Technical Comments

Indirect Aerosol Forcing

Science  20 Oct 2000:
Vol. 290, Issue 5491, pp. 407a-407
DOI: 10.1126/science.290.5491.407a

Crowley (1) used estimates of radiative forcing, together with an energy balance model, to estimate the temperature response to variations in volcanic emissions, solar irradiance, increases in greenhouse gases, and aerosol forcing. He did not account for the indirect effects of aerosols, because evaluations by the Intergovernmental Panel for Climate Change (IPCC) indicated that confidence in estimates of this forcing was very low [p. 272 of (1)]. We believe, however, that such an omission could lead to large systematic errors.

Although the IPCC's confidence in indirect aerosol forcing was indeed low, the panel used values of −0.8 Wm−2 in its energy balance model simulations (2). Its estimates for indirect forcing were developed by considering a range of model estimates for the indirect effect. When the IPCC assessment was prepared, such models had considered only the increase in cloud albedo that resulted from increases in droplet concentration (3). Yet the total indirect forcing also includes a second part: the increase in cloud extent and liquid water content associated with reduced precipitation efficiency due to the decrease in cloud droplet radius (4–6). Models evaluating this effect have been subject to question because they have included a component of response as well as the forcing (7). Recent modeling by us (8), however, suggests that the response component of this part of the indirect forcing may be only of order 10% (in the global mean), which in turn argues that the evaluation of forcing for this second indirect effect should be included in estimates of the total indirect forcing.

A number of models (8–12) have suggested that total indirect aerosol forcing—that is, both the first indirect effect related to increased droplet concentration, cited in (2), and the second indirect effect related to decreased droplet radius—varies from 1.3 to 2.2 times the value for the first effect alone. Applying that multiplier to the forcing figure of −0.8 Wm−2 cited in (2), these models thus suggest a total indirect forcing ranging from −1.0 to −1.8 Wm−2. Moreover, the evaluation in (2) did not include indirect forcing by biomass aerosols, so the forcing figure of −0.8 Wm−2 itself may be underestimated, perhaps by as much as a factor of 2.4 (13).

The magnitude of the indirect forcing is difficult to obtain from observations, although such effects are clearly present at scales that appear to be significant (6, 14). It thus seems prudent to include this forcing in climate simulations, such as those of Crowley (1), that are designed to test our understanding of the 20th-century temperature rise. This forcing, if included, would undoubtedly necessitate consideration of larger estimates of the climate sensitivity parameter to fit the instrumental record of temperature change. Considering this forcing could also lead to larger residuals between Crowley's estimates of the forced climate change and the reconstructed temperature time series. Because the magnitude of those residuals is cited as support for the low-frequency climate variability estimated by coupled models, such statements also impact our evaluation of the climate model “noise” that provides the measure upon which studies of climate change detection and attribution rest (15, 16). The use of the full array of climate forcings in such comparison would no doubt point toward the importance of improved evaluations of the indirect forcing. Indeed, an improved evaluation might also include a positive forcing if consideration of changes in ice clouds were included. Thus, we urge those involved in climate simulations, analyses, and assessments to include a range of estimates for indirect forcing.


  1. F1.
  2. F2.
  3. F3.
  4. F4.
  5. F5.
  6. F6.
  7. F7.
  8. F8.
  9. F9.
  10. F10.
  11. F11.
  12. F12.
  13. F13.
  14. F14.
  15. F15.
  16. F16.
  17. F17.
Figure 11

Comparison of model calculations with reconstructed (2) Northern Hemisphere temperature variations (labeled CL2.Jns.11) over the interval from 1600 to 1993, with the instrumental record (8) spliced in after 1860. Figure includes original 11-point smoothed (sm 11) calculation (2) using all forcing terms [volcanism, solar variability, greenhouse gases, and direct aerosol forcing, with the 14C solar index from (9)] and a climate sensitivity of 2.0°C (labeled All.C14Brd 2.0 sm 11). Also shown is a calculation with the same sensitivity and indirect forcing equivalent to −1.4 Wm−2(labeled All.C14Brd/ID 2.0 sm11), as suggested by Penner and Rotstayn. A similar calculation with a sensitivity of 4.5°C for a doubling of CO2 (labeled All.C14Brd/ID 4.5 sm11) still does not simulate the late-20th-century warming. For the 2.0°C-sensitivity runs, the time series have been fit to the pre-1850 (pre-anthropogenic perturbation) part of the record; because of the nature of the response, the higher sensitivity run was fit to the interval from 1860 to1960. (The main conclusion is not sensitive to the interval over which the curves have been fit.)

Response: It is certainly important to constrain the radiative effects of indirect sulfate aerosol forcing, as well as biomass burning and mineral dust. But not all studies (1) estimate as large an effect as do Penner and Rotstayn—and, if the indirect aerosol forcing is indeed as large as they suggest, simply increasing the sensitivity will not resolve the problem.

To illustrate this point, I include their midrange indirect aerosol forcing of −1.4 Wm−2 and, using the same energy balance model employed in (2), compute its time variation as correlative with the direct forcing term. Regardless of whether a sensitivity of 2.0° or 4.5°C is used, the late-20th-century warming cannot be simulated (Fig. 1); however, as Penner and Rotstayn suggest, a high sensitivity causes a significant divergence between the model and paleotemperature data (3). The reason for the lack of late-20th-century warming with higher climate sensitivity is simple: The total direct and indirect forcing from aerosols (maximum estimate of −2.1 Wm−2 in this calculation) is now approximately equivalent to the greenhouse gas forcing (4).

Thus, if indirect forcing is really as large as Penner and Rotstayn maintain, either we cannot simulate the late-20th-century warming, or there are additional compensating terms that provide positive feedbacks. For example, observational studies (5,6) indicate that anthropogenic soot can sometimes reduce cloud amount, leading to a significant increase of incoming radiation that offsets some of the aerosol cooling.

Although it is important to understand the role of indirect aerosol forcing, the principle of parsimony still has merit in attempting to explain the temperature record of the last 1000 years. My approach was to use as few terms as possible, most of which are well constrained. When this is done, the main features of 20th-century warming can be simulated (2), although some offset still occurs between models and data. This result does not prove that a minimalist approach is correct in all particulars, but it does suggest that the approach is justifiable, with the attendant corollaries about coupled-model variability remaining valid. That said, I fully endorse the importance of better constraining the role of aerosol forcing in the climate system, as emphasized in the comment of Penner and Rotstayn.


  1. F1-1.
  2. F1-2.
  3. F1-3.
  4. F1-4.
  5. F1-5.
  6. F1-6.
  7. F1-7.
  8. F1-8.
  9. F1-9.

Related Content

Navigate This Article