Review Articles
Climate Change
Review Articles
Greenhouse Gas Induced Climate Change IGabriele C. Hegerl, 2Ulrich Cubasch i Max-Planckqnstitut fi~r Meteorologie, Hamburg, Germany 2 Deutsches Klimarechenzentrum, Hamburg, Germany
Correspondingauthor: Dr. Gabriele Hegerl, Max-Planck-Institut liar Meteorologie, Bundesstra~e 55, D-20146 Hamburg, Germany
Abstract Simulations using global coupled climate models predict a climate change due to the increasing concentration of greenhouse gases and aerosols in the atmosphere. Both are associated with the burning of fossil fuels. There has been considerable debate if this postulated human influence is already evident. This paper gives an overview on some recent material on this question. One particular study using optimal fingerprints (HEGERLet al., 1996) is explained in more detail. In this study, an optimal fingerprint analysis is applied to temperature trend patterns over several decades. The results show the probability being less than 5 % that the most recently observed 30year trend is due to naturally occurring climate fluctuations. This result suggests that the present warming is caused by some external influence on climate, e.g. by the increasing concentrations of greenhouse gases and aerosols. More work is needed to address the uncertainties in the magnitude of naturally occurring climate fluctuations. Also, other external influences on climate need to be investigated to uniquely attribute the present climate change to the human influence. Key words: Climate change; climate models; models, simulation of climate change; climate variability; fingerprint method; greenhouse gases; carbon dioxide; methane; sulfate aerosols; nitrous oxide; anthropogenic impact; troposphere, temperature; natural fluctuations, climate change
1
Introduction
It is known that an increase in some trace gases in the atmosphere may influence the climate (CALLENDAR, 1938). The greenhouse gases not only allow solar radiation to pass the atmosphere, but they also partly prevent heat from being radiated back into space. Thus, an increase in greenhouse gases may change the heat budget of the atmosphere and lead to an increase in the average surface temperature of the earth. Although carbon dioxide is the most important greenhouse gas, other trace gases of anthropogenic origin, such as methane (CH4) and nitrous oxide (N20) also contribute to the "global warming potential". In addition, fossil fuel burning also releases sulphur, which in the atmosphere results in a form of aerosol particles that reflect some parts of the incoming solar radiation into space and thus cause a local cooling (PENNER et al., 1995). Different from greenhouse gases, they work on a rather short time ESPR - Environ. Sci. & Pollut. Res. 3 (2) 99-102 (1996) 9 ecomed publishers, D-86899 Landsberg, Germany
scale and also more regionally than CO 2. Therefore, the influence of sulfate-aerosols may mask greenhouse warming over industrialized areas with strong sulfur emissions and thus change the expected warming pattern. There has been considerable debate whether these postulated mechanisms in fact influence our climate, and are partly responsible for the long-term warming trend in observed global mean temperature (JONESand BRIFFA, 1992). The Intergovernmental Panel on Climate Change (IPCC) evaluates the scientific material on climate change on a regular basis. Recently, the IPCC concluded in its Summary for Policymakers that " ... the balance of evidence suggests that there is a discernible h u m a n influence on climate" (HOUGHTON and MEIRa FILHO, 1996; SANTER et al., 1996a)
( ~ ESPR 3/1, 1996, p. 52). This paper presents some of the evidence upon which such a statement is based.
2
Simulations of Expected Climate Change
Our knowledge of anthropogenic climate change is mostly obtained from globally coupled ocean atmosphere models, the so-called General Circulation Models, simply referred to as climate models. Climate models generally consist of circulation models for the atmosphere and the ocean, plus a sea-ice component (CUBASCHet al., 1992; 1995). Typically, each model is based on equations describing the flow of air, water, salinity etc. in the ocean and atmosphere. These equations are solved numerically, using a representation of the earth by a limited discretization (e.g. by a finite number of grid points). Climate models are useful for detection studies for two reasons: 9 They allow predictions of the expected climate change, e.g. due to the anthropogenic increase of atmospheric greenhouse gas c o n c e n t r a t i o n s and t r o p o s p h e r i c aerosols. Note that this is a change in the mean state of climate, which is predictable in the long term. 9 Long-term climate model simulations enable us to estimate the natural variability of the climate system, which originates from the internal instability of the ocean-at-
99
Climate Change mosphere system. For detection purposes, the recent observations need to be compared with natural climate variability. Climate model simulations, which are run without external changes (e.g. in greenhouse gases) show climate fluctuations similar to the real climate system (STOUFFER et al., 1994; MITCHELL et al., 1995; VON STORCH et al., 1996). Fig. 1 shows the expected change of near surface temperature due to increasing concentrations of greenhouse gases and aerosols in the middle of the next century (HASSELMANNet al., 1995) as simulated by two simulations of a climate model developed in Hamburg (Voss et al., 1996)! The concentrations of greenhouse gases and aerosols in the model are derived from observations for the years 1880 to the present. For the future, a scenario of the evolution of anthropogenic emissions has been used (HOUGHTON et al., 1992). The indirect effect of aerosols on clouds has not been taken into account. Fig. 1 shows (with small exceptions) an overall warming, which is generally stronger in high latitudes and over continents. In the northern hemisphere, the warming is delayed due to the cooling effect of the aerosols over industrialized areas. Simulations with a climate model developed in the U.K. showed quite similar results (MITCHELLet al., 1996).
Review Articles The signature of climate change in the observations can be separated earlier and more reliably from natural climate fluctuations by "fingerprint methods" rather than using a single climate parameter, e.g. global mean temperature. Fingerprint methods use our knowledge about the structure of anthropogenic climate change due to the prediction of climate models. This may be the spatial pattern of the expected climate change (-9 Fig. i) or its time evolution (see e.g. HASSELMANN, 1993; BARNETT et al., 1991). The model-predicted pattern of anthropogenic climate change (referred to as the "fingerprint") is compared to the observations by some univariate detection variable. This may be done by a projection of the observations onto the fingerprint or by a pattern correlation (in that case the spatial pattern of the climate change is compared to the observations in a similar way as time series are correlated). Using a statistically optimal fingerprint further increases the chance to detect anthropogenic climate change. An optimal fingerprint is also derived from the model-predicted pattern of climate change. However, some components of the pattern are emphasized which are unusual for naturally occurring climate fluctuations, other components may be suppressed which are associated with strong natural climate variability (HASSELMANN et al., 1993). Thus, if the observations are projected on an optimal fingerprint for anthropogenic climate change, natural fluctuations suppressed allow an earlier detection of climate change.
4
Fig. 1:
3
Spatial pattern of near surface temperature change (in K) due to the increasing concentrations of greenhouse gas and aerosols in the atmosphere, as simulated in a climate model. The pattern represents the difference between annual mean temperature of the decade 2040-2049 relative to the mean climate at the beginning of the simulation (1880-1889)
Fingerprint M e t h o d s for Detecting Anthropogenic Climate C h a n g e
The first question for detecting a model-predicted climate change is: Are the recent observations consistent with estimates of naturally occurring climate fluctuations? If the observed warming is outside a given confidence interval, where e.g. 95 % of the climate fluctuations occur, it is considered inconsistent (with a statistical error of 5 %) with naturally occurring climate fluctuations. Then we conclude that a change in the mean state of climate has in fact occurred. 100
Results of an Optimal Fingerprint Detection Study
This optimal fingerprint method has been applied to nearsurface temperature observations (HEGERL e t al., 1994; HASSELMANNet al., 1995). Temperature has been observed for a comparatively long time, therefore providing reasonably good information on the time dependence of the observed climate change and on observed climate variability on decadal timescales. Worldwide surface temperature observations, e.g. at weather stations, have been collected and assembled on a grid covering most of the earth's surface (JONES and BRIFFA,1992). These datas have been carefully checked for quality, and then homogenized. For the detection study, linear trends have been fitted to the 30-year time series of observations at each grid-point, yielding temperature trend patterns, beginning with 1860-1889 and ending 1965-1994. Areas of the globe without sufficient data to reliably compute such trends since 1949 have been omitted from the analysis. These gaps occur mainly in high latitudes and to some extent also in the interior of continents (e.g. Africa and South America). The remaining area for the analysis covers about 75 % of the globe. The model-predicted pattern of temperature trends due to the increase in greenhouse gas and aerosol concentrations, the "guess-pattern", is derived from the dominant pattern of the average of both climate change simulations. The pattern is similar to that shown in Fig. 1. An optimal fingerprint is computed from this guess-pattern by. using a long model control simulation (YON STORCHet ESPR- Environ. Sci. & Pollut. Res. 3 (2) 1996
Review Articles al., 1996) which shows the components of the guess-pattern associated with high natural variability. Fig. 2 shows the temporal evolution of the detection variable for 30-year trends (HASSELMANN et al., 1995). It is derived by projecting the observed 30-year trends onto an optimal fingerprint for the expected climate change, and is similar to trends in the observed global mean temperature record. The time refers to the final year of the trend. The 95 % confidence interval is derived from different estimates of the 95 % confidence limit for values of the detection variable which occur due to natural fluctuations of the climate system. This confidence interval has been estimated by observations and also by the fluctuations in different climate models; the highest estimate is given in the figure. Also shown is the evolution of the detection variable when computed from model data rather than observations. All time series of detection variables show fluctuations associated with climate variations in addition to a long-term rise.
Climate Change higher variability in the observations may be due to the effect of other external forcing mechanisms on the observations (e.g. volcanic eruptions which cause cooling over several years, changes in solar irradiance, etc.). However, the disagreement between the different model simulations shows that there is still some uncertainty in our knowledge of the magnitude of decadal scale climate fluctuations. Fig. 2 shows the 95 % confidence limit derived from the data which yielded the largest estimate of climate variability (i.e. the observations). Clearly, the detection variable for the latest 30-year trend exceeds this confidence limit. Thus, we conclude, the probability is less than 5 % (2.5 % for a one-tailed statistical test) that the latest observed 30-year trend of near surface temperature is due to our estimated natural variability and that we have detected a significant climate change. The detection variable also exceeds the confidence interval for a short period in the first part of the 20th century (ending in 1946). One possible interpretation of this feature is that this unusual trend is associated with an extreme event of natural climate variability. Also, there was probably a small greenhouse gas-forced component - if the modelestimated greenhouse gas signal is subtracted from the observations, the detection variable clearly decreases for this time period.
5 Towards Attributing Climate Change We conclude that a change in the mean state of climate has indeed occurred. However, it is not yet clear if this climate change can be attributed to the anthropogenic impact or if other causes are the responsible factors. Fig. 2:
Evolution of the detection variable of 30-year trend patterns computed from the observations (black line). The blue and green lines are computed with two simulations forced with greenhouse gases and aerosols ("A", "B"). The simulations differ due to the internal variability of the climate system. The red line uses a simulation forced with greenhouse gases only ("C"). In the latter case, the detection variable departs from the confidence interval earlier and indicates larger temperature increases. (From HASSELMANNet al., 1995)
The detection variable for the latest observed temperature trend 1965-1994 is higher than previous values. For detection we have to decide whether its value is still consistent with climate variability. To determine this, the detection variable is computed using estimates of the natural variability of the climate system which are derived from different long unforced climate simulations (STOUFFER et al., 1994). It is desirable to verify these estimates of climate variability with observations. However, the observations may already contain the greenhouse warming signal and thus cannot be used directly for a comparison. Therefore, we have subtracted a purely model-derived estimate of the greenhouse warming signal from the observations prior to the comparison. Generally, the variability of the observations exceeds that of the unforced model simulations, and the internal variability differs between the models. The ESPR- Environ. Sci. & Pollut. Res. 3 (2) 1996
If the model-predicted pattern ( ~ Fig. 1) and that of the observations agree to a high extent, this gives some confidence that the recent climate change is in fact of human origin. Note that perfect agreement cannot be expected, since climate fluctuations will always mask a climate change pattern. Studies using pattern correlations support the conclusion that the present warming is of anthropogenic origin. For example, SANTER et al. (1995) found unusually high (relative to climate variability) trends in a pattern correlation between observed near surface temperature and a model simulation forced by greenhouse gases and aerosols, especially for the summer and autumn period. This suggests that the pattern of man-made climate change is emerging above the level of natural climate variability, and bears a spatial correspondence to observations. Similar conclusions have been reached with analysis of resuits from the coupled climate model developed at the Hadley Centre in the United Kingdom (MITCHELLet al., 1995). For the last two decades, the correlation between a combined greenhouse gas and aerosol-forced simulation and observations is larger than expected from natural climate variability showing a 90 % significance level. An increasing agreement between observations and simulations of the combined greenhouse gas and aerosol effect can also be found for the vertical structure of temperature in the atmosphere (SANTERet al., 1996b). The correlations increase
101
Climate Change if ozone depletion (which also has a thermal effect in the upper atmosphere) is taken into account. Although a high similarity of observed and simulated climate change patterns suggests that the climate change has in fact been caused by the anthropogenic influence, it is still conceivable that other forcing mechanisms, together with natural climate variability, might cause a similar pattern in the observations. To rigorously attribute the observed climate change to the human influence, other possible mechanisms need to be specified and ruled out, e.g. using a multiple fingerprint technique (HASSELMANN,1993). For example, it has been hypothesized that changes in solar irradiance may be responsible for the present climate change. However, even a rather high estimate of changes in solar irradiance seems too small to explain the present warming t r e n d (CROWLEY and KIM, 1996; CUBASCH, in preparation). Preliminary results of a two-fingerprint approach using optimal fingerprints for CO 2 only and for the combined forcing, show that the observations are in fact inconsistent with a pure greenhouse gas signal but consistent with the combined greenhouse gas and aerosol model simulation (HEGERLet al., in prep.). 6
Conclusions
There is increasing evidence that the present warming, especially over the last few decades, disagrees from naturally occurring climate fluctuations. This disagreement is highly significant for the recently observed 30-year temperature trend pattern. There is also an increasing similarity between model-simulated patterns (if the effect of aerosol forcing on climate is taken into account) and observed climate patterns. All detection efforts, however, suffer from uncertainties associated with the estimate of the magnitude of naturally occurring climate fluctuations. Better and more realistic climate models will further increase our confidence that model fluctuations are a good estimate of internal climate variability. Although the observed climate change is consistent with the model prediction, and we can offer no other convincing explanation for the observed climate change, we cannot unequivocally attribute the observed climate change to greenhouse gas and aerosol forcing. To definitely rule out other possible climate change mechanisms, further research is necessary. Nevertheless, results to date increasingly point to the anthropogenic perturbation as the source of the recent changes in climate. Acknowledgments The authors are grateful to Tom CROWLEY,Klaus HASSELMANN,Reinhard Voss and Phil JONES for help in various aspects of this work. Jiargen WASZKEWITZand Marion GRLrNERTdrew the diagrams. The work was sponsored by the Max-Planck-Gesellschaft and by the EC Environmental Program under contract No. EV5V-CT92-0123 and ENV4CT95-0102.
7
Review Articles Critical Appraisal of Simulations and Observations. M. E. SCHLES~NGER (Ed.), Elsevier Science Publishers, Amsterdam, 537-558 CALLENDAR,G. S. (1938): The artificial production of carbon dioxide and its influence on temperature. Quart. J. Roy. Met. Soc. 64, 223-235 CROWLEY,T. J.; K-Y. KIM (1996): Comarison of proxy records of climate change and solar forcing. Geophys. Res. Letters. 23, 359-362 CUBASCHW.; K. HASSELMANN;H. HOCK; E. MAIER-REIMER;U. MIKOLAJEWICZ; B. D. SANTER;R. SAUSEN(1992): Time-dependent greenhouse warming computations with a coupled ocean-atmosphere model. Climate Dynamics 8, 55-69 CUBASCHU.; G. C. HEGERL;A. HELLBACH;H. HOCK; U. MIKOLAJEWICZ; B. D. SAVORER;R. Voss (1995): A Climate change simulation starting from 1935. Climate Dynamics 11, 71-84 HASSELMANNK. (1993): Optimal fingerprints for the Detection of Time dependent Climate Change. J. Climate 6, 1957-1971 HASSELMANNK.; L. BENGTSSON;U. CUBASCH;G. C. HEGERL;H. RODHE; E. ROECKNER;H. VON STORCH; R. VOSS; J. WASZKEWITZ(1995): Detection of anthropogenic climate change using a fingerprint method. Max- Planck Institut fi~r Meteorologie; Report 168 and Proc. "Modern Dynamical Meteorology", Symposium in Honor of AKSELWIIN-NIELSEN, 1995. Ed. E DITLEVSEN(ECMWF press 1995), 203-221 HEGERL, G. C.; H. VON STORCH;K. HASSELMANN;B. D. SANTER;U. CUBASCH;P. D. JONES (1996): Detecting Greenhouse Gas induced Climate Change with an Optimal Fingerprint Method. J. Climate, in press HOUGHTON J. T.; B. A. CALLANDER;S. K. VARNEY (1992): Climate Change 1992. The supplementary report to the IPCC Scientific Assessment. Cambridge University Press, Cambridge, 200 pp HOUGHTONJ. T.; L. G. MEIRAFILHO(1996): Climate Change 1995. The IPCC second scientific assessment. Cambridge University Press, Cambridge, 572 pp JONES P. D.; K. R. BRIFFA(1992): Global surface air temperature variations during the twentieth century: Part 1, spatial, temporal and seasonal details. The Holocene 2, 165-179 MITCHELL, J. E B.; T. J. JOHNS; J. M. GREGORY;S. E B. TETT (1995): Transient climate response to increasing sulphate aerosols and greenhouse gases. Nature 376, 501-504 PENNER,J. E.; R. J. CHARLSON;J. M. HALES;N. S. LAULAINEN;R. LEIFER; T. NOVAKOV;J. OGRED; L. E RADKE; S. E. SCHWARTZ;L. TRAVIS (1995): Quantifying and Minimizing the Uncertainty of Climate Forcing by Anthropogenic Aerosols. Bull. Am. Met. Soc. 75, 375-400 SANTERB. D.; K. E. TAYLOR;J. E. PENNER;T. M. L. WIGLEY;U. CUBASCH; P.D. JONES(1995): Towards the detection and attribution of an anthropogenic effect on climate. Climate Dynamics 12, 77-100 SANTER, B. D.; T. M. L. WIGLEY;T. P. BARNETt;E. ANYAMBA(1996a): Detection of climate change and attribution of causes. Climate change 1995. The IPCC Second Scientific Assessment. (Ed.) J. T. HOUGHTONet al. 407-444 SANTERB. D.; K. E. TAYLOR;T. M. L. WIGLEY;P. D. JONES;D. J. KAROLY; J. E B. MITCHELL;A. H. OORT;J. E. PENNER;V. RAMASWAMY;M. D. SCHWARZKOPF;R. J. STOUFFER;S. TETr (1996b): A search for human influences on the thermal structure in the atmosphere. Nature, in press STOUFFERR.J.; S. MANABE;K. Y. VINNIKOV(1994): Model assessment of the role of natural variability in recent global warming. Nature 367, 634-636 VON STORCHJ.; V. KHARIN;U. CUBASCH;G. C. HEGERL;D. SCHRIEVER; H. YON STORCH;E. ZORITA (1996): A 1260-year control integration with the coupled ECHAM1/LSG general circulation model. Submitted to J. Climate Voss, R.; R. SAUSEN;U. CUBASCH(1996): Periodically synchronously coupled integrations with the atmosphere-ocean general circulation model ECHAM3/LSG. Part I: Simulations of the present-day climate. In preparation
References
BARNETT,T. P.; M. E. SCHLESINGER;X. JIANG (1991): On greenhouse gas detection strategies. Greenhouse-Gas-Induced-Climatic Change: A
102
ESPR - Environ. Sci. & Pollut. Res. 3 (2) 1996
Climate Change
Review Articles
Greenhouse Gas Induced Climate Change IGabriele C. Hegerl, 2Ulrich Cubasch i Max-Planckqnstitut fi~r Meteorologie, Hamburg, Germany 2 Deutsches Klimarechenzentrum, Hamburg, Germany
Correspondingauthor: Dr. Gabriele Hegerl, Max-Planck-Institut liar Meteorologie, Bundesstra~e 55, D-20146 Hamburg, Germany
Abstract Simulations using global coupled climate models predict a climate change due to the increasing concentration of greenhouse gases and aerosols in the atmosphere. Both are associated with the burning of fossil fuels. There has been considerable debate if this postulated human influence is already evident. This paper gives an overview on some recent material on this question. One particular study using optimal fingerprints (HEGERLet al., 1996) is explained in more detail. In this study, an optimal fingerprint analysis is applied to temperature trend patterns over several decades. The results show the probability being less than 5 % that the most recently observed 30year trend is due to naturally occurring climate fluctuations. This result suggests that the present warming is caused by some external influence on climate, e.g. by the increasing concentrations of greenhouse gases and aerosols. More work is needed to address the uncertainties in the magnitude of naturally occurring climate fluctuations. Also, other external influences on climate need to be investigated to uniquely attribute the present climate change to the human influence. Key words: Climate change; climate models; models, simulation of climate change; climate variability; fingerprint method; greenhouse gases; carbon dioxide; methane; sulfate aerosols; nitrous oxide; anthropogenic impact; troposphere, temperature; natural fluctuations, climate change
1
Introduction
It is known that an increase in some trace gases in the atmosphere may influence the climate (CALLENDAR, 1938). The greenhouse gases not only allow solar radiation to pass the atmosphere, but they also partly prevent heat from being radiated back into space. Thus, an increase in greenhouse gases may change the heat budget of the atmosphere and lead to an increase in the average surface temperature of the earth. Although carbon dioxide is the most important greenhouse gas, other trace gases of anthropogenic origin, such as methane (CH4) and nitrous oxide (N20) also contribute to the "global warming potential". In addition, fossil fuel burning also releases sulphur, which in the atmosphere results in a form of aerosol particles that reflect some parts of the incoming solar radiation into space and thus cause a local cooling (PENNER et al., 1995). Different from greenhouse gases, they work on a rather short time ESPR - Environ. Sci. & Pollut. Res. 3 (2) 99-102 (1996) 9 ecomed publishers, D-86899 Landsberg, Germany
scale and also more regionally than CO 2. Therefore, the influence of sulfate-aerosols may mask greenhouse warming over industrialized areas with strong sulfur emissions and thus change the expected warming pattern. There has been considerable debate whether these postulated mechanisms in fact influence our climate, and are partly responsible for the long-term warming trend in observed global mean temperature (JONESand BRIFFA, 1992). The Intergovernmental Panel on Climate Change (IPCC) evaluates the scientific material on climate change on a regular basis. Recently, the IPCC concluded in its Summary for Policymakers that " ... the balance of evidence suggests that there is a discernible h u m a n influence on climate" (HOUGHTON and MEIRa FILHO, 1996; SANTER et al., 1996a)
( ~ ESPR 3/1, 1996, p. 52). This paper presents some of the evidence upon which such a statement is based.
2
Simulations of Expected Climate Change
Our knowledge of anthropogenic climate change is mostly obtained from globally coupled ocean atmosphere models, the so-called General Circulation Models, simply referred to as climate models. Climate models generally consist of circulation models for the atmosphere and the ocean, plus a sea-ice component (CUBASCHet al., 1992; 1995). Typically, each model is based on equations describing the flow of air, water, salinity etc. in the ocean and atmosphere. These equations are solved numerically, using a representation of the earth by a limited discretization (e.g. by a finite number of grid points). Climate models are useful for detection studies for two reasons: 9 They allow predictions of the expected climate change, e.g. due to the anthropogenic increase of atmospheric greenhouse gas c o n c e n t r a t i o n s and t r o p o s p h e r i c aerosols. Note that this is a change in the mean state of climate, which is predictable in the long term. 9 Long-term climate model simulations enable us to estimate the natural variability of the climate system, which originates from the internal instability of the ocean-at-
99
Climate Change mosphere system. For detection purposes, the recent observations need to be compared with natural climate variability. Climate model simulations, which are run without external changes (e.g. in greenhouse gases) show climate fluctuations similar to the real climate system (STOUFFER et al., 1994; MITCHELL et al., 1995; VON STORCH et al., 1996). Fig. 1 shows the expected change of near surface temperature due to increasing concentrations of greenhouse gases and aerosols in the middle of the next century (HASSELMANNet al., 1995) as simulated by two simulations of a climate model developed in Hamburg (Voss et al., 1996)! The concentrations of greenhouse gases and aerosols in the model are derived from observations for the years 1880 to the present. For the future, a scenario of the evolution of anthropogenic emissions has been used (HOUGHTON et al., 1992). The indirect effect of aerosols on clouds has not been taken into account. Fig. 1 shows (with small exceptions) an overall warming, which is generally stronger in high latitudes and over continents. In the northern hemisphere, the warming is delayed due to the cooling effect of the aerosols over industrialized areas. Simulations with a climate model developed in the U.K. showed quite similar results (MITCHELLet al., 1996).
Review Articles The signature of climate change in the observations can be separated earlier and more reliably from natural climate fluctuations by "fingerprint methods" rather than using a single climate parameter, e.g. global mean temperature. Fingerprint methods use our knowledge about the structure of anthropogenic climate change due to the prediction of climate models. This may be the spatial pattern of the expected climate change (-9 Fig. i) or its time evolution (see e.g. HASSELMANN, 1993; BARNETT et al., 1991). The model-predicted pattern of anthropogenic climate change (referred to as the "fingerprint") is compared to the observations by some univariate detection variable. This may be done by a projection of the observations onto the fingerprint or by a pattern correlation (in that case the spatial pattern of the climate change is compared to the observations in a similar way as time series are correlated). Using a statistically optimal fingerprint further increases the chance to detect anthropogenic climate change. An optimal fingerprint is also derived from the model-predicted pattern of climate change. However, some components of the pattern are emphasized which are unusual for naturally occurring climate fluctuations, other components may be suppressed which are associated with strong natural climate variability (HASSELMANN et al., 1993). Thus, if the observations are projected on an optimal fingerprint for anthropogenic climate change, natural fluctuations suppressed allow an earlier detection of climate change.
4
Fig. 1:
3
Spatial pattern of near surface temperature change (in K) due to the increasing concentrations of greenhouse gas and aerosols in the atmosphere, as simulated in a climate model. The pattern represents the difference between annual mean temperature of the decade 2040-2049 relative to the mean climate at the beginning of the simulation (1880-1889)
Fingerprint M e t h o d s for Detecting Anthropogenic Climate C h a n g e
The first question for detecting a model-predicted climate change is: Are the recent observations consistent with estimates of naturally occurring climate fluctuations? If the observed warming is outside a given confidence interval, where e.g. 95 % of the climate fluctuations occur, it is considered inconsistent (with a statistical error of 5 %) with naturally occurring climate fluctuations. Then we conclude that a change in the mean state of climate has in fact occurred. 100
Results of an Optimal Fingerprint Detection Study
This optimal fingerprint method has been applied to nearsurface temperature observations (HEGERL e t al., 1994; HASSELMANNet al., 1995). Temperature has been observed for a comparatively long time, therefore providing reasonably good information on the time dependence of the observed climate change and on observed climate variability on decadal timescales. Worldwide surface temperature observations, e.g. at weather stations, have been collected and assembled on a grid covering most of the earth's surface (JONES and BRIFFA,1992). These datas have been carefully checked for quality, and then homogenized. For the detection study, linear trends have been fitted to the 30-year time series of observations at each grid-point, yielding temperature trend patterns, beginning with 1860-1889 and ending 1965-1994. Areas of the globe without sufficient data to reliably compute such trends since 1949 have been omitted from the analysis. These gaps occur mainly in high latitudes and to some extent also in the interior of continents (e.g. Africa and South America). The remaining area for the analysis covers about 75 % of the globe. The model-predicted pattern of temperature trends due to the increase in greenhouse gas and aerosol concentrations, the "guess-pattern", is derived from the dominant pattern of the average of both climate change simulations. The pattern is similar to that shown in Fig. 1. An optimal fingerprint is computed from this guess-pattern by. using a long model control simulation (YON STORCHet ESPR- Environ. Sci. & Pollut. Res. 3 (2) 1996
Review Articles al., 1996) which shows the components of the guess-pattern associated with high natural variability. Fig. 2 shows the temporal evolution of the detection variable for 30-year trends (HASSELMANN et al., 1995). It is derived by projecting the observed 30-year trends onto an optimal fingerprint for the expected climate change, and is similar to trends in the observed global mean temperature record. The time refers to the final year of the trend. The 95 % confidence interval is derived from different estimates of the 95 % confidence limit for values of the detection variable which occur due to natural fluctuations of the climate system. This confidence interval has been estimated by observations and also by the fluctuations in different climate models; the highest estimate is given in the figure. Also shown is the evolution of the detection variable when computed from model data rather than observations. All time series of detection variables show fluctuations associated with climate variations in addition to a long-term rise.
Climate Change higher variability in the observations may be due to the effect of other external forcing mechanisms on the observations (e.g. volcanic eruptions which cause cooling over several years, changes in solar irradiance, etc.). However, the disagreement between the different model simulations shows that there is still some uncertainty in our knowledge of the magnitude of decadal scale climate fluctuations. Fig. 2 shows the 95 % confidence limit derived from the data which yielded the largest estimate of climate variability (i.e. the observations). Clearly, the detection variable for the latest 30-year trend exceeds this confidence limit. Thus, we conclude, the probability is less than 5 % (2.5 % for a one-tailed statistical test) that the latest observed 30-year trend of near surface temperature is due to our estimated natural variability and that we have detected a significant climate change. The detection variable also exceeds the confidence interval for a short period in the first part of the 20th century (ending in 1946). One possible interpretation of this feature is that this unusual trend is associated with an extreme event of natural climate variability. Also, there was probably a small greenhouse gas-forced component - if the modelestimated greenhouse gas signal is subtracted from the observations, the detection variable clearly decreases for this time period.
5 Towards Attributing Climate Change We conclude that a change in the mean state of climate has indeed occurred. However, it is not yet clear if this climate change can be attributed to the anthropogenic impact or if other causes are the responsible factors. Fig. 2:
Evolution of the detection variable of 30-year trend patterns computed from the observations (black line). The blue and green lines are computed with two simulations forced with greenhouse gases and aerosols ("A", "B"). The simulations differ due to the internal variability of the climate system. The red line uses a simulation forced with greenhouse gases only ("C"). In the latter case, the detection variable departs from the confidence interval earlier and indicates larger temperature increases. (From HASSELMANNet al., 1995)
The detection variable for the latest observed temperature trend 1965-1994 is higher than previous values. For detection we have to decide whether its value is still consistent with climate variability. To determine this, the detection variable is computed using estimates of the natural variability of the climate system which are derived from different long unforced climate simulations (STOUFFER et al., 1994). It is desirable to verify these estimates of climate variability with observations. However, the observations may already contain the greenhouse warming signal and thus cannot be used directly for a comparison. Therefore, we have subtracted a purely model-derived estimate of the greenhouse warming signal from the observations prior to the comparison. Generally, the variability of the observations exceeds that of the unforced model simulations, and the internal variability differs between the models. The ESPR- Environ. Sci. & Pollut. Res. 3 (2) 1996
If the model-predicted pattern ( ~ Fig. 1) and that of the observations agree to a high extent, this gives some confidence that the recent climate change is in fact of human origin. Note that perfect agreement cannot be expected, since climate fluctuations will always mask a climate change pattern. Studies using pattern correlations support the conclusion that the present warming is of anthropogenic origin. For example, SANTER et al. (1995) found unusually high (relative to climate variability) trends in a pattern correlation between observed near surface temperature and a model simulation forced by greenhouse gases and aerosols, especially for the summer and autumn period. This suggests that the pattern of man-made climate change is emerging above the level of natural climate variability, and bears a spatial correspondence to observations. Similar conclusions have been reached with analysis of resuits from the coupled climate model developed at the Hadley Centre in the United Kingdom (MITCHELLet al., 1995). For the last two decades, the correlation between a combined greenhouse gas and aerosol-forced simulation and observations is larger than expected from natural climate variability showing a 90 % significance level. An increasing agreement between observations and simulations of the combined greenhouse gas and aerosol effect can also be found for the vertical structure of temperature in the atmosphere (SANTERet al., 1996b). The correlations increase
101
Climate Change if ozone depletion (which also has a thermal effect in the upper atmosphere) is taken into account. Although a high similarity of observed and simulated climate change patterns suggests that the climate change has in fact been caused by the anthropogenic influence, it is still conceivable that other forcing mechanisms, together with natural climate variability, might cause a similar pattern in the observations. To rigorously attribute the observed climate change to the human influence, other possible mechanisms need to be specified and ruled out, e.g. using a multiple fingerprint technique (HASSELMANN,1993). For example, it has been hypothesized that changes in solar irradiance may be responsible for the present climate change. However, even a rather high estimate of changes in solar irradiance seems too small to explain the present warming t r e n d (CROWLEY and KIM, 1996; CUBASCH, in preparation). Preliminary results of a two-fingerprint approach using optimal fingerprints for CO 2 only and for the combined forcing, show that the observations are in fact inconsistent with a pure greenhouse gas signal but consistent with the combined greenhouse gas and aerosol model simulation (HEGERLet al., in prep.). 6
Conclusions
There is increasing evidence that the present warming, especially over the last few decades, disagrees from naturally occurring climate fluctuations. This disagreement is highly significant for the recently observed 30-year temperature trend pattern. There is also an increasing similarity between model-simulated patterns (if the effect of aerosol forcing on climate is taken into account) and observed climate patterns. All detection efforts, however, suffer from uncertainties associated with the estimate of the magnitude of naturally occurring climate fluctuations. Better and more realistic climate models will further increase our confidence that model fluctuations are a good estimate of internal climate variability. Although the observed climate change is consistent with the model prediction, and we can offer no other convincing explanation for the observed climate change, we cannot unequivocally attribute the observed climate change to greenhouse gas and aerosol forcing. To definitely rule out other possible climate change mechanisms, further research is necessary. Nevertheless, results to date increasingly point to the anthropogenic perturbation as the source of the recent changes in climate. Acknowledgments The authors are grateful to Tom CROWLEY,Klaus HASSELMANN,Reinhard Voss and Phil JONES for help in various aspects of this work. Jiargen WASZKEWITZand Marion GRLrNERTdrew the diagrams. The work was sponsored by the Max-Planck-Gesellschaft and by the EC Environmental Program under contract No. EV5V-CT92-0123 and ENV4CT95-0102.
7
Review Articles Critical Appraisal of Simulations and Observations. M. E. SCHLES~NGER (Ed.), Elsevier Science Publishers, Amsterdam, 537-558 CALLENDAR,G. S. (1938): The artificial production of carbon dioxide and its influence on temperature. Quart. J. Roy. Met. Soc. 64, 223-235 CROWLEY,T. J.; K-Y. KIM (1996): Comarison of proxy records of climate change and solar forcing. Geophys. Res. Letters. 23, 359-362 CUBASCHW.; K. HASSELMANN;H. HOCK; E. MAIER-REIMER;U. MIKOLAJEWICZ; B. D. SANTER;R. SAUSEN(1992): Time-dependent greenhouse warming computations with a coupled ocean-atmosphere model. Climate Dynamics 8, 55-69 CUBASCHU.; G. C. HEGERL;A. HELLBACH;H. HOCK; U. MIKOLAJEWICZ; B. D. SAVORER;R. Voss (1995): A Climate change simulation starting from 1935. Climate Dynamics 11, 71-84 HASSELMANNK. (1993): Optimal fingerprints for the Detection of Time dependent Climate Change. J. Climate 6, 1957-1971 HASSELMANNK.; L. BENGTSSON;U. CUBASCH;G. C. HEGERL;H. RODHE; E. ROECKNER;H. VON STORCH; R. VOSS; J. WASZKEWITZ(1995): Detection of anthropogenic climate change using a fingerprint method. Max- Planck Institut fi~r Meteorologie; Report 168 and Proc. "Modern Dynamical Meteorology", Symposium in Honor of AKSELWIIN-NIELSEN, 1995. Ed. E DITLEVSEN(ECMWF press 1995), 203-221 HEGERL, G. C.; H. VON STORCH;K. HASSELMANN;B. D. SANTER;U. CUBASCH;P. D. JONES (1996): Detecting Greenhouse Gas induced Climate Change with an Optimal Fingerprint Method. J. Climate, in press HOUGHTON J. T.; B. A. CALLANDER;S. K. VARNEY (1992): Climate Change 1992. The supplementary report to the IPCC Scientific Assessment. Cambridge University Press, Cambridge, 200 pp HOUGHTONJ. T.; L. G. MEIRAFILHO(1996): Climate Change 1995. The IPCC second scientific assessment. Cambridge University Press, Cambridge, 572 pp JONES P. D.; K. R. BRIFFA(1992): Global surface air temperature variations during the twentieth century: Part 1, spatial, temporal and seasonal details. The Holocene 2, 165-179 MITCHELL, J. E B.; T. J. JOHNS; J. M. GREGORY;S. E B. TETT (1995): Transient climate response to increasing sulphate aerosols and greenhouse gases. Nature 376, 501-504 PENNER,J. E.; R. J. CHARLSON;J. M. HALES;N. S. LAULAINEN;R. LEIFER; T. NOVAKOV;J. OGRED; L. E RADKE; S. E. SCHWARTZ;L. TRAVIS (1995): Quantifying and Minimizing the Uncertainty of Climate Forcing by Anthropogenic Aerosols. Bull. Am. Met. Soc. 75, 375-400 SANTERB. D.; K. E. TAYLOR;J. E. PENNER;T. M. L. WIGLEY;U. CUBASCH; P.D. JONES(1995): Towards the detection and attribution of an anthropogenic effect on climate. Climate Dynamics 12, 77-100 SANTER, B. D.; T. M. L. WIGLEY;T. P. BARNETt;E. ANYAMBA(1996a): Detection of climate change and attribution of causes. Climate change 1995. The IPCC Second Scientific Assessment. (Ed.) J. T. HOUGHTONet al. 407-444 SANTERB. D.; K. E. TAYLOR;T. M. L. WIGLEY;P. D. JONES;D. J. KAROLY; J. E B. MITCHELL;A. H. OORT;J. E. PENNER;V. RAMASWAMY;M. D. SCHWARZKOPF;R. J. STOUFFER;S. TETr (1996b): A search for human influences on the thermal structure in the atmosphere. Nature, in press STOUFFERR.J.; S. MANABE;K. Y. VINNIKOV(1994): Model assessment of the role of natural variability in recent global warming. Nature 367, 634-636 VON STORCHJ.; V. KHARIN;U. CUBASCH;G. C. HEGERL;D. SCHRIEVER; H. YON STORCH;E. ZORITA (1996): A 1260-year control integration with the coupled ECHAM1/LSG general circulation model. Submitted to J. Climate Voss, R.; R. SAUSEN;U. CUBASCH(1996): Periodically synchronously coupled integrations with the atmosphere-ocean general circulation model ECHAM3/LSG. Part I: Simulations of the present-day climate. In preparation
References
BARNETT,T. P.; M. E. SCHLESINGER;X. JIANG (1991): On greenhouse gas detection strategies. Greenhouse-Gas-Induced-Climatic Change: A
102
ESPR - Environ. Sci. & Pollut. Res. 3 (2) 1996
