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

2.4.5.2 Estimates of the Radiative Forcing from Models

General Circulation Models constitute an important and useful tool to estimate the global mean RF associated with the cloud albedo effect of anthropogenic aerosols. The model estimates of the changes in cloud reflectance are based on forward calculations, considering emissions of anthropogenic primary particles and secondary particle production from anthropogenic gases. Since the TAR, the cloud albedo effect has been estimated in a more systematic and rigorous way (allowing, for example, for the relaxation of the fixed LWC criterion), and more modelling results are now available. Most climate models use parametrizations to relate the cloud droplet number concentration to aerosol concentration; these vary in complexity from simple empirical fits to more physically based relationships. Some models are run under an increasing greenhouse gas concentration scenario and include estimates of present-day aerosol loadings (including primary and secondary aerosol production from anthropogenic sources). These global modelling studies (Table 2.7) have a limitation arising from the underlying uncertainties in aerosol emissions (e.g., emission rates of primary particles and of secondary particle precursors). Another limitation is the inability to perform a meaningful comparison between the various model results owing to differing formulations of relationships between aerosol particle concentrations and cloud droplet or ice crystal populations; this, in turn, yields differences in the impact of microphysical changes on the optical properties of clouds. Further, even when the relationships used in different models are similar, there are noticeable differences in the spatial distributions of the simulated low-level clouds. Individual models’ physics have undergone considerable evolution, and it is difficult to clearly identify all the changes in the models as they have evolved. While GCMs have other well-known limitations, such as coarse spatial resolution, inaccurate representation of convection and hence updraft velocities leading to aerosol activation and cloud formation processes, and microphysical parametrizations, they nevertheless remain an essential tool for quantifying the global cloud albedo effect. In Table 2.7, differences in the treatment of the aerosol mixtures (internal or external, with the latter being the more frequently employed method) are noted. Case studies of droplet activation indicate a clear sensitivity to the aerosol composition (McFiggans et al., 2006); additionally, radiative transfer is sensitive to the aerosol composition and the insoluble fraction present in the cloud droplets.

Table 2.7. Published model studies of the RF due to cloud albedo effect, in the context of liquid water clouds, with a listing of the relevant modelling details.

Model  Model typea  Aerosol speciesb  Aerosol mixturesc  Cloud types included  Microphysics  Radiative Forcing(W m–2)d 
Lohmann et al. (2000)  AGCM + sulphur cycle (ECHAM4)  S, OC, BC, SS, D  I  warm and mixed phase  

Droplet number concentration and LWC, Beheng (1994); Sundqvist et al. (1989). Also, mass and number from field observations

 

–1.1 (total)

–0.45 (albedo)

 
E  

–1.5 (total)

 
Jones et al. (2001)  AGCM + sulphur cycle, fixed SST (Hadley)  S, SS, D (a crude attempt for D over land, no radiation)  E  stratiform and shallow cumulus  

Droplet number concentration and LWC, Wilson and Ballard (1999); Smith (1990); Tripoli and Cotton (1980); Bower et al. (1994). Warm and mixed phase, radiativetreatment of anvil cirrus, non-spherical ice particles

 

–1.89 (total)

–1.34 (albedo)

 
Williams et al. (2001b)  GCM with slab ocean + sulphur cycle (Hadley)  S, SS  E  stratiform and shallow cumulus  

Jones et al. (2001)

 

–1.69 (total)

–1.37 (albedo)

 
AGCM, fixed SST  

–1.62 (total)

–1.43(albedo)

 
Rotstayn and Penner (2001)  AGCM (CSIRO), fixed SST and sulphur loading  S  n.a.  warm and mixed phase  

Rotstayn (1997); Rotstayn et al. (2000)

 

–1.39 (albedo)

 
Rotstayn and Liu (2003) value decreased) Interactive sulphur cycle  

Inclusion of dispersion

 

12 to 35% decrease

–1.12 (albedo, mid value decreased)

 
Ghan et al. (2001)  AGCM (PNNL) + chemistry (MIRAGE), fixed SST  S, OC, BC, SS, N, D  E (for different modes); I (within modes)  warm and mixed phase  

Droplet number concentration and LWC, crystal concentration and ice water content. Different processes affecting the various modes

 

–1.7 (total)

–0.85 (albedo)

 
Chuang et al. (2002)  CCM1 (NCAR) + chemistry (GRANTOUR), fixed SST  S, OC, BC, SS, D  E (for emitted particles); I: when growingby condensation  warm and mixed phase  

Modified from Chuang and Penner (1995), no collision/coalescence

 

–1.85 (albedo)

 
Menon et al. (2002a)  GCM (GISS) + sulphur cycle, fixed SST  S,OC, SS  E  warm  

Droplet number concentration and LWC, Del Genio et al. (1996), Sundqvist et al. (1989). Warm and mixed phase, improved vertical distribution of clouds (but only ninelayers).Global aerosol burdens poorly constrained

 

–2.41 (total)

–1.55 (albedo)

 
Kristjansson (2002)  CCM3 (NCAR) fixed SST  S, OC, BC, SS, D  E (for nucleation mode and fossil fuel BC); I (foraccumulationmode)  warm and mixed phase  

Rasch and Kristjánsson (1998). Stratiform and detraining convective clouds

 

–1.82 (total)

–1.35 (albedo)

 
Suzuki et al. (2004)  AGCM (Japan), fixed SST  S, OC, BC, SS E  stratiform  

Berry(1967), Sundqvist(1978)

 

Errata
0.54 (albedo)

 
Quaas et al. (2004)  AGCM (LMDZ) + interactive sulphur cycle, fixed SST  S  n.a.  warm and mixed phase  

Aerosol mass and cloud droplet number concentration, Boucher and Lohmann (1995); Boucher et al. (1995)

 

–1.3 (albedo)

 
Hansen et al. (2005) shallow (below  GCM (GISS) + 3 different ocean parametrizations  S, OC, BC, SS, N, D (D not included in clouds)  E  warm and 720hPa)  

Schmidt et al. (2005), 20 vertical layers. Droplet number concentration (Menon and Del Genio, 2007)

 

–0.77 (albedo)

 
Kristjansson et al. (2005)  CCM3 (NCAR) + sulphur and carbon cycles slab ocean  S, OC, BC, SS, D  E (for nucleation mode and fossil fuel BC); I (foraccumulationmode)  warm and mixed phase  

Kristjansson (2002). Stratiform and detraining convective clouds

 

–1.15 (total,at the surface)

 
Quaas and Boucher (2005)  AGCM (LMDZ) + interactive sulphur cycle, fixed SST  S, OC, BC, SS, D  E  warm and mixed phase  

Aerosol mass and cloud droplet number concentration, Boucher and Lohmann (1995); Boucher et al. (1995)control run

 

–0.9 (albedo)

 

fit to POLDER data

fit to MODIS data

 

–0.5 (albedo)e

–0.3 (albedo)e

 
Quaas et al. (2005)  AGCM (LMDZ and ECHAM4)  S, OC, BC, SS, D  E  warm and mixed phase  

Aerosol mass and cloud droplet number concentration, Boucher and Lohmann, (1995), control runs (ctl)

 

–0.84 (total LMDZ-ctl)

–1.54 (total(ECHAM4-ctl)

 

Aerosol mass and cloud droplet number concentration fitted to MODIS data

 

–0.53 (total LMDZ)e

–0.29 (total(ECHAM4)e

 
Dufresne et al. (2005)  AGCM (LMDZ) + interactive sulphur cycle, fixed SST  S  n.a.  warm  

Aerosol mass and cloud droplet number concentration, Boucher and Lohmann, (1995), fitted to POLDER data

 

–0.22 (albedo)e

 
Takemura et al. (2005)  AGCM (SPRINTARS) + slab ocean  S, OC, BC, SS, D  E (50% BC from fossil fuel); I (for OC and BC)  warm  

Activation based on Kohler theory and updraft velocity

 

–0.94 (total)

–0.52 (albedo)

 
Chen and Penner (2005)  AGCM (UM) + fixed SST  S, SS, D, OC, BC  I  warm and mixed phase  

Aerosol mass and cloud droplet number concentration(lognormal)Control (Abdul-Razzak and Ghan, 2002)

 

–1.30 (albedo,UM_ctrl)f

 

Relationship between droplet concentration and dispersion coefficient: High

 

–0.75 (albedo,UM_1)f

 

Relationship between droplet concentration and dispersion coefficient: Medium Updraft velocity

 

–0.86 (albedo,UM_2)f

–1.07 (albedo,UM_3)f

 

Relationship between droplet concentration and dispersion coefficient: Low Chuang et al. (1997)

 

–1.10 (albedo, UM_4)f –1.29 (albedo,UM_5)f

 

Nenes and Seinfeld (2003)

 

–1.79 (albedo,UM_6)f

 
Ming et al. (2005b)  AGCM (GFDL), fixed SST and sulphur loading  S  n.a.  warm  

Rotstayn et al. (2000), Khainroutdinov and Kogan (2000). Aerosols off-line

 

–2.3 (total)

–1.4 (albedo)

 
Penner et al. (2006) results from experiment 1  LMDZ, Oslo and CCSR  S, SS, D, OC, BC  E  warm and mixed phase  

Aerosol mass and cloud droplet number concentration; Boucher and Lohmann, (1995); Chen and Penner (2005); Sundqvist (1978)

 

–0.65 (albedo Oslo)

–0.68 (albedo LMDZ)

–0.74 (albedo CCSR)

 

Notes:

a AGCM: Atmospheric GCM; SST: sea surface temperature; CSIRO: Commonwealth Scientific and Industrial Research Organisation; MIRAGE: Model for Integrated Research on Atmospheric Global Exchanges; GRANTOUR: Global Aerosol Transport and Removal model; GFDL: Geophysical Fluid Dynamics Laboratory; CCSR: Centre for Climate System Research; see Table 2.4, Note (a) for listing of other models and modelling centres listed in this column.

b S: sulphate; SS: sea salt; D: mineral dust; BC: black carbon; OC: organic carbon; N: nitrate.

c E: external mixtures; I: internal mixtures.

d Only the bold numbers were used to construct Figure 2.16. Errata

e These simulations have been constrained by satellite observations, using the same empirical fit to relate aerosol mass and cloud droplet number concentration.

f The notation UM corresponds to University of Michigan, as listed in Figure 2.14.

All models estimate a negative global mean RF associated with the cloud albedo effect, with the range of model results varying widely, from –0.22 to –1.85 W m–2. There are considerable differences in the treatment of aerosol, cloud processes and aerosol-cloud interaction processes in these models. Several models include an interactive sulphur cycle and anthropogenic aerosol particles composed of sulphate, as well as naturally produced sea salt, dust and continuously outgassing volcanic sulphate aerosols. Lohmann et al. (2000) and Chuang et al. (2002) included internally mixed sulphate, black and organic carbon, sea salt and dust aerosols, resulting in the most negative estimate of the cloud albedo indirect effect. Takemura et al. (2005) used a global aerosol transport-radiation model coupled to a GCM to estimate the direct and indirect effects of aerosols and their associated RF. The model includes a microphysical parametrization to diagnose the cloud droplet number concentration using Köhler theory, which depends on the aerosol particle number concentration, updraft velocity, size distributions and chemical properties of each aerosol species. The results indicate a global decrease in cloud droplet effective radius caused by anthropogenic aerosols, with the global mean RF calculated to be –0.52 W m–2; the land and oceanic contributions are –1.14 and –0.28 W m–2, respectively. Other modelling results also indicate that the mean RF due to the cloud albedo effect is on average somewhat larger over land than over oceans; over oceans there is a more consistent response from the different models, resulting in a smaller inter-model variability (Lohmann and Feichter, 2005).

Chen and Penner (2005), by systematically varying parameters, obtained a less negative RF when the in-cloud updraft velocity was made to depend on the turbulent kinetic energy. Incorporating other cloud nucleation schemes, for example, changing from Abdul-Razzak and Ghan (2002) to the Chuang et al. (1997) parametrization resulted in no RF change, while changing to the Nenes and Seinfeld (2003) parametrization made the RF more negative. Rotstayn and Liu (2003) found a 12 to 35% decrease in the RF when the size dispersion effect was included in the case of sulphate particles. Chen and Penner (2005) further explored the range of parameters used in Rotstayn and Liu (2003) and found the RF to be generally less negative than in the standard integration.

A model intercomparison study (Penner et al., 2006) examined the differences in cloud albedo effect between models through a series of controlled experiments that allowed examination of the uncertainties. This study presented results from three models, which were run with prescribed aerosol mass-number concentration (from Boucher and Lohmann, 1995), aerosol field (from Chen and Penner, 2005) and precipitation efficiency (from Sundqvist, 1978). The cloud albedo RFs in the three models do not vary widely: –0.65, –0.68 and –0.74 W m–2, respectively. Nevertheless, changes in the autoconversion scheme led to a differing response of the LWP between the models, and this is identified as an uncertainty.

A closer inspection of the treatment of aerosol species in the models leads to a broad separation of the results into two groups: models with only a few aerosol species and those that include a more complex mixture of aerosols of different composition. Thus, in Figure 2.14, RF results are grouped according to the type of aerosol species included in the simulations. In the top panel of Figure 2.14, which shows estimates from models that mainly include anthropogenic sulphate, there is an indication that the results are converging, even though the range of models comes from studies published between 2001 and 2006. These studies show much less scatter than in the TAR, with a mean and standard deviation of –1.37 ± 0.14 W m–2. In contrast, in the bottom panel of Figure 2.14, which shows the studies that include more species, a much larger variability is found. These latter models (see Table 2.7) include ‘state of the art’ parametrizations of droplet activation for a variety of aerosols, and include both internal and external mixtures.

2.14

Figure 2.14. Radiative forcing due to the cloud albedo effect, in the context of liquid water clouds, from the global climate models that appear in Table 2.7. The labels next to the bars correspond to the published study; the notes of Table 2.7 explain the species abbreviations listed on the left hand side. Top panel: results for models that consider a limited number of species, primarily anthropogenic sulphate (S). Bottom panel: results from studies that include a variety of aerosol compositions and mixtures; the estimates here cover a larger range than those in the top panel. Chen and Penner (2005) presented a sensitivity study obtained by changing parametrizations in their model, so the results can be considered independent and are thus listed separately. Penner et al. (2006) is an intercomparison study, so the results of the individual models are listed separately.

Some studies have commented on inconsistencies between some of the earlier estimates of the cloud albedo RF from forward and inverse calculations (Anderson et al., 2003). Notwithstanding the fact that these two streams of calculations rely on very different formulations, the results here appear to be within range of the estimates from inverse calculations.