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

13.4.2 Temporal Variability

The climate change information most commonly taken from climate modelling experiments comprises mean monthly, seasonal, or annual changes in variables of importance to impact assessments. However, changes in climate will involve changes in variability as well as mean conditions. As mentioned in Section 13.3 on baseline climate, the interannual variability in climate scenarios constructed from mean changes in climate is most commonly inherited from the baseline climate, not from the climate change experiment. Yet, it is known that changes in variability could be very important to most areas of impact assessment (Mearns, 1995; Semenov and Porter, 1995). The most obvious way in which variability changes affect resource systems is through the effect of variability change on the frequency of extreme events. As Katz and Brown (1992) demonstrated, changes in standard deviation have a proportionately greater effect than changes in means on changes in the frequency of extremes. However, from a climate scenario point of view, it is the relative size of the change in the mean versus standard deviation of a variable that determines the final relative contribution of these statistical moments to a change in extremes. The construction of scenarios incorporating extremes is discussed in Section 13.4.2.2.

The conventional method of constructing mean change scenarios for precipitation using the ratio method (discussed in Section 13.3) results in a change in variability of daily precipitation intensity; that is, the variance of the intensity is changed by a factor of the square of the ratio (Mearns et al., 1996). However, the frequency of precipitation is not changed. Using the difference method (as is common for temperature variables) the variance of the time-series is not changed. Hence, from the perspective of variability, application of the difference approach to precipitation produces a more straightforward scenario. However, it can also result in negative values of precipitation. Essentially neither approach is realistic in its effect on the daily characteristics of the time-series. As mean (monthly) precipitation changes, both the daily intensity and frequency are usually affected.



Other reports in this collection