Climate Change 2001:
Working Group I: The Scientific Basis
Other reports in this collection

9.2.2.3 Multi-model ensembles

The collection of coupled climate model results that is available for this report permits a multi-model ensemble approach to the synthesis of projected climate change. Multi-model ensemble approaches are already used in short-range climate forecasting (e.g., Graham et al., 1999; Krishnamurti et al., 1999; Brankovic and Palmer, 2000; Doblas-Reyes et al., 2000; Derome et al., 2001). When applied to climate change, each model in the ensemble produces a somewhat different projection and, if these represent plausible solutions to the governing equations, they may be considered as different realisations of the climate change drawn from the set of models in active use and produced with current climate knowledge. In this case, temperature is represented as T = T0 + TF + Tm + T' where TF is the deterministic forced climate change for the real system and Tm= Tf -TF is the error in the model’s simulation of this forced response. T' now also includes errors in the statistical behaviour of the simulated natural variability. The multi-model ensemble mean estimate of forced climate change is {T} = TF + {Tm} + {T''} where the natural variability again averages to zero for a large enough ensemble. To the extent that unrelated model errors tend to average out, the ensemble mean or systematic error {Tm} will be small, {T} will approach TF and the multi-model ensemble average will be a better estimate of the forced climate change of the real system than the result from a particular model.

As noted in Chapter 8, no one model can be chosen as “best” and it is important to use results from a range of models. Lambert and Boer (2001) show that for the CMIP1 ensemble of simulations of current climate, the multi-model ensemble means of temperature, pressure, and precipitation are generally closer to the observed distributions, as measured by mean squared differences, correlations, and variance ratios, than are the results of any particular model. The multi-model ensemble mean represents those features of projected climate change that survive ensemble averaging and so are common to models as a group. The multi-model ensemble variance, assuming no correlation between the forced and variability components, is 2T = 2M + 2N, where 2M = {(Tm - {Tm})2} measures the inter-model scatter of the forced component and 2N the natural variability. The common signal is again best discerned where the signal to noise ratio {T} / T is largest.

Figure 9.3 illustrates some basic aspects of the multi-model ensemble approach for global mean temperature and precipitation. Each model result is the sum of a smooth forced signal, Tf, and the accompanying natural variability noise. The natural variability is different for each model and tends to average out so that the ensemble mean estimates the smooth forced signal. The scatter of results about the ensemble mean (measured by the ensemble variance) is an indication of uncertainty in the results and is seen to increase with time. Global mean temperature is seen to be a more robust climate change variable than precipitation in the sense that {T} / T is larger than {P} / P. These results are discussed further in Section 9.3.2.

Table 9.1: The climate change experiments assessed in this report.
Model Number (see Chapter 8, Table 8.1) Model Name and centre in italics (see Chapter 8, Table 8.1) Scenario name Scenario description Number of simulations Length of simulation or starting and final year Transient Climate Response (TCR) (Section 9.2.1) Equilibrium climate sensitivity (Section 9.2.1) (in bold used in Figure. 9.18 / Table 9.4) Effective climate sensitivity (Section 9.2.1) (from CMIP2 yrs 61-80) in bold used in Table A1 References Remarks
2 ARPEGE/OPA2
CERFACS		 CMIP2 1% CO2 1 80 1.64     Barthelet et al., 1998a  
3 BMRCa
BMRC		 ML Equilibrium 2xCO2 in mixed-layer experiment 2 60   2.2   Colman and McAvaney, 1995; Colman, 2001  
CMIP2 1% CO2 1 100 1.63      
5 CCSR/NIES
CCSR/NIES		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 40   3.6   Emori et al., 1999  
CMIP2 1% CO2 1 80 1.8      
G Historical equivalent CO2 to 1990 then 1% CO2 (approx. IS92a) 1 1890-2099        
GS As G but including direct effect of sulphate aerosols 1 1890-2099        
GS2 1% CO2 +direct effect of sulphate aerosols but with explicit representation 1 1890-2099        
31 CCSR/NIES2
CCSR/NIES		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 40   5.1   Nozawa et al., 2001  
CMIP2 1% CO2 1 80 3.1   11.6  
A1 SRES A1 scenario 1 1890-2100        
A2 SRES A2 scenario 1 1890-2100        
B1 SRES B1 scenario 1 1890-2100        
B2 SRES B2 scenario 1 1890-2100        
6 CGCM1
CCCma		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 30   3.5   Boer et al., 1992  
CMIP2 1% CO2 1 80 1.96   3.6 Boer et al., 2000a,b 1,000 yr control
G Historical equivalent CO2 to 1990 then 1% CO2 (approx. IS92a) 1 1900-2100      
GS As G but including direct effect of sulphate aerosols 3 1900-2100      
GS2050 As GS but all forcings stabilised in year 2050 1 1000 after stability      
GS2100 As GS but all forcings stabilised in year 2100 1 1000 after stability      
7 CGCM2
CCCma		 GS Historical equivalent CO2 to 1990 then 1% CO2 (approx. IS92a) and direct effect of sulphate aerosols 3 1900-2100       Flato and Boer, 2001 1,000 yr control
A2 SRES A2 scenario 3 1990-2100      
B2 SRES B2 scenario 3 1990-2100      
10 CSIRO Mk2
CSIRO		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 60   4.3   Watterson et al., 1998  
CMIP2 1% CO2 1 80 2.00   3.7 Gordon and O'Farrell, 1997  
G Historical equivalent CO2 to 1990 then 1% CO2 (approx. IS92a) 1 1881-2100        
G2080 As G but forcing stabilised at 2080 (3x initial CO2) 1 700 after stability       Hirst, 1999  
GS As G +direct effect of sulphate aerosols 1 1881-2100       Gordon and O'Farrell, 1997  
A2 SRES A2 scenario 1 1990-2100          
B2 SRES B2 scenario 1 1990-2100          
11 CSM 1.0
NCAR		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 50   2.1   Meehl et al., 2000a  
CMIP2 1% CO2 1 80 1.43   1.9  
12 CSM 1.3a
NCAR 		GS Historical GHGs +direct effect of sulph- CO2 + direct effect of sulphate aerosols includ- ing effects of pollution control policies ate aerosols to 1990 then BAU 1 1870-2100       Boville et al., 2001; Dai et al., 2001  
GS2150 Historical GHGs +direct effect of except WRE550 scenario for CO2 until it reaches 550 ppm in 2150 sulphate to aerosols to 1990 then as GS 1 1870-2100        
A1 SRES A1 scenario 1 1870-2100        
A2 SRES A2 scenario 1 1870-2100        
B2 SRES B2 scenario 1 1870-2100        
CMIP2 1% CO2 1 100 1.58   2.2  
14 ECHAM3/LSG
DKRZ		 G Historical equiv CO2 to 1990 then 1% CO2 (approx. IS92a) 1 1881-2085       Cubasch et al., 1992, 1994, 1996  
G2050 As G but forcing stabilised at 2050 (2x initial CO2) 1 850 after stability        
G2110 As G but forcing stabilised at 2110 (4x initial CO2) 2 850 after stability       Voss and Mikolajewicz, 2001 Periodically synchronous coupling
GS As G + direct effect of sulphate aerosols 2 1881-2050      
ML Equilibrium 2xCO2 in mixed-layer experiment 1 60   3.2   Cubasch et al., 1992, 1994, 1996b  
15 ECHAM4/OPYC
MPI		 CMIP2 1% CO2 1 80 1.4   2.6 Roeckner et al., 1999  
G Historical GHGs to 1990 then IS92a 1 1860-2099        
GS As G +direct effect of sulphate aerosol interactively calculated 1 1860-2049        
GSIO As GS +indirect effect of sulphate aerosol +ozone 1 1860-2049        
A2 SRES A2 scenario 1 1990-2100       Stendel et al., 2000  
B2 SRES B2 scenario 1 1990-2100        
16 GFDL_R15_a
GFDL		 ML Equilibrium 2xCO2 in mixed-layer experiment 2 40   3.7
(3.9)b		   Manabe et al., 1991 15,000 year control
CMIP2 1% CO2 2 80 2.15   4.2 Stouffer and Manabe, 1999
CMIP270 As CMIP2 but forcing stabilised at year 70 (2x initial CO 2 ) 1 4000   (4.5)c  
CMIP2140 As CMIP2 but forcing stabilised at year 140 (4 x initial CO2) 1 5000      
G Historical equivalent CO2 to 1990 then 1% CO 2 (approximate IS92a) 1 1766-2065       Haywood et al., 1997; Sarmiento et al., 1998
GS As G + direct effect of sulphate aerosols 2 1766-2065      
17 GFDL_R15_b
GFDL		 CMIP2 1% CO2 1 80 Data unavailable        
GS Historical equivalent CO2 to 1990 then 1% CO2 (approximate IS92a) + direct effect of sulphate aerosols 3 1766-2065       Dixon and Lanzante, 1999  
3 1866-2065        
3 1916-2065        
18 GFDL_R30_c
GFDL		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 40   3.4     2 x1,000year control runs with different oceanic dia- pycnal mixing
CMIP2 1% CO2 2 80 1.96      
CMIP270 As CMIP2 but forcing stabilised at year 70 (2 x initial CO2) 1 140 after stability         Different oceanic diapycnal mixing
CMIP2140 As CMIP2 but forcing stabilised at year 140 (4 x initial CO2) 1 160 after stability        
GS 1% CO (approximate IS92a) + direct effect of sulphate aerosols Historical equivalent CO2 to 1990 then 9 1866-2090       Knutson et al., 1999  
A2 SRES A2 scenario 1 1960-2090          
B2 SRES B2 scenario 1 1960-2090          
20 GISS2
GISS		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 40   (3.1) d   Yao and Del Genio, 1999  
CMIP2 1% CO2 1 80 1.45     Russell et al., 1995; Russell and Rind, 1999  
21 GOALS
IAP/LASG		 CMIP2 1% CO2 1 80 1.65        
22 HadCM2
UKMO		 ML Equilibrium 2 xCO2 in mixed-layer experiment 1 40   4.1   Senior and Mitchell, 2000  
CMIP2 1% CO2 1 80 1.7   2.5 Keen and Murphy, 1997 1,000 year control run
CMIP270 As CMIP2 but forcing stabilised at year 70 (2 x initial CO2) 1 900 after stability       Senior and Mitchell, 2000
G Historical equivalent CO2 to 1990 then 1% CO2 (approximate IS92a) 4 1881-2085       Mitchell et al., 1995; Mitchell and Johns, 1997
G2150 As G but all forcings stabilised in year 2150 1 110 after stability       Mitchell et al., 2000
GS As G + direct effect of sulphate aerosols 4 1860-2100       Mitchell et al., 1995; Mitchell and Johns, 1997
23 HadCM3
UKMO		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 30   3.3   Williams et al., 2001  
CMIP2 1% CO2 1 80 2.0   3.0   1,800 year control run
G Historical GHGs to 1990 then IS95a 1 1860-2100       Mitchell et al., 1998; Gregory and Lowe, 2000 Johns et al., 2001
GSIO As G + direct and indirect effect of sulphate aerosols + ozone changes 1 1860-2100      
A2 SRES A2 scenario 1 1990-2100      
B2 SRES B2 scenario 1 1990-2100      
25 IPSL-CM2
IPSL/LMD		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 25   (3.6)e   Ramstein et al., 1998  
CMIP2 1% CO2 1 140 1.96     Barthelet et al., 1998b  
CMIP270 As CMIP2 but forcing stabilised at year 70 (2 x initial CO2) 1 50 after stability        
CMIP2140 As CMIP2 but forcing stabilised at year 140 (4 x initial CO2) 1 60 after stability        
26 MRI1 f
MRI		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 60   4.8   Noda et al., 1999a  
CMIP2 1% CO2 1 150 1.6   2.5 Tokioka et al., 1995, 1996  
CMIP2S As CMIP2 + direct effect of sulphate aerosols 1 100       Japan Met. Agency, 1999  
27 MRI2
MRI		 ML Equilibrium 2xCO2 in mixed-layer experiment 1 50   2.0   Yukimoto et al., 2001; Noda et al., 2001  
CMIP2 1% CO2 1 150 1.1   1.5  
G Historical equivalent CO2 to 1990 then 1% CO2 (approx IS92a) 1 1900-2100        
GS As G + explicit representation of direct effect of sulphate aerosols 1 1900-2100        
A2 SRES A2 scenario 1 1900-2100        
B2 SRES B2 scenario 1 1900-2100        
30 DOE PCM
NCAR		 ML in mixed-layer exp. Equilibrium 2xCO2 1 50   2.1   Washington et al., 2000 Meehl et al., 2001  
CMIP2 1% CO2 5 80 1.27   1.7  
G Historical GHGs +direct effect of sulph- CO2 + direct effect of sulphate aerosols includ- ing effects of pollution control policies ate aerosols to 1990 then BAU   1870-2100        
GS Historical GHGs +direct effect of except WRE550 scenario for CO2 until it reaches 550 ppm in 2150 sulphate to aerosols to 1990 then as GS 5 1870-2100        
GS2150 Historical GHGs to 1990 then as GS except WRE550 scenario for CO2 until it reaches 550 ppm in 2150. 5 1870-2100        
A2 SRES A2 scenario 1 1870-2100        
B2 SRES B2 scenario 1 1870-2100        
a CSM 1.3 was at the time of the printing of this report not archived completely in the DDC. It is therefore not considered in calculations and diagrams refering to the DDC experiments with the exception of Figure 9.5.
b The equilibrium climate sensitivity if the control SSTs from the coupled model are used.
c The equilibrium climate sensitivity calculated from the coupled model.
d The ML experiment used in Table 9.2 for the GISS model were performed with a different atmospheric model to that used in the coupled model listed here.
e The ML experiment used in Table 9.2 for the IPSL-CM2 model were performed with a slightly earlier version of the atmospheric model than that used in the coupled model, but tests have suggested the changes would not affect the equilibrium climate sensitivity.
f Model MRI1 exists in two versions. At the time of writing, more complete assessment data was available for the earlier version, whose control run is in the CMIP1 database. This model is used in Chapter 8. The model used in Chapter 9 has two extra ocean levels and a modified ocean mixing scheme. Its control run is in the CMIP2 database. The equilibrium climate sensitivities and Transient Climate Responses (shown in this table) of the two models are the same.

Other reports in this collection