One of the main objectives of this project is to generate a measurement based description of The Global Aerosol System. The approach to achieving this goal is based on evaluation of the operational satellite data (presentand future) with AERONET, use of new remote sensing approaches to derive aerosol absorption, use of the data to improve the sources and processes in the GOCART model by means of an inverse technique, and update of model aerosol distributions by 3D variational data assimilation.
The final product is a combination of the observational data themselves and the
global, gridded aerosol fields from GOCART simulations and GAAS assimilation experiments.
For more information on these subjects, scroll down or click on the following links:
Evaluation of data |
Use of lidar data |
Remote sensing of aerosol absorption |
Analyzing the GOCART model |
The improved GOCART model and assimilation | The Global Aerosol System data set
|
Current comparison of MODIS and TOMS operational data, which now has over 6000 points over the land and 2000 points over the ocean.
|
The MODIS operational optical thickness was evaluated using scatter plots against AERONET over land and ocean [Chu et al 2002; Remer et al 2002]. The MODIS performance was found to be within the pre-launch specified error bars [Kaufman et al 1997; Tanré et al 1997]: Dt=±0.03±0.05t over the ocean and Dt=±0.05±0.15t over the land. The comparison now has over 6000 points over the land and 2000 points over the ocean (see image), with average bias over the ocean of only Dt=+0.005. However a point appears in the scatter plots only if both the satellite and AERONET found it to be cloud free, leaving a room for additional evaluation. In a different study we concluded, using AERONET data, that the 10:30 or 1:30 orbits of Terra and Aqua should represent adequately the daily average aerosol loading [Kaufman et al 2000]. There is a need to put together these two issues and test the satellite (MODIS, TOMS etc.) average monthly data vs. these of AERONET. In desert regions TOMS adjusted using MODIS data outside the desert and AERONET data will be the main sources of information.
Back to top
We plan to use lidar data from the GLAS and CALIPSO missions and from the MPL network to supplement the dataset with aerosol profiles. We shall use a combined inversion of the
MODIS+lidar data as suggested by Leon et al [2003] and Kaufman et al [2003] to get separately the profiles of the fine and coarse particles.
Back to top
|
Graph demonstrating the derivation of the single scattering albedo of dust from MODIS spectral measurements.
|
MODIS and TOMS data are sensitive to aerosol absorption and can be used to derive the single scattering albedo across the solar spectrum: in the UV from TOMS [Torres et al 1998, 2002a] and in the visible to mid IR from MODIS [Kaufman et al 2001b, 2002]. MODIS spectral measurements of absorption can cover the parts of the 0.47-2.1 mm spectrum for which the surface reflectance is brighter than 0.15 and the aerosol optical thickness is larger than 0.3 [Kaufman, 1987]. While the TOMS technique is used operationally, and depends on a good estimate of the aerosol height (e.g. from GOCART), the MODIS technique has only been applied only to a limited number of cases (see image). We plan to make the technique more automatic by deriving the difference in the top of the atmosphere reflectance, for couples of days with the same view direction (16 days apart), for which the aerosol optical thickness changed by more than 0.4. Preliminary tests show that the technique is feasible.
The TOMS method of aerosol absorption characterization takes advantage of the large Rayleigh scattering component in the near ultraviolet region. Because of the interaction of aerosol absorption and multiple molecular scattering, the length of a photon's path through an absorbing aerosol layer is increased, so that the chance of aerosol absorption is significantly enhanced. The resulting spectral contrast of the measured near UV radiances at the top of the atmosphere can, therefore, be used to separate aerosol absorption from scattering effects [Torres et al., 1998]. Both gray [carbonaceous particles] and colored aerosols [desert dust] can be detected and measured by this approach.
Back to top
Simulation of the trace-back retreival of dust emission sources from MODIS: |
Initial guess emission sources |
"True" emission sources used in the GOCART simulations of the MODIS data |
Emission sources retrieved by trace-back inversion from the simulated MODIS data |
The knowledge of inversion processes gained from the routine inversion of AERONET
measurements [Dubovik and King 2000] is used to develop a technique that uses the
measurement field (MODIS, AERONET, TOMS) and finds the optimum strength of sources that fits the observed aerosol field. This method was developed and found to be correct against simulated examples. A poor fit of the model to the data over the ocean, after
optimization can be used to assess missing sink processes or inaccurate rates of processes in the atmosphere. It follows the suggestion of Rosenfeld [2002] that the addition of large sea salt CCNs to the aerosol and the formation of clouds on the mixture of pollution with sea salt generates precipitation despite the pollution and cleans the air. We plan to further develop the method and apply it to multiyears of MODIS, AERONET and TOMS data sets to evaluate the global sources of natural and anthropogenic fine and coarse aerosols. The results will be compared with the location and activity of cities and the location and strength of fires detected from the MODISes 4 passes in 24 hours.
Back to top
The improved GOCART model will be evaluated first by AERONET
[Chin et al 2002] and the satellite data. Residuals between the GOCART generated
aerosol field, the satellite data and an assimilation of the aerosol field will be calculated and
used to evaluate the regional radiative forcing. The calculation of radiative forcing will be
carried out with the Goddard radiative transfer model [Chou and Suarez, 1999] which is coupled
with the GOCART model. The short-wave radiative processes in the model are parameterized in 11
spectral bands (0.2 - 10 µm), accounting for the absorption by O3, O2, CO2, H2O and scattering by
gas and particles [Chin et al., 2001]. The fields from GAAS will be evaluated in a similar fashion,
but with the radiative parameterization of the finite-volume GCM used instead for the calculation
of the radiative forcing.
Back to top
The Global Aerosol System data set is produced on a 1°x1° resolution, daily and
monthly, for key aerosol parameters derived from the measurements and/or from the
measurement driven models:
|
Image of GOCART, AERONET and MODIS aerosol optical thickness.
|
- Each MODIS measurement over the ocean identifies a fine mode (out of 4) and coarse mode (out of 5). The average optical thickness over the 1°x1° grid box is accumulated in the 4 fine modes and the 5 coarse modes. We shall archive these results and sum them into fine, tf, and coarse,tc optical thicknesses. Similar analysis will be performed over the land but for fewer aerosol modes. The TOMS optical thickness in the UV, the absorption optical thickness, tabs(l) are archived. The aerosol reflected and absorbed solar flux, based on the measurements is also archived.
- AERONET measurements of tf, tc and tabs(l), size distribution, where available.
- Lidar profiles, where available.
- GOCART produced tf, tc and tabs(l) in the same location of the satellite data for comparison.
Produce the optical thicknesses also in the cloudy regions where the satellites do not measure
aerosol. Produce the ratio of the daily average optical thicknesses to those during the satellite
passArchive the optical thicknesses and the composition of the aerosol (see image), the
reflected and absorbed fluxes, profiles of the aerosol and fraction above the clouds.
- GOCART and GAAS assimilated gridded aerosol fields, globally
- Residuals between the GOCART parameters and the measurements (in cloud free) and the
assimilation.
- Through the study we shall devise automatic criteria for selecting the parameters out of
duplicates of sources based on the expected error in each. For example over the ocean for high concentration of aerosol (t>0.3) the preference is given to satellite observations. For lower optical thickness the error in the satellite derivation (±0.03±0.05t) will be compared to
estimated error in the model based on comparison with near by AERONET stations or
comparison with satellite data in regions with concentrated aerosol.
Back to top
Evaluation of data |
Use of lidar data |
Remote sensing of aerosol absorption |
Analyzing the GOCART model |
The improved GOCART model and assimilation |
The Global Aerosol System data set
General |
The global aerosol system |
The direct effect |
Effects on clouds
Back to top
