IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group I: The Physical Science Basis

9.7 Combining Evidence of Anthropogenic Climate Change

The widespread change detected in temperature observations of the surface (Sections 9.4.1, 9.4.2, 9.4.3), free atmosphere (Section 9.4.4) and ocean (Section 9.5.1), together with consistent evidence of change in other parts of the climate system (Section 9.5), strengthens the conclusion that greenhouse gas forcing is the dominant cause of warming during the past several decades. This combined evidence, which is summarised in Table 9.4, is substantially stronger than the evidence that is available from observed changes in global surface temperature alone (Figure 3.6).

The evidence from surface temperature observations is strong: The observed warming is highly significant relative to estimates of internal climate variability which, while obtained from models, are consistent with estimates obtained from both instrumental data and palaeoclimate reconstructions. It is extremely unlikely (<5%) that recent global warming is due to internal variability alone such as might arise from El Niño (Section 9.4.1). The widespread nature of the warming (Figures 3.9 and 9.6) reduces the possibility that the warming could have resulted from internal variability. No known mode of internal variability leads to such widespread, near universal warming as has been observed in the past few decades. Although modes of internal variability such as El Niño can lead to global average warming for limited periods of time, such warming is regionally variable, with some areas of cooling (Figures 3.27 and 3.28). In addition, palaeoclimatic evidence indicates that El Niño variability during the 20th century is not unusual relative to earlier periods (Section; Chapter 6). Palaeoclimatic evidence suggests that such a widespread warming has not been observed in the NH in at least the past 1.3 kyr (Osborn and Briffa, 2006), further strengthening the evidence that the recent warming is not due to natural internal variability. Moreover, the response to anthropogenic forcing is detectable on all continents individually except Antarctica, and in some sub-continental regions. Climate models only reproduce the observed 20th-century global mean surface warming when both anthropogenic and natural forcings are included (Figure 9.5). No model that has used natural forcing only has reproduced the observed global mean warming trend or the continental mean warming trends in all individual continents (except Antarctica) over the second half of the 20th century. Detection and attribution of external influences on 20th-century and palaeoclimatic reconstructions, from both natural and anthropogenic sources (Figure 9.4 and Table 9.4), further strengthens the conclusion that the observed changes are very unusual relative to internal climate variability.

The energy content change associated with the observed widespread warming of the atmosphere is small relative to the energy content change of the ocean, and also smaller than that associated with other components such as the cryosphere. In addition, the solid Earth also shows evidence for warming in boreholes (Huang et al., 2000; Beltrami et al., 2002; Pollack and Smerdon, 2004). It is theoretically feasible that the warming of the near surface could have occurred due to a reduction in the heat content of another component of the system. However, all parts of the cryosphere (glaciers, small ice caps, ice sheets and sea ice) have decreased in extent over the past half century, consistent with anthropogenic forcing (Section 9.5.5, Table 9.4), implying that the cryosphere consumed heat and thus indicating that it could not have provided heat for atmospheric warming. More importantly, the heat content of the ocean (the largest reservoir of heat in the climate system) also increased, much more substantially than that of the other components of the climate system (Figure 5.4; Hansen et al., 2005; Levitus et al., 2005). The warming of the upper ocean during the latter half of the 20th century was likely due to anthropogenic forcing (Barnett et al., 2005; Section; Table 9.4). While the statistical evidence in this research is very strong that the warming cannot be explained by ocean internal variability as estimated by two different climate models, uncertainty arises since there are discrepancies between estimates of ocean heat content variability from models and observations, although poor sampling of parts of the World Ocean may explain this discrepancy. However, the spatial pattern of ocean warming with depth is very consistent with heating of the ocean resulting from net positive radiative forcing, since the warming proceeds downwards from the upper layers of the ocean and there is deeper penetration of heat at middle to high latitudes and shallower penetration at low latitudes (Barnett et al., 2005; Hansen et al., 2005). This observed ocean warming pattern is inconsistent with a redistribution of heat between the surface and the deep ocean.

Thus, the evidence appears to be inconsistent with the ocean or land being the source of the warming at the surface. In addition, simulations forced with observed SST changes cannot fully explain the warming in the troposphere without increases in greenhouse gases (e.g., Sexton et al., 2001), further strengthening the evidence that the warming does not originate from the ocean. Further evidence for forced changes arises from widespread melting of the cryosphere (Section 9.5.5), increases in water vapour in the atmosphere (Section and changes in top-of-the atmosphere radiation that are consistent with changes in forcing.

The simultaneous increase in energy content of all the major components of the climate system and the pattern and amplitude of warming in the different components, together with evidence that the second half of the 20th century was likely the warmest in 1.3 kyr (Chapter 6) indicate that the cause of the warming is extremely unlikely to be the result of internal processes alone. The consistency across different lines of evidence makes a strong case for a significant human influence on observed warming at the surface. The observed rates of surface temperature and ocean heat content change are consistent with the understanding of the likely range of climate sensitivity and net climate forcings. Only with a net positive forcing, consistent with observational and model estimates of the likely net forcing of the climate system (as used in Figure 9.5), is it possible to explain the large increase in heat content of the climate system that has been observed (Figure 5.4).

Table 9.4. A synthesis of climate change detection results: (a) surface and atmospheric temperature evidence and (b) evidence from other variables. Note that our likelihood assessments are reduced compared to individual detection studies in order to take into account remaining uncertainties (see Section 9.1.2), such as forcing and model uncertainty not directly accounted for in the studies. The likelihood assessment is indicated in percentage terms, in parentheses where the term is not from the standard IPCC likelihood levels.


Result Region  Likelihood Factors contributing to likelihood assessment 
Surface temperature 
Warming during the past half century cannot be explained without external radiative forcing Global Extremely likely (>95%) Anthropogenic change has been detected in surface temperature with very high significance levels (less than 1% error probability). This conclusion is strengthened by detection of anthropogenic change in the upper ocean with high significance level. Upper ocean warming argues against the surface warming being due to natural internal processes. Observed change is very large relative to climate-model simulated internal variability. Surface temperature variability simulated by models is consistent with variability estimated from instrumental and palaeorecords. Main uncertainty from forcing and internal variability estimates (Sections,,,, 9.7). 
Warming during the past half century is not solely due to known natural causes  Global Very Likely  This warming took place at a time when non-anthropogenic external factors would likely have produced cooling. The combined effect of known sources of forcing would have been extremely likely to produce a warming. No climate model that has used natural forcing only has reproduced the observed global warming trend over the 2nd half of the 20th century. Main uncertainties arise from forcing, including solar, model-simulated responses and internal variability estimates (Sections 2.9.2, 9.2.1,,; Figures 9.5, 9.6, 9.9).  
Greenhouse gas forcing has been the dominant cause of the observed global warming over the last 50 years. Global Very likely  All multi-signal detection and attribution studies attribute more warming to greenhouse gas forcing than to a combination of all other sources considered, including internal variability, with a very high significance. This conclusion accounts for observational, model and forcing uncertainty, and the possibility that the response to solar forcing could be underestimated by models. Main uncertainty from forcing and internal variability estimates (Section; Figure 9.9). 
Increases in greenhouse gas concentrations alone would have caused more warming than observed over the last 50 years because volcanic and anthropogenic aerosols have offset some warming that would otherwise have taken place. Global Likely  Estimates from different analyses using different models show consistently more warming than observed over the last 50 years at the 5% significance level. However, separation of the response to non-greenhouse gas (particularly aerosol) forcing from greenhouse gas forcing varies between models (Section; Figure 9.9). 
There has been a substantial anthropogenic contribution to surface temperature increases in every continent except Antarctica since the middle of the 20th century  Africa, Asia, Australia, Europe, North America and South America Likely  Anthropogenic change has been estimated using detection and attribution methods on every individual continent (except Antarctica). Greater variability compared to other continental regions makes detection more marginal in Europe. No climate model that used natural forcing only reproduced the observed continental mean warming trend over the second half of the 20th century. Uncertainties arise because sampling effects result in lower signal-to-noise ratio at continental than at global scales. Separation of the response to different forcings is more difficult at these spatial scales (Section 9.4.2; FAQ 9.2, Figure 1). 
Early 20th-century warming is due in part to external forcing. Global Very Likely  A number of studies detect the influence of external forcings on early 20th-century warming, including a warming from anthropogenic forcing. Both natural forcing and response are uncertain, and different studies find different forcings dominant. Some studies indicate that internal variability could have made a large contribution to early 20th-century warming. Some observational uncertainty in early 20th-century trend (Sections,; Figures 9.4, 9.5). 
Result Region  Likelihood Factors contributing to likelihood assessment 
Surface temperature 
Pre-industrial temperatures were influenced by natural external forcing (period studied is past 7 centuries)  NH (mostly extratropics) Very Likely  Detection studies indicate that external forcing explains a substantial fraction of inter-decadal variability in NH temperature reconstructions. Simulations in response to estimates of pre-industrial forcing reproduce broad features of reconstructions. Substantial uncertainties in reconstructions and past forcings are unlikely to lead to a spurious agreement between temperature reconstructions and forcing reconstructions as they are derived from independent proxies (Section 9.3.3; Figures 9.4, 6.13). 
Temperature extremes have changed due to anthropogenic forcing NH land areas and Australia combined. Likely  A range of observational evidence indicates that temperature extremes are changing. An anthropogenic influence on the temperatures of the 1, 5, 10 and 30 warmest nights, coldest days and coldest nights annually has been formally detected and attributed in one study, but observed change in the temperature of the warmest day annually is inconsistent with simulated change. The detection of changes in temperature extremes is supported by other comparisons between models and observations. Model uncertainties in changes in temperature extremes are greater than for mean temperatures and there is limited observational coverage and substantial observational uncertainty (Section 9.4.3). 
Free atmosphere changes 
Tropopause height increases are detectable and attributable to anthropogenic forcing (latter half of the 20th century) Global  Likely  There has been robust detection of anthropogenic influence on increasing tropopause height. Simulated tropopause height increases result mainly from greenhouse gas increases and stratospheric ozone decreases. Detection and attribution studies rely on reanalysis data, which are subject to inhomogeneities related to differing availability and quality of input data, although tropopause height increases have also been identified in radiosonde observations. Overall tropopause height increases in recent model and one reanalysis (ERA-40) appear to be driven by similar large-scale changes in atmospheric temperature, although errors in tropospheric warming and stratospheric cooling could lead to partly spurious agreement in other data sets (Section; Figure 9.14).  
Tropospheric warming is detectable and attributable to anthropogenic forcing (latter half of the 20th century) Global Likely  There has been robust detection and attribution of anthropogenic influence on tropospheric warming, which does not depend on including stratospheric cooling in the fingerprint pattern of response. There are observational uncertainties in radiosonde and satellite records. Models generally predict a relative warming of the free troposphere compared to the surface in the tropics since 1979, which is not seen in the radiosonde record (possibly due to uncertainties in the radiosonde record) but is seen in one version of the satellite record, although not others (Section 9.4.4).  
Simultaneous tropospheric warming and stratospheric cooling due to the influence of anthropogenic forcing has been observed (latter half of the 20th century)  Global Very Likely Simultaneous warming of the troposphere and cooling of the stratosphere due to natural factors is less likely than warming of the troposphere or cooling of the stratosphere alone. Cooling of the stratosphere is in part related to decreases in stratospheric ozone. Modelled and observational uncertainties as discussed under entries for tropospheric warming with additional uncertainties due to stratospheric observing systems and the relatively poor representations of stratospheric processes and variability in climate models (Section 9.4.4). 


Result Region  Likelihood Factors contributing to likelihood assessment 
Ocean changes 
Anthropogenic forcing has warmed the upper several hundred metres of the ocean during the latter half of the 20th century  Global (but with limited sampling in some regions) Likely  Robust detection and attribution of anthropogenic fingerprint from three different models in ocean temperature changes, and in ocean heat content data, suggests high likelihood, but observational and modelling uncertainty remains. 20th-century simulations with MMD models simulate comparable ocean warming to observations only if anthropogenic forcing is included. Simulated and observed variability appear inconsistent, either due to sampling errors in the observations or under-simulated internal variability in the models. Limited geographical coverage in some ocean basins (Section; Figure 9.15). 
Anthropogenic forcing contributed to sea level rise during the latter half 20th century Global Very likely Natural factors alone do not satisfactorily explain either the observed thermal expansion of the ocean or the observed sea level rise. Models including anthropogenic and natural forcing simulate the observed thermal expansion since 1961 reasonably well. Anthropogenic forcing dominates the surface temperature change simulated by models, and has likely contributed to the observed warming of the upper ocean and widespread glacier retreat. It is very unlikely that the warming during the past half century is due only to known natural causes. It is therefore very likely that anthropogenic forcing contributed to sea level rise associated with ocean thermal expansion and glacier retreat. However, it remains difficult to estimate the anthropogenic contribution to sea level rise because suitable studies quantifying the anthropogenic contribution to sea level rise and glacier retreat are not available, and because the observed sea level rise budget is not closed (Table 9.2; Section 9.5.2). 
Sea level pressure shows a detectable anthropogenic signature during the latter half of the 20th century Global  Likely  Changes of similar nature are observed in both hemispheres and are qualitatively, but not quantitatively consistent with model simulations. Uncertainty in models and observations. Models underestimate the observed NH changes for reasons that are not understood, based on a small number of studies. Simulated response to 20th century forcings is consistent with observations in SH if effect of stratospheric ozone depletion is included (Section; Figure 9.16). 
Anthropogenic forcing contributed to the increase in frequency of the most intense tropical cyclones since the 1970s  Tropical regions  More likely than not (>50%)  Recent observational evidence suggests an increase in frequency of intense storms. Increase in intensity is consistent with theoretical expectations. Large uncertainties due to models and observations. Modelling studies generally indicate a reduced frequency of tropical cyclones in response to enhanced greenhouse gas forcing, but an increase in the intensity of the most intense cyclones. Observational evidence, which is affected by substantial inhomogeneities in tropical cyclone data sets for which corrections have been attempted, suggests that increases in cyclone intensity since the 1970s are associated with SST and atmospheric water vapour increases (Section 3.8.3, Box 3.5 and Section 
Precipitation, Drought, Runoff 
Volcanic forcing influences total rainfall  Global land areas  More likely than not (>50%) Model response detectable in observations for some models and result supported by theoretical understanding. However, uncertainties in models, forcings and observations. Limited observational sampling, particularly in the SH (Section; Figure 9.18). 
Increases in heavy rainfall are consistent with anthropogenic forcing during latter half 20th century Global land areas (limited sampling) More likely than not (>50%) Observed increases in heavy precipitation appear to be consistent with expectations of response to anthropogenic forcing. Models may not represent heavy rainfall well; observations suffer from sampling inadequacies (Section 
Result Region  Likelihood Factors contributing to likelihood assessment 
Precipitation, Drought, Runoff 
Increased risk of drought due to anthropogenic forcing during latter half 20th century Global land areas More likely than not (>50%) One detection study has identified an anthropogenic fingerprint in a global Palmer Drought Severity Index data set with high significance, but the simulated response to anthropogenic and natural forcing combined is weaker than observed, and the model appears to have less inter-decadal variability than observed. Studies of some regions indicate that droughts in those regions are linked either to SST changes that, in some instances, may be linked to anthropogenic aerosol forcing (e.g., Sahel) or to a circulation response to anthropogenic forcing (e.g., southwest Australia). Models, observations and forcing all contribute uncertainty (Section 
Anthropogenic forcing has contributed to reductions in NH sea ice extent during the latter half of the 20th century Arctic Likely  The observed change is qualitatively consistent with model-simulated changes for most models and expectation of sea ice melting under arctic warming. Sea ice extent change detected in one study. The model used has some deficiencies in arctic sea ice annual cycle and extent. The conclusion is supported by physical expectations and simulations with another climate model. Change in SH sea ice probably within range explained by internal variability (Section 
Anthropogenic forcing has contributed to widespread glacier retreat during the 20th century  Global Likely Observed changes are qualitatively consistent with theoretical expectations and temperature detection. Anthropogenic contribution to volume change difficult to estimate. Few detection and attribution studies, but retreat in vast majority of glaciers consistent with expected reaction to widespread warming (Section