Climate Change 2001:
Working Group I: The Scientific Basis
Other reports in this collection

9.3.6 Changes of Extreme Events

In this section, possible future changes in extreme weather and climate phenomena or events (discussed in Chapter 2) will be assessed from global models. Regional information derived from global models concerning extremes will be discussed in Chapter 10.

Although the global models have improved over time (Chapter 8), they still have limitations that affect the simulation of extreme events in terms of spatial resolution, simulation errors, and parametrizations that must represent processes that cannot yet be included explicitly in the models, particularly dealing with clouds and precipitation (Meehl et al., 2000d). Yet we have confidence in many of the qualitative aspects of the model simulations since they are able to reproduce reasonably well many of the features of the observed climate system not only in terms of means but also of variability associated with extremes (Chapter 8). Simulations of 20th century climate have shown that including known climate forcings (e.g., greenhouse gases, aerosols, solar) leads to improved simulations of the climate conditions we have already observed. Ensembles of climate change experiments are now being performed to enable us to better quantify changes of extremes. Temperature

Figure 9.29: The change in 20-year return values for daily maximum (upper panel) and minimum (lower panel) surface air temperature (or screen temperature) simulated in a global coupled atmosphere-ocean model (CGCM1) in 2080 to 2100 relative to the reference period 1975 to 1995 (from Kharin and Zwiers, 2000). Contour interval is 4°C. Zero line is omitted.

Models described in the IPCC First Assessment Report (Mitchell et al., 1990) showed that a warmer mean temperature increases the probability of extreme warm days and decreases the probability of extreme cold days. This result has appeared consistently in a number of more recent different climate model configurations (Dai et al., 2001; Yonetani and Gordon, 2001). There is also a decrease in diurnal temperature range (DTR) since the night-time temperature minima warm faster than the daytime maxima in many locations (e.g., Dai et al., 2001). Although there is some regional variation as noted in Chapter 10, some of these changes in DTR have also been seen over a number of areas of the world in observations (see Chapter 2). In general, the pattern of change in return values for 20-year extreme temperature events from an equilibrium simulation for doubled CO2 with a global atmospheric model coupled to a non-dynamic slab ocean shows moderate increases over oceans and larger increases over land masses (Zwiers and Kharin, 1998; Figure 9.29). This result from a slab ocean configuration without ocean currents is illustrative and could vary from model to model, though it is similar to results from the fully coupled version in a subsequent study (Kharin and Zwiers, 2000).

The greatest increase in the 20-year return values of daily maximum temperature (Figure 9.29, top) is found in central and southeast North America, central and south-east Asia and tropical Africa, where there is a decrease in soil moisture content. Large extreme temperature increases are also seen over the dry surface of North Africa. In contrast, the west coast of North America is affected by increased precipitation resulting in moister soil and more moderate increases in extreme temperature. There are small areas of decrease in the Labrador Sea and Southern Ocean that are associated with changes in ocean temperature. The changes in the return values of daily minimum temperature (Figure 9.29, bottom) are larger than those of daily maximum temperature over land areas and high latitude oceans where snow and ice retreat. Somewhat larger changes are found over land masses and the Arctic while smaller increases in extreme minimum temperatures occur at the margins of the polar oceans. Thus, there is some asymmetry between the change in the extremes of minimum and maximum temperature (with a bigger increase for minima than maxima). This has to do with the change in the nature of the contact between atmosphere and the surface (e.g., minima increase sharply where ice and snow cover have retreated exposing either ocean or land, maxima increase more where the land surface has dried). Consequently there is a seasonal dependence related to changes in underlying surface conditions, which indroduces uncertainties in some regions in some models (Chapter 10).

Simulations suggest that both the mean and standard deviation of temperature are likely to change with a changed climate, and the relative contribution of the mean and standard deviation changes depends on how much each moment changes. Increased temperature variance adds to the probability of extreme high temperature events over and above what could be expected simply from increases in the mean alone. The increased variance of daily temperature in summer in northern mid-continental areas noted above has also been seen in other global models (Gregory and Mitchell, 1995). However, as noted in Chapter 10, such changes can vary from region to region and model to model (e.g., Buishand and Beersma (1996), who showed some small decreases over an area of Europe). The change in the mean is usually larger than the change in variance for most climate change simulations. Climate models have also projected decreased variability of daily temperature in winter over mid-continental Europe (Gregory and Mitchell, 1995). Such a decrease is partly related to a reduction of cold extremes, which are primarily associated with the increased mean of the daily minimum temperature. The detrimental effect of extreme summer heat is likely to be further exacerbated by increased atmospheric moisture. One model scenario shows an increase of about 5°C in July mean "heat index" (a measure which includes both the effects of temperature and moisture, leading to changes in the heat index which are larger than changes in temperature alone; it measures effects on human comfort; see further discussion in Chapter 10) over the southeastern USA by the year 2050 (Delworth et al., 2000). Changes in the heating and cooling degree days are another likely extreme temperature-related effect of future greenhouse warming. For example, analysis of these measures shows a decrease in heating degree days for Canada and an increase in cooling degree days in the southwest USA in model simulations of future climate with increased greenhouse gases (Zwiers and Kharin, 1998; Kharin and Zwiers, 2000), though this can be considered a general feature associated with an increase in temperature.

Other reports in this collection