproducing realistic weather patterns. This affects not only the ability to simulate regional climate, but also to some extent to correctly maintain the large-scale atmospheric and ocean circulation, since the large-scale circulation in turn is partly driven dynamically by smaller weather systems.

Another major problem concerns the representation of physical processes in large-scale models. Radiation and clouds, deep convection, and near surface- and free atmospheric turbulence are examples of atmospheric processes that are extremely difficult to handle, partly due to the lack of suitable observational data as well as to lack of proper understanding of complex atmospheric processes. Similar difficulties occur in ocean and land surface modeling, where the processes regulating the exchange of heat, water, and momentum on the scale of climate models are not completely known.

The coupling between the atmosphere and the ocean is a particular problem. Minor changes in cloud cover and sea ice distribution drastically influence the exchange of heat and water between the atmosphere and the ocean. Small systematic errors in the atmospheric and ocean model components can then generate an erroneous climate drift in a long integration over several centuries. Many models handle this by introducing a small systematic correction of the surface fluxes to ensure that no systematic errors will occur in the equilibrium case when no climate change occurs. Climate change model studies are therefore in essence perturbation studies and are likely to become misleading when the perturbations become too large. For the global change studies assuming a doubling or a trebling of the greenhouse gas concentration, this does not appear to be a serious restriction.

The so-called flux adjustment (Sausen et al, 1988) has been criticized and used as an example that coupled models making use of this assumption are less credible. Recently, there have been a few integrations where the flux adjustment has been significantly relaxed (e.g., only using annual means; Roeckner ct al, 1999) or even fully eliminated (Mitchell, personal communication). However, it appears that whether or not a flux adjustment was used has no apparent effect on the overall result. It is nevertheless to be expected that the new generation of coupled models are able to reduce systematic error to such a low level that flux adjustment or any other empirical correction is no longer required.

An outstanding issue, which finally must be stressed, is the inherent stochastic variability in models and as I believe also in nature itself. This means that just by chance we can have climate anomalies lasting for several decades influencing regional climate in a significant way. Such anomalies are often mistakenly taken for a genuine climate change as happened after the warm period in the 1930s and again after the cold periods in the 1960s and 1970s when even some climate scientists seriously suggested the possibility of a persistent change toward a much colder climate.

However, we still do not know whether characteristic climate anomalies like ENSO and NAO will systematically change in a warmer climate. This cannot be ruled out at present, and in fact some models indicate that the present probability distribution of both ENSO and NAO may be different (Timmermann et al, 1998). Such a change may have more serious consequences for the regional climate than an overall superimposed warming.

Finally, let me return to the question of the overall dynamical stability of the earth's climate, that is, whether climate is transient or intransient. The most likely event creating a switch to another regime of climate, although essentially only of regional influence, could be caused by a reduction or even a halt in the thermohaline circulation of the North Atlantic, leading to a situation with reduced sea surface temperatures in that region (Marotzke and Willebrand, 1991; Rahmstorf, 1996). This is currently an issue of considerable interest and concern, since models have indicated that such an instability could be initiated (e.g., Manabe and Stouffer, 1988) by increased precipitation in the North Atlantic storm track or by increased melting of glaciers in southern Greenland. Whether such event may take place in reality or not is still an open question and more advanced modeling studies are urgently required here.

0 0

Post a comment