Other reports in this collection
This subsection sketches the most evident singularities discussed in the context of climate change and reviews the pertinent literature on their potential impacts.
That the occurrence of weather events is essentially stochastic is a well-established fact (e.g., Lorenz, 1982; Somerville, 1987). Most climatic impacts arise from extreme weather events or from climatic variables exceeding some critical level and thereby affecting the performance or behavior of a biological or physical system (e.g., Downing et al., 1999). The same holds for the impacts of climate change (see Chapters 1, 2, and 3, especially Table 3-9; Pittock and Jones, 2000).
For many important climate impacts, we are interested in the effects of specific extreme events or threshold magnitudes that have design or performance implications. To help in zoning and locating developments or in developing design criteria for the capacities of spillways and drainage structures, the heights of levee banks, and/or the strengths of buildings, for example, planners and engineers routinely use estimated "return periods" (the average time between events) at particular locations for events of particular magnitudes. Such event magnitudes include flood levels (Hansen, 1987; Handmer et al., 1999) and storm-surge heights (Middleton and Thompson, 1986; Hubbert and McInnes, 1999). Return period estimates normally are based on recent instrumental records, sometimes augmented by estimates from other locations, or statistical or physically based modeling (Middleton and Thompson, 1986; Hansen, 1987; Beer et al., 1993; National Research Council, 1994; Pearce and Kennedy, 1994; Zhao et al., 1997; Abbs, 1999). The assumption usually is made that these statistics, based on past events, are applicable to the future—but climate change means that this often will not be the case.
Thus, a central problem in planning for or adapting to climate change and estimating the impacts of climate change is how these statistics of extreme events are likely to change. Similar problems arise in nonengineering applications such as assessing the economic performance or viability of particular enterprises that are affected by weather—for example, farming (Hall et al., 1998; Kenny et al., 1999; Jones, 2000)—or health effects (Patz et al., 1998; McMichael and Kovats, 2000; see also Chapter 9).
Relatively rapid changes in the magnitude and frequency of specified extreme events arise because extremes lie in the low-frequency tails of frequency distributions, which change rapidly with shifts in the means. Moreover, there also can be changes in the shape of frequency distributions, which may add to or subtract from the rate of change of extremes in particular circumstances (Mearns et al., 1984; Wigley, 1985, 1988; Hennessy and Pittock, 1995; Schreider et al., 1997). Such changes in the shape of frequency distributions require special attention. Evidence suggests that they are particularly important for changes in extreme rainfall (Fowler and Hennessy, 1995; Gregory and Mitchell, 1995; Walsh and Pittock, 1998), possibly in the intensities of tropical cyclones (Knutson et al., 1998; Walsh and Ryan, 2000), and in ENSO behavior (Dilley and Heyman, 1995; Bouma et al., 1997; Bouma, 1999; Timmermann et al., 1999; Fedorov and Philander, 2000). Return periods can shorten, however, even if none of these higher moment effects emerge; simply moving mean precipitation higher, for example, could make the 100-year flood a 25-year flood.
It is noteworthy that the central role in impact assessments of the occurrence of extreme weather events gives rise to multiple sources of uncertainty in relation to climate change. The stochastic nature of the occurrence of extremes and the limited historical record on which to base the frequency distribution for such events give rise, even in a stationary climate, to a major uncertainty. Beyond that, any estimate of a change in the frequency distribution under a changing climate introduces new uncertainties. Additional uncertainties relate to our limited understanding of the impacted systems and their relevant thresholds, as well as the possible effects of adaptation, or societal change, in changing these thresholds. If this were not complicated enough, many impacts of weather extremes arise from sequences of extremes of the same or opposite sign—such as sequences of droughts and floods affecting agriculture, settlements, pests, and pathogens (e.g., Epstein, 2000) or multiple droughts affecting the economic viability of farmers (e.g., Voortman, 1998).
Planned adaptation to climate change therefore faces particular difficulty in this environment because projections of changes in the frequency of extreme events and threshold exceedence require a multi-decadal to century-long projected (or "recent" observed) data series, or multiple ensemble predictions (which is one way of generating improved statistics). Thus, it is difficult to base planned adaptation on the record of the recent past, even if there is evidence of a climate change trend in the average data. Planned adaptation therefore must rely on model predictions of changes in the occurrence of extreme and threshold events (e.g., see Pittock et al., 1999), with all their attendant uncertainties. Real-life adaptation therefore will most likely be less optimal (more costly or less effective) than if more precise information on future changes in such thresholds and extremes were available.
Nonetheless, planned adaptation will most likely proceed in response to changes in the perceived relative frequency of extreme events. Properly done, it can have immediate benefit by reducing vulnerability to current climate as well as future benefit in reducing exposure to future climate change. As suggested above, however, there are many ways to respond inappropriately if care is not taken. In short, changes in extremes and in the frequency of exceeding impacts thresholds are a vital feature of vulnerability to climate change that is likely to increase rapidly in importance because the frequency and magnitude of such events will increase as global mean temperature rises.
|
Other reports in this collection |
|