Reduction of tropical cloudiness by soot

Science

Volumn 288, Issue 5468, 2000, Pages 1042-1047

  Ackerman, A S ^a   Toon, O B ^b   Stevens, D E ^c   Heymsfield, A J ^d   Ramanathan, V ^e   Welton, E J ^f  

^a NASA AMES RESEARCH CENTER   (United States)

^b UNIVERSITY OF COLORADO   (United States)

^c LAWRENCE LIVERMORE NATIONAL LABORATORY   (United States)

^d NATIONAL CENTER FOR ATMOSPHERIC RESEARCH   (United States)

^e UNIVERSITY OF CALIFORNIA   (United States)

^f SCIENCE SYSTEMS AND APPLICATIONS INC   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

AEROSOL; CLOUD COVER; SOOT; TROPICAL METEOROLOGY;

AEROSOL; ARTICLE; ATMOSPHERE; CLOUD; COOLING; MODEL; POLLUTION; PRIORITY JOURNAL; SIMULATION; SOLAR RADIATION; SOOT; TROPIC CLIMATE; WATER CONTENT;

EID: 0034640239     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.288.5468.1042     Document Type: Article

References (54)

1. 4 Scripps Institution of Oceanography, UCSD, La Jolla, CA, 1996) p. 83.
2. S. Twomey, Atmos. Environ. 8, 1251 (1974).
3. B. A. Albrecht, Science 245, 1227 (1989).
4. A. S. Ackerman, O. B. Toon, P. V. Hobbs, Science 262, 226 (1993).
5. R. Pincus and M. B. Baker, Nature 372, 250 (1994).
6. L. F. Radke, J. A. Coakley Jr., M. D. King, Science 246, 1146 (1989).
7. E. E. Hindman, W. M. Porch, J. G. Hudson, P. A. Durkee, Atmos. Environ. 28, 3393 (1994).
8. J. P. Taylor and A. S. Ackerman, Q. J. R. Meteorol. Soc. 125, 2643 (1999).
9. A. Jayaraman et al., J. Geophys. Res. 103, 13827 (1998).
10. 2 for an average aerosol optical depth of 0.2 (at 0.5 μm). They also report a range of 0.88 to 0.90 for measurements of the aerosol single scattering albedo at 0.5 μm (which take into account the effects of relative humidity on particle size).
11. Satellite retrievals of monthly diurnal-average low-cloud fractional coverage during February over the years 1989 to 1993 range from ∼0.2 to 0.3 at the equator to ∼0.1 to 0.2 at 10°N [W. B. Rossow and R. A. Schiffer, Bull. Am. Meteorol. Soc. 72, 2 (1991)]. METEOSAT-5 retrievals by VR yield comparable values for 1999.
12. 2.
13. M. C. Wyant, C. S. Bretherton, H. A. Rand, D. E. Stevens, J. Atmos. Sci. 54, 168 (1997).
14. B. A. Albrecht, J. Atmos. Sci. 38, 97 (1981).
15. B. Stevens et al., 13th Symposium on Boundary Layers and Turbulence, 10 to 15 January 1999, Dallas, TX (American Meteorological Society, Boston, 1999), 269-270. The simulations here are derived from a model intercomparison setup described therein.
16. K. Brill and B. Albrecht, Mon. Weather Rev. 110, 601 (1982).
17. The dynamics model [D. E. Stevens, C. S. Bretherton, J. Comput. Phys. 129, 284 (1997)] solves the anelastic Navier-Stokes equations in conservative form using 5 s time steps on a domain spanning 6.4 km by 6.4 km horizontally and 3 km vertically, which is uniformly discretized into 32 × 32 × 75 grid cells. The boundary conditions are doubly periodic in the horizontal, and rigid at the top and bottom. Surface fluxes are parameterized through similarity relations [J. A. Businger, J. C. Wyngaard, Y. Izumi, E. F. Bradley, J. Atmos. Sci. 28, 181 (1971)]. A sponge layer at the top of the model dampens trapped buoyancy waves at altitudes >500 m above the trade inversion, which is defined as the horizontal-average height where the total water mixing ratio is 6.5 g of water per kg of air. First-order turbulence closure is used for subgrid-scale mixing [J. Smagorinsky, Mon. Weather Rev. 91, 99 (1963); D. K. Lilly, Tellus 14, 1012 (1962)] with a stability-dependent mixing length [J. W. Deardorff, Boundary-Layer Meteorol. 18, 495 (1980)] modified to account for the effects of evaporation [P. J. Mason and M. K. MacVean, J. Atmos. Sci. 47, 1012 (1990)]. Large-scale subsidence is calculated as the product of the divergence rate (assumed constant) and altitude. Subsidence and radiative forcings are linearly attenuated to zero in the 300 m above the trade inversion to prevent drift of the overlying atmospheric properties due to any imbalanced forcings.
18. The dynamics model [D. E. Stevens, C. S. Bretherton, J. Comput. Phys. 129, 284 (1997)] solves the anelastic Navier-Stokes equations in conservative form using 5 s time steps on a domain spanning 6.4 km by 6.4 km horizontally and 3 km vertically, which is uniformly discretized into 32 × 32 × 75 grid cells. The boundary conditions are doubly periodic in the horizontal, and rigid at the top and bottom. Surface fluxes are parameterized through similarity relations [J. A. Businger, J. C. Wyngaard, Y. Izumi, E. F. Bradley, J. Atmos. Sci. 28, 181 (1971)]. A sponge layer at the top of the model dampens trapped buoyancy waves at altitudes >500 m above the trade inversion, which is defined as the horizontal-average height where the total water mixing ratio is 6.5 g of water per kg of air. First-order turbulence closure is used for subgrid-scale mixing [J. Smagorinsky, Mon. Weather Rev. 91, 99 (1963); D. K. Lilly, Tellus 14, 1012 (1962)] with a stability-dependent mixing length [J. W. Deardorff, Boundary-Layer Meteorol. 18, 495 (1980)] modified to account for the effects of evaporation [P. J. Mason and M. K. MacVean, J. Atmos. Sci. 47, 1012 (1990)]. Large-scale subsidence is calculated as the product of the divergence rate (assumed constant) and altitude. Subsidence and radiative forcings are linearly attenuated to zero in the 300 m above the trade inversion to prevent drift of the overlying atmospheric properties due to any imbalanced forcings.
19. The dynamics model [D. E. Stevens, C. S. Bretherton, J. Comput. Phys. 129, 284 (1997)] solves the anelastic Navier-Stokes equations in conservative form using 5 s time steps on a domain spanning 6.4 km by 6.4 km horizontally and 3 km vertically, which is uniformly discretized into 32 × 32 × 75 grid cells. The boundary conditions are doubly periodic in the horizontal, and rigid at the top and bottom. Surface fluxes are parameterized through similarity relations [J. A. Businger, J. C. Wyngaard, Y. Izumi, E. F. Bradley, J. Atmos. Sci. 28, 181 (1971)]. A sponge layer at the top of the model dampens trapped buoyancy waves at altitudes >500 m above the trade inversion, which is defined as the horizontal-average height where the total water mixing ratio is 6.5 g of water per kg of air. First-order turbulence closure is used for subgrid-scale mixing [J. Smagorinsky, Mon. Weather Rev. 91, 99 (1963); D. K. Lilly, Tellus 14, 1012 (1962)] with a stability-dependent mixing length [J. W. Deardorff, Boundary-Layer Meteorol. 18, 495 (1980)] modified to account for the effects of evaporation [P. J. Mason and M. K. MacVean, J. Atmos. Sci. 47, 1012 (1990)]. Large-scale subsidence is calculated as the product of the divergence rate (assumed constant) and altitude. Subsidence and radiative forcings are linearly attenuated to zero in the 300 m above the trade inversion to prevent drift of the overlying atmospheric properties due to any imbalanced forcings.
20. The dynamics model [D. E. Stevens, C. S. Bretherton, J. Comput. Phys. 129, 284 (1997)] solves the anelastic Navier-Stokes equations in conservative form using 5 s time steps on a domain spanning 6.4 km by 6.4 km horizontally and 3 km vertically, which is uniformly discretized into 32 × 32 × 75 grid cells. The boundary conditions are doubly periodic in the horizontal, and rigid at the top and bottom. Surface fluxes are parameterized through similarity relations [J. A. Businger, J. C. Wyngaard, Y. Izumi, E. F. Bradley, J. Atmos. Sci. 28, 181 (1971)]. A sponge layer at the top of the model dampens trapped buoyancy waves at altitudes >500 m above the trade inversion, which is defined as the horizontal-average height where the total water mixing ratio is 6.5 g of water per kg of air. First-order turbulence closure is used for subgrid-scale mixing [J. Smagorinsky, Mon. Weather Rev. 91, 99 (1963); D. K. Lilly, Tellus 14, 1012 (1962)] with a stability-dependent mixing length [J. W. Deardorff, Boundary-Layer Meteorol. 18, 495 (1980)] modified to account for the effects of evaporation [P. J. Mason and M. K. MacVean, J. Atmos. Sci. 47, 1012 (1990)]. Large-scale subsidence is calculated as the product of the divergence rate (assumed constant) and altitude. Subsidence and radiative forcings are linearly attenuated to zero in the 300 m above the trade inversion to prevent drift of the overlying atmospheric properties due to any imbalanced forcings.
21. The dynamics model [D. E. Stevens, C. S. Bretherton, J. Comput. Phys. 129, 284 (1997)] solves the anelastic Navier-Stokes equations in conservative form using 5 s time steps on a domain spanning 6.4 km by 6.4 km horizontally and 3 km vertically, which is uniformly discretized into 32 × 32 × 75 grid cells. The boundary conditions are doubly periodic in the horizontal, and rigid at the top and bottom. Surface fluxes are parameterized through similarity relations [J. A. Businger, J. C. Wyngaard, Y. Izumi, E. F. Bradley, J. Atmos. Sci. 28, 181 (1971)]. A sponge layer at the top of the model dampens trapped buoyancy waves at altitudes >500 m above the trade inversion, which is defined as the horizontal-average height where the total water mixing ratio is 6.5 g of water per kg of air. First-order turbulence closure is used for subgrid-scale mixing [J. Smagorinsky, Mon. Weather Rev. 91, 99 (1963); D. K. Lilly, Tellus 14, 1012 (1962)] with a stability-dependent mixing length [J. W. Deardorff, Boundary-Layer Meteorol. 18, 495 (1980)] modified to account for the effects of evaporation [P. J. Mason and M. K. MacVean, J. Atmos. Sci. 47, 1012 (1990)]. Large-scale subsidence is calculated as the product of the divergence rate (assumed constant) and altitude. Subsidence and radiative forcings are linearly attenuated to zero in the 300 m above the trade inversion to prevent drift of the overlying atmospheric properties due to any imbalanced forcings.
22. The dynamics model [D. E. Stevens, C. S. Bretherton, J. Comput. Phys. 129, 284 (1997)] solves the anelastic Navier-Stokes equations in conservative form using 5 s time steps on a domain spanning 6.4 km by 6.4 km horizontally and 3 km vertically, which is uniformly discretized into 32 × 32 × 75 grid cells. The boundary conditions are doubly periodic in the horizontal, and rigid at the top and bottom. Surface fluxes are parameterized through similarity relations [J. A. Businger, J. C. Wyngaard, Y. Izumi, E. F. Bradley, J. Atmos. Sci. 28, 181 (1971)]. A sponge layer at the top of the model dampens trapped buoyancy waves at altitudes >500 m above the trade inversion, which is defined as the horizontal-average height where the total water mixing ratio is 6.5 g of water per kg of air. First-order turbulence closure is used for subgrid-scale mixing [J. Smagorinsky, Mon. Weather Rev. 91, 99 (1963); D. K. Lilly, Tellus 14, 1012 (1962)] with a stability-dependent mixing length [J. W. Deardorff, Boundary-Layer Meteorol. 18, 495 (1980)] modified to account for the effects of evaporation [P. J. Mason and M. K. MacVean, J. Atmos. Sci. 47, 1012 (1990)]. Large-scale subsidence is calculated as the product of the divergence rate (assumed constant) and altitude. Subsidence and radiative forcings are linearly attenuated to zero in the 300 m above the trade inversion to prevent drift of the overlying atmospheric properties due to any imbalanced forcings.
23. 2; the ocean surface albedo is assumed to be independent of wavelength and is determined from the wind speed at 10 m (averaging ∼8 m/s) and the solar zenith angle using the parameterization of J. Hansen et al. [Mon. Weather Rev. 111, 609 (1983)].
24. 2; the ocean surface albedo is assumed to be independent of wavelength and is determined from the wind speed at 10 m (averaging ∼8 m/s) and the solar zenith angle using the parameterization of J. Hansen et al. [Mon. Weather Rev. 111, 609 (1983)].
25. 2; the ocean surface albedo is assumed to be independent of wavelength and is determined from the wind speed at 10 m (averaging ∼8 m/s) and the solar zenith angle using the parameterization of J. Hansen et al. [Mon. Weather Rev. 111, 609 (1983)].
26. 2; the ocean surface albedo is assumed to be independent of wavelength and is determined from the wind speed at 10 m (averaging ∼8 m/s) and the solar zenith angle using the parameterization of J. Hansen et al. [Mon. Weather Rev. 111, 609 (1983)].
27. 2; the ocean surface albedo is assumed to be independent of wavelength and is determined from the wind speed at 10 m (averaging ∼8 m/s) and the solar zenith angle using the parameterization of J. Hansen et al. [Mon. Weather Rev. 111, 609 (1983)].
28. E. Augstein, H. Riehl, F. Ostapoff, V. Wagner, Mon. Weather Rev. 101, 101 (1973).
29. -1, the mean value derived from the observations (16). Pseudo-random perturbations of temperature and water vapor mixing ratio are imposed below the inversion to promote turbulence initially; the amplitudes of the perturbations (which horizontally average to zero) are 0.1 K and 0.025 g/kg, respectively (15).
30. 2, an average precipitation rate of ∼0.1 mm/day, and 10-m wind speed of ∼8 m/s) the parameterization of P. J. Webster, C. A. Clayson, J. A. Curry [J. Clim. 9, 1712 (1996)] yields an amplitude of only 0.2 K for the diurnal variation in sea surface temperature.
31. Advective fluxes for the ATEX triangle were calculated by (19), from which advective forcings can be computed (as small differences between large terms): in the lower 60 mbar of the atmosphere, a drying tendency of 1.4 g/kg/day and a cooling tendency of 0.9 K/day are indicated. Following (15), we parameterize these forcings to fade linearly from a maximum at the surface to zero at the trade inversion. The surface drying tendency we use (1.5 g/kg/day) is taken directly from (15). We double the surface cooling tendency of (15) to 2 K/day, which compensates for a diurnal-average clear-sky cooling rate (1.7 K/day infrared cooling, 0.7 K/day solar heating) that is half the value imposed in (15).
32. In the real atmosphere, changes in aerosol-induced heating rates (tending to decrease cloud coverage) are linked to changes in cloud droplet concentrations (tending to increase cloud coverage) through microphysical details of the aerosol population (specifically, their chemical composition and size distribution). Any net effect of these opposed tendencies depends on such microphysical details, as well as the meteorology. Rather than attempt to comprehensively evaluate any net effects, we instead decouple the forcings by varying the haze properties and cloud droplet concentrations separately.
33. The model domain is initially cloudless; the cited droplet concentrations apply only to grid cells in which clouds appear. In all simulations the number concentration of haze particles increases linearly from zero at the ocean surface up to 600 m, maintains a uniform value up to the trade inversion, and vanishes linearly in the overlying 300 m. The haze particle size distribution is log-normal with a geometric mean radius of 0.1 μm and a geometric SD of 1.8.
34. 2 of solar radiation (diurnally averaged), which is comparable to the absorption measured during INDOEX 1998 (10). Optical properties of black carbon (soot) are taken from [P. Chylek, V. Ramaswamy, R. J. Cheng, J. Atmos. Sci. 41, 3076 (1984)].
35. To address the sensitivity of the simulations to small variations in initial conditions, for the baseline we ran an ensemble of four simulations that differed only in the pseudo-random distribution of initial perturbations of temperature and water vapor. Output from one member of the ensemble is shown in Figs. 2 and 3.
36. Vertically resolved cloud fraction, defined as the fraction of cells in each layer with cloud water >0.05 g/kg, is distinct from the fractional cloud coverage we define subsequently, which is evaluated from vertically integrated columns.
37. During the first 5 days of ATEX (the observation period upon which our meteorology is based), surface reports of the cloud fractional coverage, which may include clouds above the boundary layer, ranged from an early morning maximum ∼0.9 on two separate days to a minimum of ∼0.2 one afternoon; the average value over the period was 0.5 [B. Albrecht, J. Atmos. Sci. 48, 1519 (1991)]. For comparison, the simulated cloud coverage depends on a number of factors, including the criterion used to count cloudy grid columns, the model resolution, subgrid-scale mixing assumptions, and the water vapor above the inversion. With our model setup, the diurnal average (0.23) and range (0.1 to 0.4) from the baseline simulations are more comparable to those found over the Indian Ocean during the northeast monsoon (11).
38. We assume that all the enhanced solar absorption occurs only within the haze, though some fraction of the soot is likely to be incorporated into cloud droplets through nucleation and coagulation/coalescence. Cloud droplets (of typical radius 10 μm) collect significantly more sunlight than do haze (of typical radius 0.1 μm), and hence a fixed amount of soot will absorb more sunlight when embedded within cloud droplets (particularly when most of the haze lies below the bulk of cloud cover); such an effect could be expected to increase the impact of the soot on boundary-layer dynamics. Yet this expectation is not borne out by simulations in which all the soot is assumed to be within cloud droplets (when present within a grid cell) because such a small volume of boundary-layer air is occupied by cloud in our simulations.
39. "Cloud-burning" is the response of clouds to increased atmospheric heating, which includes reductions in cloud coverage and liquid and water path.
40. -3. However, the departure of that maximum from the average cloud-burning effect is not significant compared to the noise found for the baseline ensemble.
41. As seen in Fig. 5B, the average liquid water path is roughly independent of droplet concentration for any particular aerosol. The increases of cloud coverage with droplet concentration in these simulations (Fig. 5A) are largely due to enhancement of total droplet cross-sectional area and therefore optical depth (which vary as the cube root of droplet concentration, holding liquid water path fixed). Because cloud coverage is defined as the fraction of columns exceeding an optical depth threshold (2.5), columns do not need as much liquid water path at increased droplet concentrations to be counted as cloudy.
42. Simulations with moisture enhanced above the inversion layer (at 6 g/kg, up from 4.5 g/kg used in our other simulations) produce moister clouds with greater fractional coverage that are more strongly influenced by both the soot cloud-burning and the conventional indirect aerosol effects.
43. An increase in cloud coverage results in more reflection of solar energy (a cooling effect), while at the same time allowing less infrared energy to escape to space (a warming effect). The solar forcing dominates any infrared compensation in trade cumulus, which therefore exert a net cooling influence (compare clear-sky to cloudy net fluxes in Fig. 5C).
44. J. T. Houghton et al., Climate Change 1995: The Science of Climate Change (Cambridge Univ. Press, Cambridge, 1996), pp. 112-117.
45. S. K. Satheesh et al. [J. Geophys. Res. 104, 27421 (1999)] estimate that 60% of the aerosol optical depth in the INDOEX 1998 haze was anthropogenic (accounting for an optical depth of 0.12 at 0.5 μm), which is 70% of the optical depth in our baseline haze.
46. The scope of our calculations (17 simulations of 30 hours) demand a number of computational efficiencies, which include parameterized cloud microphysics, moderate grid resolution and domain area, and no treatment of horizontal radiative transfer.
47. Y. J. Kaufman and R. S. Fraser, Science 277, 1636 (1997).
48. Y. J. Kaufman and T. Nakajima, J. Appl. Meteorol. 32, 729 (1993).
49. Distributed over a 2.5-km-deep boundary layer (not in a trade cumulus regime), the diurnal-average aerosol forcings reported by P. B. Russel et al. [J. Geophys. Res. 104, 2289 (1999)] and P. Hignett, J. P. Taylor, P. N. Francis, M. D. Glew [J. Geophys. Res. 104, 2279 (1999)) yield clear-sky heating rates of 0.5 and 0.6 K/day, respectively. In comparison, the diurnal-average aerosol-induced heating rates in the boundary layer range from 0.5 to 1 K/day for our idealized INDOEX hazes.
50. Distributed over a 2.5-km-deep boundary layer (not in a trade cumulus regime), the diurnal-average aerosol forcings reported by P. B. Russel et al. [J. Geophys. Res. 104, 2289 (1999)] and P. Hignett, J. P. Taylor, P. N. Francis, M. D. Glew [J. Geophys. Res. 104, 2279 (1999)) yield clear-sky heating rates of 0.5 and 0.6 K/day, respectively. In comparison, the diurnal-average aerosol-induced heating rates in the boundary layer range from 0.5 to 1 K/day for our idealized INDOEX hazes.
51. R. R. Draxler and G. D. Hess, NOAA Tech. Memo. ERL ARL-224 (1997).
52. V. Manghani et al., Boundary-Layer Meteorol., in press.
53. E. J. Welton et al., Tellus B 52, 635 (2000).
54. Supported by NASA grants NAG5-6504 and NAGS-8362, the U.S. Department of Energy, and NSF. The Micro Pulse Lidar measurements were supported by NASA contract NA55-31363 and the NASA SIMBIOS project. We thank J. Coakley for performing radiative transfer calculations for comparison against ours, C. Twohy for helpful discussions regarding droplet nuclei measurements from INDOEX, the NOAA Pacific Marine Environmental Laboratory (PMEL) and the crews of the R/V Ronald H. Brown and the National Center for Atmospheric Research (NCAR) C-130 aircraft for making the INDOEX measurements possible, and two anonymous reviewers for comments that improved the manuscript.