Scientists have confidence in a result to the extent that it can be derived by different investigators. Their confidence is increased if different techniques lead to the same conclusion. Concurrence provides evidence that the conclusion does not depend upon assumptions that occasionally are insufficiently supported. In contrast, two articles published last December on the same day arrive at very different and incompatible estimates of the effect of human-made aerosols on the radiative budget of the planet (Bellouin et al., 2005; Chung et al., 2005). They follow an earlier estimate published last year, (which included Dorothy as a co-author) that was in the middle (Yu et al., 2005). Aerosols are important to climate partly because their concentration is increased by the same industrial processes that increase the atmospheric concentration of greenhouse gases; yet aerosols generally oppose greenhouse warming. Because aerosols cause respiratory and other health problems and acid rain, they have been regulated more aggressively than greenhouse gases. Concentrations of some aerosols have decreased over the United States and Europe in recent decades as a result of environmental laws, although an increase has been observed in many thrid world regions, where economic development is a priority. In the twenty-first century, aerosol levels are anticipated to drop faster than greenhouse gases in response to future emission reductions, which will leave greenhouse warming unopposed and unmoderated.
Each published calculation of aerosol radiative forcing was a tour de force for integrating a wide variety of measurements ranging from absorption of radiation by individual particles to satellite estimates of aerosol amount. The disparate results emphasize the complexity and difficulty of the calculation. But let’s start at the beginning….
Aerosols are solid particles or liquid droplets that are temporarily suspended within the atmosphere. Naturally occurring examples are sea spray or sulfate droplets, along with soil particles (dust) eroded by the wind. During the twentieth century, natural sources of sulfate aerosols were overwhelmed by the contribution from pollution, in particular from the burning of fossil fuels. The number of soot particles in the atmosphere was increased by industry and the burning of forests to clear land for agriculture. Sulfate aerosols are reflective and act to cool the planet. Soot particles are also reflective, but can absorb sunlight and cause warming. Soot production is greater if combustion occurs at low temperatures, as with cooking fires or inefficient power generation. Aerosols also scatter longwave radiation, although this is significant only for larger aerosols like soil dust, and is neglected by all three of the studies discussed here.
In addition to their ability to scatter radiation and change the net energy gain at the top of the atmosphere (the ‘direct’ effect), aerosols modify the reflectance and lifetime of clouds (the ‘indirect’ radiative effects). Aerosols act as nuclei for the condensation of water vapor, resulting in the distribution of water over a larger number of cloud droplets compared to condensation in clean air. This increases the cloud’s ability to reflect sunlight, while increasing the number of droplet collisions required to form a raindrop large enough to fall out of the cloud, effectively increasing the cloud lifetime. Observations and models provide a weaker constraint upon the size of the indirect effects, so the studies discussed here confine themselves to calculating only the direct radiative effect of anthropogenic aerosols.
According to the latest (2001) IPCC report, direct radiative forcing by anthropogenic aerosols cools the planet, but the forcing magnitude is highly uncertain, with a global, annual average between -0.35 and -1.35 W/m2 at the top of the atmosphere (TOA). The uncertainty of the total indirect effect is even larger. Aerosols eventually fall out of the atmosphere or are washed out by rainfall. The smaller particles having the largest radiative effect typically reside in the atmosphere for only a few days to a few weeks. This time is too short for them to be mixed uniformly throughout the globe (unlike CO2), so there are large regional variations in aerosol radiative forcing, with the largest effects predictably downwind of industrial centers like the east coast of North America, Europe, and East Asia. Consequently, aerosol effects upon climate are larger in particular regions, where they are key to understanding twentieth century climate change.
Aerosol concentrations have been measured downwind of sources over the past few decades, but the number of observing sites is limited and the analysis is laborious. Since the late 1970’s, satellite instruments have detected aerosols routinely with nearly global coverage. However, only the combined effect of all aerosols upon radiation impinging upon the satellite was originally measured. The original instruments couldn’t distinguish between dust and sulfate aerosols where both were present, over the Mediterranean or East Asia, for example. Recent instruments, like the Moderate Resolution Imaging Spectroradiometer (MODIS) measure radiation at multiple wavelengths. This allows particle size to be distinguished with greater confidence, which can be used with some assumptions to infer the aerosol species.
The new generation of satellite instruments is at the heart of recent attempts to reduce the large uncertainty of direct radiative forcing by aerosols. Each of these studies provides an estimate of the most likely value, along with a range of uncertainty. Bellouin et al. (2005) in Nature arrive at TOA forcing of -0.8 ± 0.1 W/m2. While near the center of the range published by the IPCC, this estimate is noteworthy for its comparatively small uncertainty. Yet on the same day, Chung et al. (2005) published an article in the JGR, estimating based upon similarly extensive calculations that the forcing by aerosols at TOA is -0.35 ± 0.25 W/m2. A few months earlier, Yu et al. (2005) had estimated a more conciliatory value of -0.5 ± 0.33 W/m2. The wide range of estimates give some indication the difficulty of the problem.
Forcing estimates differ not only at TOA but also at the surface: Bellouin et al. predict that aerosols reduce the net radiation incident upon the surface by 1.9 ± 0.2 W/m2 compared to 3.4 ± 0.1 W/m2 for Chung et al. (2005). That is, Chung et al. estimate much greater atmospheric absorption. Because radiation into the surface is mainly balanced by evaporation, except within extremely arid regions, the discrepancy has implications for the supply of moisture to the atmosphere. Chung et al. estimate a much larger reduction in global rainfall by aerosols.
What are the sources of disagreement and uncertainty? Ideally, one would know the three-dimensional distribution of each aerosol species and its evolution throughout the year. One would also be able to distinguish natural and human fractions of each species. For sulfate aerosols, this means distinguishing droplets created by industrial sources, compared to biogenic sources. In addition, the ability of each particle to scatter radiation would be known as a function of its age and aggregation with other species (in the way that dust can be coated with sulfates when passing over industrial areas, for example). Many of these processes are included in aerosol models, but some of the key parameters are uncertain given limited observations.
Bellouin et al. attempt an empirical end-run around this uncertainty by dividing the planet into six regions where aerosol concentration is high, and using a ‘typical’ value of particle absorption based on surface measurements. The measured absorption is a single value that reflects the combined effect of both anthropogenic and natural aerosols, although the six representative sites were chosen where contribution by the former dominates. Regions with a preponderance of sulfates, such as the eastern coast of North America and downwind, were assigned greater reflectance and lesser absorption than particles over the Indian Ocean where dark soot particles are more common. This is based upon contrasting surface measurements at Washington DC and the Maldive Islands in the Indian Ocean. The total aerosol mass was inferred from MODIS estimates of the aerosol optical thickness (AOT), which measures attenuation of a light beam passing through an aerosol layer. To estimate the anthropogenic fraction of aerosols, Bellouin et al. made use of the fact that anthropogenic aerosols such as sulfate and soot are generally smaller than natural aerosols such as soil dust and sea salt. MODIS provides not only the total AOT but also the fractional contribution corresponding to smaller particles whose diameter is less than one micron (a thousandth of a millimeter). Bellouin et al. attributed the total AOT to human influence in regions where the fine fraction AOT exceeds 85% of the total. Conversely, regions where larger particles make the predominant contribution to AOT were excluded from the anthropogenic total. While MODIS is able to make this distinction between small and large particles over ocean, the distinction is more uncertain over land, and here Bellouin et al. resorted to the anthropogenic fraction computed by five aerosol models, a number chosen to reduce the uncertainty associated with any single model.
Despite their different result compared to Bellouin et al., the calculations by Chung et al. and Yu et al. are similar. Chung et al. assign the total AOT using MODIS, and adjust this value using local measurements by the AERONET array of sun photometers. (These instruments point toward the sun and record incident radiation at various wavelengths.) The main difference is that Chung et al. compute the anthropogenic fraction over both land and ocean using a single aerosol model, and they use this model along with AERONET measurements to specify the radiative properties of the combined aerosol population within each column. Consequently, these properties vary within each region as opposed to the regionally averaged values used by Bellouin et al. based upon a single putatively representative site. Yu et al. use an even broader array of measurements and models.
Why do similar methods result in forcing estimates whose uncertainty ranges don’t overlap? This is difficult to know, although here we speculate upon the effect of some of the differing assumptions. Chung et al. specify greater particle absorption compared to all but one of the six regional values used by Bellouin et al. Because the TOA forcing becomes less negative as absorption increases, this accounts for some of the difference. Similarly, Chung et al.’s replacement of their model estimate of anthropogenic particle fraction over the ocean with the MODIS estimate (following Bellouin et al.) narrows the difference.
Treatment of aerosol forcing over cloudy regions also contributes to the difference. Both studies estimate nearly identical forcing at the surface in the absence of clouds. While aerosol absorption and reflection have opposing effects at TOA, they both reduce sunlight beneath the aerosol layer, contributing to negative forcing at the surface. Thus, forcing at the surface is less sensitive to the relative strength of absorption versus reflection. When cloudy regions are included, Chung et al. calculate a much larger reduction of surface radiation than Bellouin et al., who assume that aerosol forcing in these regions is zero. At TOA, Chung et al. calculate positive aerosol forcing within cloudy regions, accounting for some of the global disagreement with Bellouin et al. TOA forcing depends strongly upon the relative position of the cloud and aerosol layer. An absorbing soot layer above a bright cloud absorbs more radiation than if the layer were beneath the cloud. Unlike AOT, the vertical distribution of aerosols is not measured routinely, and is comparatively uncertain.
The disagreement among forcing estimates raises the more general point of whether any study really captures the full range of uncertainty. The number of calculations needed to sample the uncertainty can increase exponentially with the number of uncertain parameters. While parametric uncertainty is straightforward to estimate, the dearth of observations makes it difficult to estimate the effect of assuming a bulk absorption that represents an ‘average’ aerosol rather than computing absorption by each species separately. The latter is an example of a structural uncertainty that is typically difficult to characterize. Given the difficulty of measuring the aerosol mass over the entire planet, along with myriad aspects of the aerosol life cycle that are poorly measured and impossible to model precisely, the most reliable estimate of forcing uncertainty may be derived by combining the central forcing estimate from a number of studies, as opposed to taking the uncertainty range of any single study. Yu et al. seem to acknowledge the large outstanding uncertainty by relegating their estimate of anthropogenic aerosol forcing to a table, rather than highlighting it in the abstract or conclusions.
Progress will come by more systematic comparisons among studies to identify key uncertainties. The unambiguous distinction between individual aerosol species within models will eventually become possible by direct observation as a result of more discerning instruments. Nonetheless, models will remain valuable for their ability to distinguish natural and anthropogenic sources of the same aerosol species. While Bellouin et al. assume that all soot particles over the ocean are anthropogenic, naturally occurring forest fires contribute as well. As consensus emerges regarding the global aerosol forcing, attention will turn to regional values that cause local changes to climate and heat redistribution by the atmosphere. Because of the added complexity of cloud physics, the aerosol indirect effect may be even more resistant to consensus. Aerosol forcing remains a crucial problem because its offset of greenhouse warming is expected to decrease with time as governments address the health problems associated with aerosols. Because of their comparatively short lifetimes, the concentration of aerosols decreases much faster than that of CO2 given a reduction in fossil fuel use. Regardless of the absolute amount of the forcing, future reductions in aerosol emissions will be a positive forcing, amplyfying the warming effects of increasing greenhouse gases.
Bellouin, N., O. Boucher, J.Haywood and M. S. Reddy. Global estimate of aerosol direct radiative forcing from satellite measurements, Nature, 438, 1138-1141 (22 December 2005) | doi:10.1038/nature04348 (pdf)
Chung, C. E., V. Ramanathan, D. Kim and I. Podgorny, 2005: Global anthropogenic aerosol direct forcing derived from satellite and ground-based observations. J. Geophys. Res., 110, D24207, doi:10.1029/2005JD006356. (pdf)
Yu, H., Y.J. Kaufman, M. Chin, G. Feingold, L.A. Remer, T.L. Anderson, Y. Balkanski, N. Bellouin, O. Boucher, S. Christopher, P. DeCola, R. Kahn, D. Koch, N. Loeb, M.S. Reddy, M. Schulz, T. Takemura, and M. Zhou 2006. A review of measurement-based assessment of aerosol direct radiative effect and forcing. Atmos. Chem. Phys., in press. (pdf)