The issues involved in measuring and assessing changes in extremes have recently been comprehensively reviewed by Trenberth and Owen (1999), Nicholls and Murray (1999), and Folland et al. (1999b). Despite some progress described below, there remains a lack of accessible daily climate data sets which can be intercompared over large regions (Folland et al., 2000). Extremes are a key aspect of climate change. Changes in the frequency of many extremes (increases or decreases) can be surprisingly large for seemingly modest mean changes in climate (Katz, 1999) and are often the most sensitive aspects of climate change for ecosystem and societal responses. Moreover, changes in extremes are often most sensitive to inhomogeneous climate monitoring practices, making assessment of change more difficult than assessing the change in the mean.
Given the number of ways in which extreme climate events and variability about the mean can be defined, (e.g., extreme daily temperatures, large areas experiencing unusual temperatures, severity of heat waves, number of frosts or freezes, changes in interannual variability of large area temperatures, etc.) extreme care must be exercised in generalising results. Here we assess the evidence for changes in temperature extremes or variability, first based on global analyses and then on more detailed regional analyses.
Parker et al. (1994) compared the interannual variability of seasonal temperature anomalies from the 1954 to 1973 period to the 1974 to 1993 period for most of the globe. They found a small increase in variability overall with an especially large increase in central North America. By restricting the analyses to the latter half of the 20th century, Parker et al. (1994) minimised the potential biases due to an increasing number of observations in this period. Several other studies found a reduction in other aspects of variability over longer time periods. Jones (1999) also analysed global data and found no change in variability, but since 1951 the rise in global mean temperatures can be attributed to an increase (decrease) in areas with much above (below) normal temperatures. They also analysed the change in the aggregated total of much below and much above normal temperatures (upper and lower ten percentiles). They found little overall change, except for a reduced number of much above or below normal temperatures during the 1960s and 1970s. Michaels et al. (1998) examined 5° latitude x 5° longitude monthly temperature anomalies for many grid cells around the world and found an overall decrease in intra-annual variance over the past 50 to 100 years. They also examined the daily maximum and minimum temperatures from the United States, China, and the former Soviet Union and found a general decline in the intra-monthly temperature variability. As reported in the SAR, a related analysis by Karl et al. (1995b) found reduced day-to-day variability during the 20th century in the Northern Hemisphere, particularly in the United States and China. Recently, Collins et al. (2000) has identified similar trends in Australia. By analysing a long homogenised daily temperature index for four stations in Northern Europe, Moberg et al. (2000) also found a progressive reduction in all-seasons inter-daily variability of about 7% between 1880 and 1998. Balling (1998) found an overall decrease in the spatial variance of both satellite-based lower-tropospheric measurements from 1979 to 1996 and in near-surface air temperatures from 1897 to 1996.
Consequently, there is now little evidence to suggest that the interannual variability of global temperatures has increased over the past few decades, but there is some evidence to suggest that the variability of intra-annual temperatures has actually quite widely decreased. Several analyses find a decrease in spatial and temporal variability of temperatures on these shorter time-scales.
Figure 2.33: Changes in the number of frost days (a, b) and in heat-wave duration (c, d) from Frich et al. (2001). Panel (a) shows the percent changes in the total number of days with a minimum temperature of less than 0°C between the first and last half of the period, approximately 1946 to 1999. The red circles indicate negative changes and the blue circles indicate positive changes. Panel (c) shows percentage changes in the maximum number of consecutive days (for periods with >5 such days) with maximum temperatures >5°C above the 1961 to 1990 daily normal. The changes are for the first and second half of the period, approximately 1946 to 1999. The red circles indicate positive changes and the blue circles indicate negative changes. In both panels, the size of each circle reflects the size of the change and solid circles represent statistically significant changes. Panels (b) and (d) show the average annual values of these quantities expressed as percentage differences from their 1961 to 1990 average values. The trend shown in panel (b) is statistically significant at the 5% level.
There have been a number of new regional studies related to changes in extreme temperature events during the 20th century. Gruza et al. (1999) found statistically significant increases in the number of days with extreme high temperatures across Russia using data back to 1961 and on a monthly basis back to 1900. Frich et al. (2001) analysed data spanning the last half of the 20th century across most of the Northern Hemisphere mid- and high latitudes and found a statistically significant increase (5 to >15%) in the growing season length in many regions. Heino et al. (1999) also found that there has been a reduction in the number of days with frost (the number of days with minimum temperature 0°C) in northern and central Europe. Thus, some stations now have as many as 50 fewer days of frost per year compared with earlier in the 20th century. Easterling et al. (2000) found there has been a significant decrease in the number of days below freezing over the central United States (about seven per year). For Canada, Bonsal et al. (2001) also found fewer days with extreme low temperatures during winter, spring and summer, and more days with extreme high temperatures during winter and spring. This has led to a significant increase in the frost-free period. Decreasing numbers of days with freezing temperatures have been found in Australia and New Zealand over recent decades (Plummer et al., 1999; Collins et al., 2000). In addition, while increases in the frequency of warm days have been observed, decreases in the number of cool nights have been stronger. Frich et al. (2001) show a reduced number of days with frost across much of the globe (Figure 2.33) while Michaels et al. (2000) find that much of the warming during the 20th century has been during the cold season in the mid- to high latitudes, consistent with the reduction of extremely low temperatures. Frich et al. (2001) have also found a statistically significant reduction in the difference between the annual extremes of daily maximum and minimum temperatures during the latter half of the 20th century. In China, strong increases in the absolute minimum temperature have been observed, with decreases in the 1-day seasonal extreme maximum temperature (Zhai et al., 1999a) since the 1950s. Wang and Gaffen (2001), however, for a similar period, found an increase in "hot" days in China. Hot days were defined as those days above the 85th percentile during July and August based on an "apparent temperature" index related to human discomfort in China (Steadman, 1984). The number of extremely cold days has also been shown to be decreasing in China (Zhai et al., 1999a). Manton et al. (2001) found significant increases in hot days and warm nights, and decreases in cool days and cold nights since 1961 across the Southeast Asia and South Pacific Region. Jones et al. (1999c) have analysed the 230-year-long daily central England data set that has been adjusted for observing inhomogeneities. They found that the increase in temperature observed in central England corresponds mainly to a reduction in the frequency of much below normal daily temperatures. An increase of the frequency of much above normal temperatures was less apparent.
Analyses of 20th century trends in the United States of short-duration episodes (a few days) of extreme hot or cold weather did not show any significant changes in frequency or intensity (Kunkel et al., 1996, 1999; Karl and Knight, 1997). For Australia, Collins et al. (2000) found higher frequencies of multi-day warm nights and days, and decreases in the frequency of cool days and nights. In an extensive assessment of the change in frequency of heat waves during the latter half of the 20th century, Frich et al. (2001) find some evidence for an increase in heat-wave frequency, but several regions have opposite trends (Figure 2.33c). The extreme heat in the United States during several years in the 1930s dominates the time-series of heat waves in that region. On the other hand, trends in the frequency of extreme apparent temperatures are significantly larger for 1949 to 1995 during summer over most of the USA (Gaffen and Ross, 1998). Warm humid nights more than doubled in number over 1949 to 1995 at some locations. Trends in nocturnal apparent temperature in the USA, however, are likely to be associated, in part, with increased urbanisation. Nevertheless, using methods and data sets to minimise urban heat island effects and instrument changes, Easterling et al. (2000) arrived at conclusions similar to those of Gaffen and Ross (1998).
During the 1997/98 El Niño event, global temperature records were broken for sixteen consecutive months from May 1997 through to August 1998. Karl et al. (2000) describe this as an unusual event and such a monthly sequence is unprecedented in the observational record. More recently, Wigley (2000) argues that if it were not for the eruption of Mt. Pinatubo, an approximately equal number of record-breaking temperatures would have been set during the El Niño of 1990/91. As temperatures continue to warm, more events like these are likely, especially when enhanced by other factors, such as El Niño.
Other reports in this collection