0.025°C, respectively. Note that neither data set has a global-mean value of precisely zero over its reference period. Primarily, this is because the reference period applies to the individual grid points and because there are coverage changes over the reference periods. In the Jones data set, the land and marine components have different reference periods. For compatibility in merging the data sets, the land data have to have their reference period adjusted from 1951–1970 to 1950–1979, an adjustment that can only be made approximately. This adjustment is not a problem with the IPCC data because both land and marine components use the same reference period (1961–1990). For land station data to be included in the analysis, however, full reference-period coverage was not required; only a minimum of 20 out of 30 years was needed. This can lead to biases in the anomalies relative to the reference period, which are reflected in the spatial mean anomaly averaged over the reference period. In Fig. 1, both data sets have been adjusted to have zero means over 1961– 1990 by subtraction of 0.025°C from the IPCC data and 0.083°C from the Jones data.
The second reason for differences (raw marine data differences) has not been specifically quantified, but it is likely to be relatively small. The third reason (different correction methods) is more important. Below, the combined influences of these two effects are calculated by differencing. To do this, we assume the effect of land data differences (the fifth reason) is negligible, calculate the effect of the fourth reason independently, and subtract this from the difference between the data sets after the effect of the first reason has been removed (see above and Fig. 1).
The fourth reason (different hemispheric averaging methods) has not previously been discussed and is quantified here for the first time (to our knowledge). The two different methods are as follows. In calculating the global mean for the IPCC data set, the data for individual grid boxes are simply area-weighted and averaged. Because the fractional coverage in each hemisphere varies with time, with a relatively greater fraction covered in the Northern Hemisphere in the earlier years, this method may potentially bias the global mean toward the Northern Hemisphere in these years. The Jones global mean is calculated by area-averaging the hemispheres separately first and then averaging the hemispheric means. This method may put undue weight on the sparsely covered Southern Hemisphere in the early years. It is impossible to decide a priori which method is better.
Fig. 2 shows the difference between the IPCC data recalculated using the Jones method and the standard IPCC values (recalculated values minus original values). This isolates the effect of the fourth reason. The differences are small and somewhat erratic, with no overall trend.
Fig. 3 shows the residual difference, Jones data minus recalculated IPCC data, after adjustment of both data sets for the effect of reference-period differences. This plot has been calculated by subtracting the “error” shown in Fig. 2 from the annual data used to produce Fig. 1 This essentially isolates the influences of the different sea surface temperature data sets and the different ways these data sets have been corrected for instrumental biases. The low-frequency changes in this plot arise largely from the different instrumentation correction schemes, whereas the shorter time scale differences mainly reflect differences in the raw data. A clear overall trend (arising mainly over the period 1880–1910) is evident. This trend is reflected in the data differences shown in Fig. 1 and explains why the Jones data have a slightly greater overall warming trend than the IPCC data.
Why has the globe warmed? Because we are confident that human activities have substantially changed the atmospheric composition in terms of greenhouse gases (GHGs; especially carbon dioxide) and aerosols, we are also confident that at least part of the observed warming is human-induced. The leading question is how much? To answer this, we first need to estimate the magnitude of the expected anthropogenic warming. To do this requires a knowledge of the anthropogenic forcing change, and a suitable model to convert this forcing to an estimated climate change.
Fig. 4 shows the current central estimate of forcing changes as used in the latest IPCC calculations of global-mean temperature and sea-level change (ref. 17; further details are given in ref. 18). It is clear that CO2 is the main single factor, but