Working Group I: The Scientific Basis

Other reports in this collection

2.2 How Much is the World Warming? 2.2.1 Background

The SAR concluded that, on a global average, land-surface air and sea surface temperature rose by between 0.3°C and 0.6°C between the late 19th century and 1994. In this section, the recent warming is re-examined, using updated data. We include recent analyses of the diurnal asymmetry of the warming and its geographical structure. Conventional temperature observations are supplemented by indirect evidence and by satellite-based data. For the first time, we make objective estimates of uncertainties in the surface temperature data, though these are preliminary. We also assess recent work in compiling hemispheric and global temperature records from palaeoclimatic data, especially for the most recent millennium.

2.2.2 Temperature in the Instrumental Record for Land and Oceans

Figure 2.1: (a) Annual anomalies of global average land-surface air temperature (°C), 1861 to 2000, relative to 1961 to 1990 values. Bars and solid curve are from CRU (Jones et al., 2001). Values are the simple average of the anomalies for the two hemispheres. The smoothed curve was created using a 21-point binomial filter giving near decadal averages. (b) As (a) but smoothed curves only from NCDC (updated from Peterson and Vose, 1997) � thin solid curve; GISS (adapted from Hansen et al., 1999) � thick dashed curve; SHI (updated from Vinnikov et al., 1990) � thin dashed curve to 1999 only; Peterson and Vose (1997) � thin solid curve. Thick solid curve � as in (a). Two standard error uncertainties are centred on the CRU curve and are estimated using an optimum averaging method (Folland et al., 2001) and include uncertainties due to urbanisation but not due to uncertainties in thermometer exposures. The NCDC curve is the weighted average of the two hemispheres according to the area sampled, which accounts for most of the differences from the CRU curve.

Note that all data sets are adjusted to have zero anomaly when averaged over the period 1961 to 1990. Land-surface air temperature

The SAR reviewed the three databases of land-surface air temperature due to Jones (1994), Hansen and Lebedeff (1988) and Vinnikov et al. (1990). The first and second databases have been updated by Jones et al. (2001) and Hansen et al. (1999), respectively, and a further analysis has become available (Peterson and Vose, 1997; Peterson et al., 1998a, 1999). The last paper also separates rural temperature stations in the Global Historical Climatology Network (GHCN) (Peterson and Vose, 1997) from the full set of stations which, in common with the other three analyses, have been screened for urbanisation effects. While there is little difference in the long-term (1880 to 1998) rural (0.70°C/century) and full set of station temperature trends (actually less at 0.65°C/century), more recent data (1951 to 1989), as cited in Peterson et al. (1999), do suggest a slight divergence in the rural (0.80°C/century) and full set of station trends (0.92°C/century). However, neither pair of differences is statistically significant. In addition, while not reported in Peterson et al., the 1951 to 1989 trend for urban stations alone was 0.10°C/decade. We conclude that estimates of long-term (1880 to 1998) global land-surface air temperature variations and trends are relatively little affected by whether the station distribution typically used by the four global analyses is used, or whether a special effort is made to concentrate on rural stations using elaborate criteria to identify them. Part of the reason for this lack of sensitivity is that the average trends in available worldwide urban stations for 1951 to 1989 are not greatly more than those for all land stations (0.09°C/decade). The differences in trend between rural and all stations are also virtually unaffected by elimination of areas of largest temperature change, like Siberia, because such areas are well represented in both sets of stations.

These results confirm the conclusions of Jones et al. (1990) and Easterling et al. (1997) that urban effects on 20th century globally and hemispherically averaged land air temperature time-series do not exceed about 0.05°C over the period 1900 to 1990 (assumed here to represent one standard error in the assessed non-urban trends). However, greater urbanisation influences in future cannot be discounted. Note that changes in borehole temperatures (Section 2.3.2), the recession of the glaciers (Section, and changes in marine temperature (Section, which are not subject to urbanisation, agree well with the instrumental estimates of surface warming over the last century. Reviews of the homogeneity and construction of current surface air temperature databases appear in Peterson et al. (1998b) and Jones et al. (1999a). The latter shows that global temperature anomalies can be converted into absolute temperature values with only a small extra uncertainty.

Figure 2.1a shows the Jones et al. (2001) CRU (Climatic Research Unit) annual averages, together with an approximately decadally smoothed curve, to highlight decadal and longer changes. This is compared with smoothed curves from the other three analyses in Figure 2.1b. We do not show standard errors for the CRU land data using the Jones et al. (1997b) method as tests suggest that these may not be reliable for land data on its own. Instead we use an optimum averaging method (Folland et al., 2001) where the calculated uncertainties are centred on the simple CRU average. We have added an estimate of the additional, independent, uncertainty (twice the standard error) due to urbanisation increasing from zero in 1900 to 0.12°C in 2000. (The Jones et al. (1990) estimates can be interpreted as one standard error equal to 10% of the global warming to that time of about 0.05°C, see also Box 2.1 on urbanisation.) Note that the warming substantially exceeds the calculated uncertainties. (We have not included the possible refinement of assuming urbanisation uncertainties to apply to the cold side of the trend line only, which would reduce the total uncertainty range in Figure 2.1.)

Box 2.1: Urban Heat Island and the Observed Increases in Land Air Temperature.

There are two primary reasons why urban heat islands have been suspected as being partially responsible for the observed increases in land air temperatures over the last few decades. The first is related to the observed decrease in the diurnal temperature range and the second is related to a lower rate of warming observed over the past twenty years in the lower troposphere compared with the surface.

Since the 1950s both daily maximum and minimum temperatures are available over more than 50% of the global land area. These data indicate that on average the mean minimum temperature has increased at nearly twice the rate of the maximum temperature, reducing the daily temperature range by about 0.4°C over these areas. This has raised questions related to whether the growth of urban heat islands may be responsible for a substantial portion of the observed mean temperature increase, because it is well-known that compared to non-urban areas urban heat islands raise night-time temperatures more than daytime temperatures. Nonetheless, the relatively strong correlation between observed decreases in the daily temperature range with increases of both precipitation (leading to more moist surface conditions) and total cloud amount support the notion that the reduction in diurnal temperature range is in response to these physical changes.

Since 1979 satellite observations and weather balloons (which generally agree well) show substantially less warming of the global lower troposphere (around 2 km) than surface temperatures (0.03 and 0.04°C/decade, respectively, compared to 0.16°C/decade at the surface). However, over the Northern Hemisphere land areas where urban heat islands are most apparent, both the trends of lower-tropospheric temperature and surface air temperature show no significant differences. In fact, the lower-tropospheric temperatures warm at a slightly greater rate over North America (about 0.28°C/decade using satellite data) than do the surface temperatures (0.27°C/decade), although again the difference is not statistically significant. In the global average, the trend differences arise largely from the tropical and sub-tropical oceans. In many such regions, the near-surface marine air temperatures tend to be cool and dense compared with conditions aloft, allowing for the lapse rate with height, disconnecting near-surface (up to about 1 km) conditions from higher layers in the atmosphere. Thus the surface marine layer and the troposphere above can have differing variations and trends.

Clearly, the urban heat island effect is a real climate change in urban areas, but is not representative of larger areas. Extensive tests have shown that the urban heat island effects are no more than about 0.05°C up to 1990 in the global temperature records used in this chapter to depict climate change. Thus we have assumed an uncertainty of zero in global land-surface air temperature in 1900 due to urbanisation, linearly increasing to 0.06°C (two standard deviations 0.12°C) in 2000.

Continues on next page

Other reports in this collection

IPCC Homepage