Aerosol Workshop — June 2-3, 1997
Session 3: Satellite Aerosol Retrievals
(Facilitator: Pat McCormick, Recorder: Ina Tegen)
AVHRR Decadal Aerosol Time Series (Pathfinder Project)
Larry Stowe, NOAA NESDIS
Thirteen years (1981-1993) of September monthly mean values of aerosol optical thickness, AOT, at 0.63 microns and cloud-free broadband reflected flux, Fsw, at the top of atmosphere for ocean grid cells (110 km resolution) have been used to estimate the maximum shortwave aerosol forcing observed by AVHRR over this period. A map of variance explained by linear regression (r2) of Fsw as a function of AOT over the thirteen years showed that primarily grid cells in the region from about 20S to 30N exhibited an aerosol forcing signature sufficiently large to be detected by AVHRR (i.e., regions where r2 > 30%). In this region, the maximum monthly mean aerosol forcing [Fsw(year of AOTmin ) - Fsw(year of AOTmax )] was negative ranging between -10 and -40 Wm-2. Maps of the maximum change in optical depth (AOTmax - AOTmin ) over the thirteen years suggest that this forcing is due to the combined effects of stratosphere aerosol from Mt. Pinatubo and biomass burning aerosol over Indonesia and west of tropical Africa and South America (0.2 < ΔAOT < 1.0). It is likely that for other months of the year, aerosol forcing signals would be detected with the PATMOS datasets in other regions and at higher latitudes associated with Saharan dust outbreaks and industrial pollution. It is proposed that these observations be used to test the performance of climate models in simulating "aerosol events" over this thirteen year period.
Combining AVHRR and GOES to Obtain Better Aerosol Data
Phil Durkee, Naval Postgraduate School
Aerosol optical retrievals from polar-orbiting satellites generally offer higher spatial, radiometric and spectral resolution than geostationary satellites. The NOAA POES series nominally provides two views per day of a region at nearly constant local times. Geostationary satellites provide imagery at up to 15 minute resolution. The GOES 8/9/10 visible channel offers sufficient radiometric resolution to retrieve aerosol optical depth but lacks other wavelengths needed to account for particle size variations and subsequent effects on directional scattering.
A technique that combines POES and GOES can provide aerosol optical depth at high temporal resolution while accounting for variations in directional scatter. This technique compares favorably to POES retrievals and airborne sunphotometer measurements.
Use of GOES in optical depth retrievals will greatly improve analysis of aerosol radiative forcing in an "events" strategy. Plumes from aerosol sources can be tracked at high temporal resolution. This will greatly aide in identifying source regions, vertical distributions, transport processes and diurnal forcing. In addition, GOES analysis can provide regional mean scattering phase function characteristics from observations at varying sun-earth-satellite geometry.
TOMS UV Measurements of Aerosols
Jay Herman, Goddard Space Flight Center
TOMS can observe both absorbing (smoke, dust, volcanic ash) and small-particle nonabsorbing aerosols (sulfates etc.) over land and water. We have a daily global record of aerosols from 1979 to the present. Recently the observed aerosol index has been converted to optical depth and compared with sunphotometer data. The observations show that the TOMS aerosol index is linearly related to the sunphotometer optical depth with the slope corresponding to different aerosol types. The form is (for absorbing aerosols): Aerosol Index = Ki τ (1 - w) (1 - P), where Ki = slope, w = single scatter albedo, P = pressure in atmospheres, τ = optical depth.
Time series for nonabsorbing aerosols show the strong sources from the U.S. eastern region and over western Europe. The amount of nonabsorbing aerosols decreases rapidly with latitude with almost no nonabsorbing aerosols in the Southern Hemisphere. The nonabsorbing aerosol index is given by Aerosol Index = Ci (340 - 380)
Aerosol Optical Depth and Size from 2-Channel AVHRR and ADEOS/OCTS
Terry Nakajima, University of Tokyo
Aerosol optical properties. An efficient algorithm has been developed for retrieving the aerosol optical thickness and Angstrom exponent from channel-1 and -2 of AVHRR (Nakajima and Higurashi, J. Geophys. Res., in press, 1997; Higurashi and Nakajima, An analysis of radiative fields in a coupled atmosphere-ocean system, International Radiation Symposium, University of Alaska, Fairbanks, August 19-24, 1996). It is found that the set of the aerosol optical thickness and Angstrom exponent can show a distinct features of difference in the airmass origin on global scale. The Angstrom exponent significantly increased over large areas of several hundred kilometers extent around industrial areas and biomass burning regions; small values are found in the most areas of oceanic atmospheres and areas loaded by dust particles. Combining the present results with the absorbing aerosol characteristics derived from TOMS UV analyses is a promising approach for improving our knowledge of the global distribution of aerosol optical properties.
Cloud microphysical parameters. The algorithm of Nakajima and Nakajima (J. Atmos. Sci., 52, 4043-4059, 1995) has been extended for global analyses. The results show a similar characteristics of the effective cloud particle radius as found by Han et al. (J. Climate, 7, 465-497, 1994), i.e., the effective particle radius is larger over ocean areas than over land areas. It is found, however, our values are systematically larger by about 2 micron than those of Han et al. (1994) for a case of 1988. Calibration constants of AVHRR might be the cause of the difference.
The following strategies are important for future aerosol studies:
- Satellite-guided experiments. There are characteristic regions where the in situ aerosol chemical and optical properties should be studied to confirm and validate the retrievals from AVHRR two channel and TOMS UV algorithms.
- Homogeneous time series of aerosol parameters. It is important to define good aerosol-related parameters which can be retrieved from 14 year satellite records of AVHRR and TOMS. Those parameters should be simple enough to be retrieved homogeneously on global scale from existing satellite radiance data.
SAGE Measurements of El Chichon and Pinatubo Aerosols
Andrew Lacis, NASA Goddard Institute for Space Studies
A global climatology of stratospheric aerosol radiative properties has been compiled from SAGE II multiwavelength extinction measurements for the period 1984-1994 (Thomason and Poole, JGR, 102, 8967-8976, 1997). With the assumption that the aerosol composition is adequately represented by 75% solution of H2SO4 (based on in situ measurements by Deshler et al., GRL, 20, 1435-1438, 1993), the aerosol optical depth and effective radius can be accurately retrieved, based on Mie scattering. Because the SAGE measurements do not have a sufficiently broad spectral baseline to constrain the size distribution variance, the effective size distribution variance is instead inferred from ISAMS limb scan measurements at 12.1µm (Lambert et al., GRL, 20, 1287-1290, 1993). The size distribution variance, while it has little impact on the spectral extinction at visible wavelengths, has substantial impact on the amount of extinction at thermal wavelengths, which account for most of the stratospheric heating produced by volcanic aerosols. Based on these measurements, a monthly-mean stratospheric aerosol climatology of optical depth and particle size with 4° × 5° horizontal and 5 km vertical resolution has been developed for GCM use (Hansen et al., NATO ASI Series, I42, 233-272, 1996). The GCM-simulated global surface cooling of about 0.5°C, and stratospheric warming of several degrees, are in basic agreement with observed surface and stratospheric temperature changes (Hansen et al., GRL, 23, 1665-1668, 1996; Angell, JGR, 102, 9479-9486, 1997). This provides a closure between the observed radiative forcing, the climate model simulation of the climate response, and the observed real world climate response.
POAM II Stratospheric Aerosol Data
Cora Randall, University of Colorado
The POAM satellite instruments series is similar in concept to SAGE, measuring aerosol extinction profiles at 5 wavelengths from 0.353 to 1.06 microns. POAM II (Oct. 1993 to Nov. 1996) and POAM III (to be launched in January, 1998) make measurements in polar regions, OOAM (to be launched in August, 1997) makes measurements at mid and low latitudes.
POAM II measurements compare well with coincident SAGE II observations, and extend the SAGE measurements to higher latitudes year-round.
The POAM II and POAM III polar measurements extend the SAM II aerosol measurements into the post-Pinatubo era, and can provide additional parameters such as volume density, surface area density, effective radius and number densities (with appropriate assumptions regarding size distribution functions). OOAM provides the same data for mid and low latitudes.
The combination of POAM III, OOAM and SAGE III (and SAGE II if still operational) solar occultations, and SAGE III lunar occultations will provide unprecedented global coverage of aerosol vertical profile measurements. Transmission profiles for these instruments should reach the mid-troposphere.
Limb occultation measurements such as those of the POAM and SAGE instrument series are the best means of deriving climatologies of sub-visual cirrus clouds.
Recommendations: These mainly stratospheric data should be utilized to the fullest advantage to understand aerosol climate forcing effects during volcanic periods. They should further be used in conjunction with the column measurements such as TOMS and AVHRR to separate tropospheric aerosols from stratospheric aerosols (or to separate low troposphere from high troposphere, when the transmissions permit). Climatologies of cirrus clouds should be expanded to include on-going measurements and measurements at high latitudes (see note below).
Note: Regarding subvisual cirrus clouds: It was not clear from the workshop discussions if these were deemed important with regard to climate forcing. What is the general opinion on this issue? We did not have time to discuss the capabilities of limb occultation measurements regarding cirrus cloud detection; but if the general agreement is that they play an important role in the indirect forcing, I think we should include efforts to use limb occultation measurements to document their climatology and year-to-year variability in any recommendations made to NASA.
Note: I have been asked about availability of aerosol date from SME (Solar Mesospheric Explorer). If making that data public in a useable form is a priority for modelers, we can pursue this at LASP to see what resources it would require.
Session 3 Summary by
This session provided an overview of currently available satellite retrieval methods to obtain global aerosol distributions for the period from the beginning of the 1980's until present. Retrieval methods to obtain tropospheric aerosols (from AVHRR, GOES, TOMS) and stratospheric aerosols (SAGE, POAM II) were presented. Although quantitative retrieval of tropospheric aerosol distributions and properties is difficult, those products may provide useful information if combined with each other, in-situ measurements, and global modeling results. The combination of information from several channels and different satellites (geostationary and polar orbiting) appears to be a promising approach for obtaining quantitative information about tropospheric aerosols. Some starting points for derivation of information about aerosol processes and interannual changes from those retrievals were presented.
Larry Stowe estimated aerosol radiative forcing changes at low and middle latitudes from his AVHRR aerosol product for the month of September. He found an optical thickness of about 0.2-0.3 due to Pinatubo aerosols and 0.7 near Indonesia due to biomass burning. The latter was estimated to cause a regional forcing of -10 to -40 W/m2. Near the United States East Coast (location of the TARFOX experiment) a positive forcing of 20-30 W/m2 was found. For most regions the tropospheric aerosol effect was too small to be detected by this method. He indicated that other months would probably have measurable tropospheric forcings due to dust and industrial pollutants, as concentrations of these aerosols are not at their maximum in September.
Discussion after Stowe's paper:
Question: How would an error in the assumption of the aerosol absorption change the results?
Stowe: An error in the assumption of the imaginary part of refractive index would lead to errors of only 20% in these retrievals. Also, even if errors in the retrievals occur due to too many assumptions, interannual changes in aerosol patterns can be recognized.
Phil Durkee showed measurements taken during a TARFOX flight (east coast of United States), which combined results from the polar orbiting AVHRR instrument and the geostationary GOES satellite. Intercomparison with sunphotometer measurements (4 wavelengths) showed good agreement for optical thickness. Retrieved optical thickness values ranged between 0.2 and 0.3.
While GOES provides a substantially better temporal resolution than AVHRR (of the order of hours compared to days), the spectral resolution from AVHRR is necessary to provide additional information about aerosol properties. An important point in satellite AOT retrievals is the use of the correct aerosol model - it turns out that when tested with 2 phase functions the results can vary by a factor of 3 with the scattering angle.
Discussion after Durkee's paper:
Question: Does the change in phase functions depend on the assumed particle size?
Durkee: Yes, size was allowed to vary in the retrieval; refractive index was held constant.
Question: Is it possible to find one 'best' phase function?
Durkee: It is unlikely that all parameters will ever be accurately defined, uncertainties must be expected. But the actual range in radiances is smaller than would be implied by phase function differences.
Jay Herman presented retrievals for both absorbing and non-absorbing aerosols for period of 1979 onward based on TOMS measurements. For absorbing aerosols (dust and carbonaceous particles) the TOMS product gives an 'Aerosol Absorption Index' for land ocean regions.
Some limitations of the absorbing aerosol product are that no absorbing aerosols can detected close to the ground (below 1.5 km) and under cloudy conditions sensitivity is reduced. For non-absorbing (sulfate) aerosols the sensitivity is less than for absorbing aeroso interference of surface reflection occurs.
Case studies at some specific sites show that at Tenerife the aerosol index p January, which is caused by aerosol transport rather than source strength at this time of year. In Spain, on the other hand, the non-absorbing aerosol which is caused by in-situ aerosol production is stronger than the absorbing (Saharan dust) which is caused by in-situ aerosol production. In South America the absorbing aerosol background signal is small, the biomass burning sign found to be negatively correlated with rainfall.
Comparison with in-situ sunphotometer data shows a generally high positive correlati between AOT and TOMS absorption index.
Terry Nakajima described a 2-channel retrieval method which provides information on aerosol size as well as optical thickness. Explicit information on aerosol composition in not provided by this method, although in some cases it can be inferred from aerosol history.
His results show that, for biomass burning particles, South American aerosols appear to have different properties than South African aerosols, specifically differing in size.
In areas influenced by air masses from industrial regions particles are smaller at higher optical thicknesses, while for very high optical thicknesses at some locations the particle sizes are larger because of hygroscopic growth.
Comparison of cloud and aerosol optical thicknesses and particle sizes shows no correlation between cloud optical thickness and aerosol optical thickness on a global scale.
Discussion after Nakajima's paper:
Question: How do the results for cloud particle sizes compare to the results of Han and Rossow?
Nakajima: The results have a similar magnitude.
Question: For the case of soil dust aerosol, can a difference in particle size be observed for different distances from the source regions?
Nakajima: No, such a size change was not observed.
Question: Can interannual changes be observed in these retrievals?
Nakajima: Only one year (1991) has been analyzed so far.
Andy Lacis described aerosol information extracted from multispectral SAGE data. Specification of the radiative forcing requires knowledge of the aerosol effective size, vertical distribution and refractive index. Spectral extinction ratios are sensitive to particle size, and as a result, if the size distribution is monomodel, lookup tables can be devised to determine effective size from the spectral extinctions. With the size information optical thicknesses can be derived at other wavelengths. Although bi-model size distributions may be more realistic for stratospheric aerosols, the use of such distributions was not found to improve the overall fit with observations.
For calculations of the aerosol effect in a GCM the aerosol optical thickness and effective radius are needed as input. In the GISS GCM it is found that the main local effect of the aerosols is stratospheric heating via absorption of thermal radiation, the solar contribution being relatively smaller.
Discussion after Lacis' paper:
Question: Are the retrieved sizes vertically resolved?
Lacis: No, in the present retrieval they are the column average.
Question: Will SAGE III improve the accuracy of stratospheric aerosol retrievals?
Lacis: Yes, for several reasons, but especially because of somewhat broader spectral coverage.
Cora Randall described 'POAM II Polar Ozone and Aerosol Measurement) stratospheric aerosol data. The measurement concept is similar to that of SAGE. POAM II has 1km resolution at 15-30 km altitude. Observations from this instrument exist for the years 1994-1996. Comparisons with SAGE in this period show good agreement. POAM II shows that lower aerosol extinction occurs inside of the polar vortex.
Two follow-up instruments will be launched: OOAM (1997), POAM III (1998).