IPCC Report.pdf - Adam Curry

Chapter 3Changes in Climate Extremes and their Impacts on the Natural Physical EnvironmentRecent publications have used other approaches for evaluating characteristics of extremes or changes in extremes, for instance, analyzingtrends in record events or investigating whether records in observed time series are being set more or less frequently than would beexpected in an unperturbed climate (Benestad, 2003, 2006; Zorita et al., 2008; Meehl et al., 2009c; Trewin and Vermont, 2010).Furthermore, besides the actual magnitude of extremes (quantified in terms of probability/return frequency or absolute threshold), otherrelevant aspects for the definition of climate extremes from an impact perspective include the event’s duration, the spatial area affected,timing, frequency, onset date, continuity (i.e., whether there are ‘breaks’ within a spell), and preconditioning (e.g., rapid transition from aslowly developing meteorological drought into an agricultural drought, see Box 3-3). These aspects, together with seasonal variations inclimate extremes, are not as frequently examined in climate models or observational analyses, and thus can only be partly assessedwithin this chapter.As noted in the discussion of ‘extreme weather or climate events’ in Section 3.1.2, thresholds, percentiles, or return values used for thedefinition of climate extremes are generally defined with respect to a given reference period (generally historical, i.e., 1961-1990, butpossibly also based on climate model data). In some cases, a transient baseline can also be considered (i.e., the baseline uses data fromthe period under examination and changes as the period being considered changes, rather than using a standard period such as1961-1990). The choice of the reference period may be relevant for the magnitude of the assessed changes as highlighted, for example,in Lorenz et al. (2010). The choice of the reference period (static or transient) could also affect the assessment of the respective role ofchanges in mean versus changes in variability for changes in extremes discussed in Section 3.1.6. If extremes are based on the probabilitydistribution from which they are drawn, then a simple change in the mean (and keeping the same distribution) would, strictly speaking,produce no relative change in extremes at all. The question of the choice of an appropriate reference period is tied to the notion ofadaptation. Events that are considered extreme nowadays in some regions could possibly be adapted to if the vulnerability and exposureto these extremes is reduced (Chapters 1, 2, and 4 through 7). However, there are also some limits to adaptation as highlighted inChapter 8. These considerations are difficult to include in the statistical analyses of climate scenarios because of the number of (mostlynon-physical) aspects that would need to be taken into account.To conclude, there is no precise definition of an extreme (e.g., D.B. Stephenson et al., 2008). In particular, we note limitations in thedefinition of both probability-based or threshold-based climate extremes and their relations to impacts, which apply independently ofthe chosen method of analysis:• An event from the extreme tails of probability distributions is not necessarily extreme in terms of impact.• Impact-related thresholds can vary in space and time, that is, single absolute thresholds (e.g., a daily rainfall exceeding 25 mm orthe number of frost days) will not reflect extremes in all locations and time periods (e.g., season, decade).As an illustration, projected patterns (in the magnitude but not the sign) of changes in annual heat wave length were shown to be highlydependent on the choice of index used for the assessment of heat wave or warm spell duration (using the mean and maximum HeatWave Duration Indices, HWDImean and HWDImax, and the Warm Spell Duration Index, WSDI; see Orlowsky and Seneviratne, 2011),because of large geographical variations in the variability of daily temperature (Alexander et al., 2006). Similar definition issues apply toother types of extremes, especially those characterizing dryness (see Section 3.5.1 and Box 3-3).• Not all extreme weather and climate events necessarily haveextreme impacts.• The distinction between extreme weather events and extreme climateevents is not precise, but is related to their specific time scales:– An extreme weather event is typically associated with changingweather patterns, that is, within time frames of less than a dayto a few weeks.– An extreme climate event happens on longer time scales. It canbe the accumulation of several (extreme or non-extreme)weather events (e.g., the accumulation of moderately belowaveragerainy days over a season leading to substantially belowaveragecumulated rainfall and drought conditions).For simplicity, we collectively refer to both extreme weather events andextreme climate events with the term ‘climate extremes’ in this chapter.From this definition, it can be seen that climate extremes can be definedquantitatively in two ways:1) Related to their probability of occurrence2) Related to a specific (possibly impact-related) threshold.The first type of definition can either be expressed with respect to givenpercentiles of the distribution functions of the variables, or with respectto specific return frequencies (e.g., ‘100-year event’). Compound eventscan be viewed as a special category of climate extremes, which resultfrom the combination of two or more events, and which are again‘extreme’ either from a statistical perspective (tails of distribution functionsof climate variables) or associated with a specific threshold (Section3.1.3.). These two definitions of climate extremes, probability-based orthreshold-based, are not necessarily antithetic. Indeed, hazards forsociety and ecosystems are often extreme both from a probability and117