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

Continued from previous page

Over global land, a further warming of surface air temperature has occurred since the SAR. The Peterson and Vose (1997) NCDC (National Climate Data Center) series gives distinctly more warming than does the CRU series since the mid-1980s. The former series is a straightforward average of local land areas, weighted according to their size, whereas the CRU series is a simple average of the two hemispheres which gives more weight to the relatively small area of the Southern Hemisphere land. Because the Northern Hemisphere land has warmed considerably faster than the Southern Hemisphere land since the mid-1980s (reflected in Table 2.1), the simple average results in less warming. The Hansen et al. (1999) GISS (Goddard Institute for Space Studies) series has recently been revised and shows a little less warming than the CRU series since the late 1980s. One reason for this behaviour lies in the way that the Hansen series is constructed. Among other differences, this series gives much more weight to oceanic islands and Antarctica. Because the oceans and Antarctica have warmed less than the rest of the global land in the last fifteen years (see below), the Hansen series can be expected to show less warming. Some of these considerations apply to the Vinnikov et al. (1990) SHI (State Hydrological Institute) series, though this excludes areas south of 60°S.

Table 2.1: Restricted maximum likelihood linear trends in annual average land-surface air temperature (LSAT) anomalies from CRU and sea surface temperature (SST) and night marine air temperature (NMAT) anomalies from the UK Met Office (UKMO). Twice the standard errors of the trends are shown in brackets. Trends significant at the 5% level or better, according to calculations made using an appropriate form of the t test, are shown in bold type. The significances of the trends are indicated beneath their twice standard errors. The method for calculating the trends, standard errors and significances allows for serial correlation and can result in a trend for the globe that is not exactly equal to the average of the trends for the hemispheres, consistent with uncertainties in the trends. The estimates of trends and errors for the land data account for uncertainties in the annual anomalies due to data gaps and urbanisation. Uncertainties in annual marine anomalies are not available. Trends are given in C/decade.

 
1861 to 2000
1901 to 2000
1910 to 1945
1946 to 1975
1976 to 200

Northern Hemisphere
CRU LSAT
(Jones et al., 2001)

0.06
(0.02)
1%
0.07
(0.03)
1%
0.14
(0.05)
1%
-0.04
(0-06)
0.31
(0.11)
1%
Southern Hemisphere
CRU LSAT
(Jones et al., 2001)
0.03
(0.01)
1%
0.05
(0.01)
1%
0.08
(0.04)
1%
0.02
(0.05)
0.13
(0.08)
1%

Global
CRU LSAT
(Jones et al., 2001)

0.05
(0.02)
1%
0.06
(0.02)
1%
0.11
(0.03)
1%
-0.01
(0.05)
0.22
(0.08)
1%
 
Northern Hemisphere
UKMO SST
(Jones et al., 2001)
0.03
(0.01)
1%
0.05
(0.02)
1%
0.15
(0.04)
1%
-0.05
(0.10)
0.18
(0.05)
1%
Southern Hemisphere
UKMO SST
(Jones et al., 2001)
0.04
(0.01)
1%
0.06
(0.01)
1%
0.13
(0.05)
1%
0.06
(0.07)
0.10
(0.05)
1%
Global
UKMO SST
(Jones et al., 2001)
0.04
(0.01)
1%
0.06
(0.01)
1%
0.15
(0.04)
1%
0.01
(0.06)
0.14
(0.04)
1%

Global
UKMO NMAT
(Parker et al., 1995)

 
0.05
(0.02)
1%
0.14
(0.04)
1%
-0.01
(0.06)
0.11
(0.05)
1%

A new record was set in all four series in 1998 (anomalies relative to 1961 to 1990 of CRU, 0.68°C; NCDC, 0.87°C; GISS, 0.58°C; and SHI, 0.58°C). 1998 was influenced by the strong 1997/98 El Niño; the warming influence of El Niño on global temperature is empirically well attested (e.g., Jones, 1994) and the physical causes are starting to be uncovered (Meehl et al., 1998). However, 1998 was considerably warmer than 1983, a year warmed by the comparable 1982/83 El Niño. In fact 1998 was between 0.34 and 0.54°C warmer than 1983 over land, depending on the temperature series used, though there was some offsetting cooling from volcanic aerosols from the 1982 El Chichon eruption in 1983. 1999 was globally much cooler than 1998, with an anomaly of 0.40°C in the CRU series, as it was cooled by the strongest La Niña since 1988/89. Despite its relative coolness, 1999 was still the fifth warmest year in the CRU record. Depending on the record used, 1999 was between 0.11°C and 0.33°C warmer than the last comparable La Niña year, 1989. It is noteworthy, however, that north of 20°N, 1999 was nearly as warm as 1998. Mitigation of the warming trend in the early 1990s was short-lived and was mainly due to the cooling influence of the eruption of Mount Pinatubo in 1991 (Parker et al., 1996), highlighted in the SAR. The ten warmest years in all four records have occurred after 1980, six or seven of them in the 1990s in each series.

Based on the CRU series, equivalent linear trends in global, Northern and Southern Hemisphere land-surface air temperature are shown in Table 2.1. Because warming may not persist at the rates shown, all trends are shown in °C/decade. The two main periods of warming in all three series are between about 1910 to 1945 and between 1976 to 2000 (updated from Karl et al., 2000). Trends have been calculated using a restricted maximum likelihood method (Diggle et al., 1999) that allows for serial correlation in the data. It gives larger standard errors than ordinary least squares methods when data have a complex temporal structure, as is true here. Table 2.1 and Figure 2.1 show that the rate of global and hemispheric warming in land-surface air temperature from 1976 to 2000 was about twice as fast (but interannually more variable) than that for the period 1910 to 1945. However, trends over such short periods are very susceptible to end effects so the values in Table 2.1, and Table 2.2 below, should be viewed with caution for these periods. Both periods of warming are statistically significant, as is (easily) the warming since 1861 or 1901. Uncertainties in the annual values due to data gaps, including an additional estimate of uncertainties due to urbanisation, are included for land-surface air temperature but equivalent uncertainties are not currently available for the marine data alone. Thus estimates in Table 2.1 for the marine data may be conservative, though the effect of adding the influence of annual uncertainties to the land-surface air temperature data trends was small. The period 1946 to 1975 had no significant change of temperature, though there was a small non-significant, but regionally more marked, cooling over the Northern Hemisphere, as discussed by Parker et al. (1994).

The equivalent linear changes in global average CRU land-surface air temperature over 1861 to 2000 and 1901 to 2000 that take into account annual sampling errors and uncertainties due to urbanisation are 0.63 ± 0.24°C and 0.61 ± 0.18°C respectively. Corresponding Northern and Southern Hemisphere changes for 1901 to 2000 are 0.71 ± 0.31°C and 0.52 ± 0.13°C, respectively. Marine surface temperatures are discussed further in Section 2.2.2.2.

Table 2.2: As Table 2.1 but for annual optimally averaged combined CRU land-surface air temperature anomalies and UKMO sea surface temperature anomalies (CRU LSAT + UKMO SST). All of the estimates of trends and errors in the table account for uncertainties in the annual anomalies due to data gaps, urbanisation over land, and bias corrections to SST.
 
1861 to 2000
1901 to 2000
1910 to 1945
1946 to 1975
1976 to 2000
Northern Hemisphere
CRU LSAT + UKMOSST
(Folland et al., 2001)
0.05
(0.02)
1%
0.06
(0.02)
1%
0.17
(0.03)
1%
-005
(0.05)
0.24
(0.07)
1%
Southern Hemisphere
CRU LSAT + UKMOSST
(Folland et al., 2001)
0.04
(0.01)
1%
0.05
(0.02)
1%
0.09
(0.05)
1%
0.03
(0.07)
0.11
(0.05)
1%
Global
CRU LSAT + UKMOSST
(Folland et al., 2001)
0.04
(0.01)
1%
0.06
(0.02)
1%
0.14
(0.04)
1%
-0.01
(0.04)
0.17
(0.05)
1%

Maximum and minimum temperature
As reported in the SAR, and updated by Easterling et al. (1997), the increase in temperature in recent decades has involved a faster rise in daily minimum than daily maximum temperature in many continental regions. This gives a decrease in the diurnal temperature range (DTR) in many parts of the world. The analysis by Easterling et al. (1997) increased total global coverage from 37% to 54% of global land area. Large parts of the world have still not been analysed due to a lack of observations or inaccessible data, particularly in the tropics. Updating all the data remains a problem, so the analysis ends in 1993.

The overall global trend for the maximum temperature during 1950 to 1993 is approximately 0.1°C/decade and the trend for the minimum temperature is about 0.2°C/decade. Consequently, the trend in the DTR is about -0.1°C/decade. The rate of temperature increase for both maximum and minimum temperature over this period is greater than for the mean temperature over the entire 20th century, reflecting the strong warming in recent decades. Note that these trends for 1950 to 1993 will differ from the global trends due to the restricted data coverage so we only quote trends to 0.1°C.

Continues on next page



Other reports in this collection