Climate Change 2001:
Working Group II: Impacts, Adaptation and Vulnerability
Other reports in this collection Temporal Resolution (Mean versus Variability)

For the most part, climate changes calculated from climate model experiments have been mean monthly changes in relevant variables. Techniques for generating changes in variability emerged in the 1990s (Mearns et al., 1992, 1996, 1997; Wilks, 1992; Semenov and Barrow, 1997). The most common technique involves manipulation of the parameters of stochastic weather generators to simulate changes in variability on daily to interannual time scales (e.g., Bates et al., 1994, 1996). Several studies have found important differences in the estimated impacts of climate change when effects of variance change were included (Mearns et al., 1997; Semenov and Barrow, 1997). Combined changes in mean and variability also are evident in a broad suite of statistical downscaling methods (Katz and Parlange, 1996; Wilby et al., 1998). Other types of variance change still are difficult to incorporate, such as possible changes in the frequency and intensity of El Niño events (Trenberth and Hoar, 1997). However, where ENSO signals are strong, weather generators can be conditioned on ENSO phases, enabling scenarios of changed ENSO frequency to be generated stochastically (e.g., Woolhiser et al., 1993). However, climate models still are not capable of clearly indicating how ENSO events might change in the future (TAR WGI Chapter 9). Incorporation of Extremes in Scenarios

Whereas changes in both the mean and higher order statistical moments (e.g., variance) of time series of climate variables affect the frequency of extremes based on these variables (e.g., extreme high daily or monthly temperatures; drought and flood episodes), other types of extremes are based on complex atmospheric phenomena (e.g., hurricanes). Given the importance of the more complex extremes—such as hurricanes, tornadoes, and storm surges (see Table 1-1)—it would be desirable to incorporate changes in the frequency of such phenomena into scenarios. Unfortunately, very little work has been performed on how to accomplish this, and there is only limited information on how the frequency, intensity, and spatial characteristics of such phenomena might change in the future (see Section 3.8.5).

An example of an attempt to incorporate such changes into impact assessments is the study of McInnes et al. (2000), who developed an empirical/dynamical model that gives return period versus height for tropical cyclone-related storm surges for a location on the north Australian coast. The model can accept changes in tropical cyclone characteristics that may occur as a result of climate change, such as changes in cyclone intensity. Other methods for incorporating such changes into quantitative climate scenarios remain to be developed; further advances in this area of research can be expected over the next few years. Surprises: Low-Probability, High-Impact Events

Several types of rapid, nonlinear response of the climate system to anthropogenic forcing, sometimes referred to as "surprises," have been suggested. These include reorganization of the thermohaline circulation, rapid deglaciation, and fast changes to the carbon cycle (e.g., Stocker and Schmittner, 1997). For instance, it has been suggested that a sudden collapse of the thermohaline circulation in the North Atlantic—an event that has not been simulated by any AOGCM (TAR WGI Chapter 9) but cannot be ruled out on theoretical grounds (TAR WGI Chapter 7)—could cause major disruptions in regional climate over northwest Europe. Such a possibility has been used to create synthetic arbitrary climate scenarios to investigate possible extreme impacts (Alcamo et al., 1994; Klein Tank and Können, 1997).

Other reports in this collection