Stronger Constraints on the Anthropogenic Indirect Aerosol Effect

See allHide authors and affiliations

Science  01 Nov 2002:
Vol. 298, Issue 5595, pp. 1012-1015
DOI: 10.1126/science.1075405


The anthropogenic indirect aerosol effects of modifying cloud albedo and cloud lifetime cannot be deduced from observations alone but require a modeling component. Here we validate a climate model, with and without indirect aerosol effects, by using satellite observations. The model agrees better with observations when both indirect aerosol effects are included. However, the simulated clouds are more susceptible to aerosols than the observed clouds from the POLDER satellite, thus overestimating the indirect aerosol effect. By taking the difference in susceptibilities into account, the global mean total anthropogenic aerosol effect is reduced from –1.4 to –0.85 watts per square meter.

The anthropogenic component of sulfate and carbonaceous aerosols has substantially increased the global mean aerosol burden from preindustrial times to the present and can influence climate in different ways. The direct aerosol effect is caused by the absorption and scattering of solar radiation. Additionally, aerosols exert an indirect effect by acting as cloud condensation nuclei, thereby affecting the initial cloud droplet number concentration (CDNC), albedo, precipitation formation, and lifetime of warm clouds. For a constant liquid water path, a higher cloud droplet number causes an increase in cloud albedo (cloud albedo effect). Reductions in precipitation efficiency due to more but smaller cloud droplets slow down precipitation formation and increase cloud lifetime (cloud lifetime effect). The cooling from both indirect effects has been estimated by climate models to be –1 to –4.4 W m−2in the global mean (1–5), but this estimate is still very poorly constrained and is an important source of uncertainty in projections of future climate change (6,7).

Data from a regional chemical transport model, together with satellite data estimating the cloud albedo effect in the North Atlantic, show that the cloud-top spherical albedo was enhanced over two-week episodes by 0.02 to 0.15 for the same liquid water path distribution (8) relative to the unperturbed case. POLDER satellite data were used (9) to derive aerosol concentration and cloud droplet effective radii (CDR) from 8 months of space-borne measurements and to explore the effect of aerosols on cloud microphysics. It was found that the cloud droplet size decreases with increasing aerosol index (AI) (10), which is representative of the aerosol column number concentration, indicating that the effect of aerosols on cloud microphysics is significant and occurs on a global scale. However, these data alone are not sufficient to quantify the magnitude of the global indirect aerosol effects between preindustrial times and present day.

From historical climate record data of oceanic and atmospheric warming together with ensembles of simulations with one climate model of reduced complexity, the anthropogenic indirect aerosol effects have recently been constrained within the range of 0 to –1.2 W m−2 (11). Here, a complex global climate model rather than a simple climate model is used to try to determine the importance of the cloud albedo and the cloud lifetime effects by finding the model configurations that produce results that most resemble the observational data (9) in order to estimate the anthropogenic aerosol effects constrained by POLDER data.

We used the ECHAM4 general circulation model (GCM) (12) in T30 horizontal resolution to estimate the anthropogenic aerosol effect on a global scale. The GCM includes a fully coupled aerosol-cloud microphysics module (3, 13,14). The reference simulation, ECHAM-CTL, includes both indirect aerosol effects with the use of present-day emissions (15). To turn off the cloud albedo effect (simulation ECHAM-2ND-AIE), we prescribed CDNC everywhere as a function of height in the radiation calculation: CDNC equals 150 cm−3 near the surface and decreases to 50 cm−3 in the midtroposphere. In the simulation ECHAM-NO-AIE, a constant number of cloud droplets was used in the cloud microphysics and the radiation calculations (16). We also ran a preindustrial climate simulation, ECHAM-PI, in which the sulfate and carbonaceous (black carbon and organic carbon) aerosols from fossil fuel and biomass burning were set to zero (17), leaving natural emissions from forests as the only source for organic carbon, and dimethyl sulfide emissions from the ocean and volcanoes as the only sources for sulfate aerosols.

Figure 1 shows AI as obtained from the control simulation as well as the difference between ECHAM's AI and that obtained from POLDER and averaged over March, April, and May 1997 (9, 18, 19). We calculated AI offline with the use of Mie theory, assuming an externally mixed aerosol (20). ECHAM correctly simulates a large land-sea contrast in AI, with the larger AI over land resulting from pollution as observed from POLDER. ECHAM's AI agrees with that derived from POLDER over the oceans except for the transported Saharan dust. However, ECHAM largely underestimates AI over land, probably caused by too-large aerosol mode radii (21). The land-sea contrast in cloud droplet effective radius at the top of warm clouds, with cloud-top temperatures above 273.2 K and with smaller droplets over land, is also reproduced (22). However, ECHAM underestimates CDR, especially over continents of the Northern Hemisphere. Although ECHAM's liquid water path is in reasonable agreement with microwave observations (23), which are available only over oceans, its vertically integrated CDNC exceeds that derived from ISCCP satellite observations (24) in most parts of the world (not shown).

Figure 1

AI and effective cloud droplet radius [in (μm), indicated by color scale] averaged for March, April, and May 1997 from the ECHAM-CTL model simulation, and their differences between ECHAM-CTL and POLDER results.

The question to be addressed is whether anthropogenic aerosol indirect effects are needed to explain the observed anticorrelation between AI and CDR, especially the larger negative slope over oceans, or whether these anticorrelations are determined purely by geographic variations in the liquid water path. ECHAM-CTL reproduces the observed anticorrelation between AI and CDR with a larger slope over oceans (Fig. 2). The slopes and linear correlation coefficients were computed over the same AI range (0 to 0.15) and covered the same geographical region between 60°N and 45°S as in (9). Whereas (9) used back trajectories to find the AI that belongs to a given CDR, we assumed that on a monthly basis AI and CDR can be taken from the same grid boxes in the model.

Figure 2

Cloud droplet effective radius as a function of AI over oceans (diamonds) and over land (squares) from the POLDER satellite retrieval and ECHAM-CTL, ECHAM-2ND-AIE, ECHAM-NO-AIE, and ECHAM-PI model simulations. Straight lines are least-square fits for AI < 0.15. Error bars for observations and ECHAM refer to ±1 standard error. The horizontal scale is reduced for ECHAM-PI.

In the observations and in the model, the linear correlation coefficients between AI and CDR are –0.9 over ocean and even higher over land. The larger oceanic slope, especially for low values of AI, is attributed to more susceptible clouds over the ocean. That is, for a similar AI over remote continental areas as over remote oceans, the continental cloud has a smaller liquid water path, leading to smaller CDR and thus smaller decreases of CDR with increasing AI, because AI values over land and ocean converge toward the same CDR limit in the POLDER data. A smaller CDR value is reached at high AI values for continental clouds in ECHAM, probably because of an overestimate in CDNC and an underestimated liquid water path in summer, when the continents tend to dry out.

The difference in cloud susceptibility over land and ocean is confirmed by the preindustrial simulation, ECHAM-PI. In this simulation, AI never exceeds 0.04 and, thus, the slope calculations are limited to that range. The oceanic clouds start with a higher CDR at the smallest AI but CDR rapidly decreases with increasing AI, whereas the CDR over land is almost constant with AI reflecting the different distribution of liquid water path over land and ocean (Fig. 3).

Figure 3

Liquid water path as a function of AI over oceans (diamonds) and over land (squares) from the ECHAM-CTL, ECHAM-2ND-AIE, ECHAM-NO-AIE, and ECHAM-PI model simulations. The AI scale is different for ECHAM-PI.

If no indirect aerosol effect is taken into account as in ECHAM-NO-AIE, then the oceanic liquid water path decreases with increasing AI, thus increasing CDR for low AI and increasing its slope in worse agreement with observations. If only the cloud lifetime effect is considered as in ECHAM-2ND-AIE, then the difference in slope between oceanic and continental clouds increases, in better agreement with observations. Only the maritime clouds in ECHAM-NO-AIE show an anticorrelation between liquid water path and AI (Fig. 3). Again, the smallest absolute values in slope and correlation coefficient for oceanic clouds in ECHAM-CTL are most in-line with recent findings from satellite data over oceans (25) showing no systematic trend of liquid water path on column aerosol number over the full range of column aerosol number concentration. Thus, we conclude that both aerosol indirect effects, including the more contentious cloud lifetime effect, are needed to reproduce the observed variation of CDR with AI and of AI with liquid water path.

Aerosol indirect effects exist even in the preindustrial simulation (Figs. 2 and 3). With increasing AI, the cloud droplet effective radius decreases and the liquid water path increases, especially in oceanic clouds. This finding complicates any attempt to detect an anthropogenic aerosol indirect effect from observational data alone and demonstrates the importance of using models together with observations.

After showing evidence for the existence of both aerosol indirect effects, we used the climate model simulations including both indirect effects to estimate the global mean aerosol effects from preindustrial to present times. We obtain a decrease in shortwave radiation at the top of the atmosphere of 1.8 W m−2, which is accompanied by a 0.4-W m−2 increase in longwave radiation, such that the net radiation is reduced by 1.4 W m−2 in the global annual mean. However, the CDR-versus-AI plots showed that ECHAM's clouds are more susceptible to aerosols than the observed clouds, so ECHAM very likely overestimates the indirect aerosol effect. If CDR was constant with AI, then the indirect aerosol effect would be zero. Thus, we use the difference in simulated versus observed slope over land and ocean to scale our estimate of the aerosol effect. ECHAM simulates an aerosol effect of –1.28 W m−2 over oceans. Taking the factor of 1.3 in the overestimate in oceanic slope into account, the aerosol effect over oceans is reduced to –0.98 W m−2. Likewise, this method reduces the aerosol effect over land from –1.62 to –0.53 W m−2, resulting in a net global mean aerosol effect of –0.85 W m−2. This new estimate of the net aerosol effects is consistent with the range of 0 to –1.2 W m−2 estimated from historical climate record data coupled with a simple climate model for the indirect aerosol effects alone (11) and is much smaller than the purely model-based estimates (1, 2, 4).

  • * To whom correspondence should be addressed. E-mail: Ulrike.Lohmann{at}Dal.Ca


View Abstract

Stay Connected to Science

Navigate This Article