In the previous sections, we have evaluated the different lines of evidence on the causes of recent climate change. Here, we summarise briefly the arguments that lead to our final assessment. The reader is referred to the earlier sections for more detail.

20th century climate was unusual.
Palaeoclimatic reconstructions for the last 1,000 years (e.g., Chapter 2, Figure 2.21) indicate that the 20th century warming is highly unusual, even taking into account the large uncertainties in these reconstructions.

The observed warming is inconsistent with model estimates of natural internal climate variability.
While these estimates vary substantially, on the annual to decadal time-scale they are similar, and in some cases larger, than obtained from observations. Estimates from models and observations are uncertain on the multi-decadal and longer time-scales required for detection. Nonetheless, conclusions on the detection of an anthropogenic signal are insensitive to the model used to estimate internal variability. Recent observed changes cannot be accounted for as pure internal variability even if the amplitude of simulated internal variations is increased by a factor of two or more. It is therefore unlikely (bordering on very unlikely) that natural internal variability alone can explain the changes in global climate over the 20th century (e.g., Figure 12.1).

The observed warming in the latter half of the 20th century appears to be inconsistent with natural external (solar and volcanic) forcing of the climate system.
Although there are measurements of these forcings over the last two decades, estimates prior to that are uncertain, as the volcanic forcing is based on limited measurements, and the solar forcing is based entirely on proxy data. However, the overall trend in natural forcing over the last two, and perhaps four, decades of the 20th century is likely to have been small or negative (Chapter 6, Table 6.13) and so is unlikely to explain the increased rate of global warming since the middle of the 20th century.

The observed change in patterns of atmospheric temperature in the vertical is inconsistent with natural forcing.
The increase in volcanic activity during the past two to four decades would, if anything, produce tropospheric cooling and stratospheric warming, the reverse to what has occurred over this period (e.g., Figure 12.8). Increases in solar irradiance could account for some of the observed tropospheric warming, but mechanisms by which this could cool the stratosphere (e.g., through changes in stratospheric ozone) remain speculative. Observed increases in stratospheric water vapour might also account for some of the observed stratospheric cooling. Estimated changes in solar radiative forcing over the 20th century are substantially smaller than those due to greenhouse gas forcing, unless mechanisms exist which enhance the effects of solar radiation changes at the ground. Palaeo-data show little evidence of such an enhancement at the surface in the past. Simulations based solely on the response to natural forcing (e.g., Figure 12.7a) are inconsistent with the observed climate record even if the model-simulated response is allowed to scale up or down to match the observations. It is therefore unlikely that natural forcing and internal variability together can explain the instrumental temperature record.

Anthropogenic factors do provide an explanation of 20th century temperature change.
All models produce a response pattern to combined greenhouse gas and sulphate aerosol forcing that is detectable in the 20th century surface temperature record (e.g., Figures 12.10, 12.12 (one model produces an estimate of internal variability which is not consistent with that observed)). Given that sulphate aerosol forcing is negative, and hence tends to reduce the response, detection of the response to the combined forcing indicates the presence of a greenhouse gas signal that is at least as large as the combined signal.

The effect of anthropogenic greenhouse gases is detected, despite uncertainties in sulphate aerosol forcing and response.
The analysis used to derive Figures 12.10a and 12.12, left box, assumes that the ratio of the greenhouse gas and sulphate aerosol responses in each model is correct. Given the uncertainty in sulphate aerosol forcing, this may not be the case. Hence one must also consider the separate responses to greenhouse gases and aerosols simultaneously. A greenhouse gas signal is consistently detected in the observations (e.g., Figure 12.10b,c, Figure 12.12 right hand boxes; North and Wu, 2001; Tett et al. 2000). The greenhouse gas responses are consistent with the observations in all but one case. The two component studies all indicate a substantial detectable greenhouse gas signal, despite uncertainties in aerosol forcing. The spread of estimates of the sulphate signal emphasises the uncertainty in sulphate aerosol forcing and response.

It is unlikely that detection studies have mistaken a natural signal for an anthropogenic signal.
In order to demonstrate an anthropogenic contribution to climate, it is necessary to rule out the possibility that the detection procedure has mistaken part or all of a natural signal for an anthropogenic change. On physical grounds, natural forcing is unlikely to account completely for the observed warming over the last three to five decades, given that it is likely that the overall trend in natural forcing over most of the 20th century is small or negative. Several studies have involved three or more components - the responses to greenhouse gases, sulphate aerosols and natural (solar, volcanic or volcanic and solar) forcing. These studies all detect a substantial greenhouse gas contribution over the last fifty years, though in one case the estimated greenhouse gas amplitude is inconsistent with observations. Thus it is unlikely that we have misidentified the solar signal completely as a greenhouse gas response, but uncertainty in the amplitude of the response to natural forcing continues to contribute to uncertainty in the size of the anthropogenic signal.

The detection methods used should not be sensitive to errors in the amplitude of the global mean forcing or response.
Signal estimation methods (e.g., Figures 12.10, 12.11 and 12.12) allow for errors in the amplitude of the response, so the results should not be sensitive to errors in the magnitude of the forcing or the magnitude of the simulated model response. This would reduce the impact of uncertainty in indirect sulphate forcing on the estimated greenhouse and net sulphate signal amplitudes, to the extent that the pattern of response to indirect sulphate forcing resembles the pattern of response to direct sulphate forcing. Some models indicate this is may be the case, others do not, so this remains an important source of uncertainty. Note that if the spatio-temporal pattern of response to indirect sulphate forcing were to resemble the greenhouse response, it would lead to the amplitude of the greenhouse response being underestimated in cases where indirect sulphate forcing has not been included in the model. Detection and attribution results are also expected to be insensitive to all but the largest scale details of radiative forcing patterns. Detection is only possible at the largest spatial scales (e.g., Stott and Tett, 1998). In addition, atmospheric motions and large-scale feedbacks smooth out the response. All these arguments tend to reduce the impact of the large uncertainty in the magnitude of the forcing due to indirect sulphate aerosols. The inclusion of forcing from additional aerosols (see Chapter 6) is unlikely to alter our conclusion concerning the detection of a substantial greenhouse gas signal, though it is likely to affect estimates of the sulphate aerosol response. This is because part of the response to sulphate aerosols can be considered as surrogate for other aerosols, even though the patterns of forcing and response may differ on smaller scales. In general, the estimates of global mean forcing for other neglected factors are small (see Chapter 6, Figure 6.6).

Studies of the changes in the vertical patterns of temperature also indicate that there has been an anthropogenic influence on climate over the last 35 years.
One study finds that even when changes in stratospheric ozone and solar irradiance are taken into account, there is a detectable greenhouse gas signal in the vertical temperature record.

Observed and simulated vertical lapse rate changes are inconsistent over the last two decades, but there is an anthropogenic influence on tropospheric temperatures over a longer period.
Over the last twenty years, the observed warming trend in the lower troposphere has been smaller than at the surface. This contrasts with model simulations of the response to anthropogenic greenhouse gases and sulphate aerosols. Natural climate variability and the influence of natural external forcing, such as volcanism, can explain part of this difference. However, a discrepancy remains that cannot be accounted for with current climate models. The reduced warming in the lower troposphere does not, however, call into question the fact that the surface temperature has been warming over the satellite period (e.g., National Academy of Sciences, 2000). Over the longer period for which radiosonde data are available, an anthropogenic influence due to increasing greenhouse gases and decreasing stratospheric ozone is detected in all studies.

Natural factors may have contributed to the early century warming.
Most of the discussion in this section has been concerned with evidence relating to a human effect on late 20th century climate. The observed global mean surface temperature record shows two main periods of warming. Some studies detect a solar influence on surface temperature over the first five decades of the century, with perhaps a small additional warming due to increases in greenhouse gases. One study suggests that the early warming could be due to a combination of anthropogenic effects and a highly unusual internal variation. Thus the early century warming could be due to some combination of natural internal variability, changes in solar irradiance and some anthropogenic influence. The additional warming in the second half-century is most likely to be due to a substantial warming due to increases in greenhouse gases, partially offset by cooling due to aerosols, and perhaps by cooling due to natural factors towards the end of the period.

IPCC Homepage