Aerosols in the atmosphere arise mainly from dust (e.g., from the Sahara and China), smoke (from biomass burning in South America and Africa), and sulfates (from industrial pollution). Aerosols are detected using the differences between the measured 340 and 388 nm radiances I340 and I388 after removal of the surface and Rayleigh scattering contributions. Surface contributions are removed by using seasonal minimum reflectivity values derived from 14 years of TOMS reflectivity data (Herman and Celarier, 1997). The contribution is quite small, since reflectivity for the surface ranges from 2 to 8% at UV wavelengths.
A direct method for detecting aerosols in the atmosphere consists of using an aerosol index AI. This quantity is much simpler to compute than the optical depth, and does not need aerosol properties (e.g., refractive index and particle size) for its computation.
The sign has been selected so that AI>0 for absorbing aerosols (e.g., dust and smoke) and AI<0 for non-absorbing aerosols (e.g., sulfates) (Herman et al., 1997; Torres et al., 1998). There are two terms in the equation for AI. The first represents the measured ratio of radiances and contains the effects of Rayleigh scattering, surface reflectance, and aerosols. The second term is calculated for a pure Rayleigh scattering atmosphere, at the same geometry appropriate for the measured radiances, over the measured and climatological surface reflectivity (Herman and Celarier, 1997). AI is a measure of the deviation of I340/I388 from a pure Rayleigh atmosphere.
The value of AI is zero for the large scattering particles (~10 microns or larger) in clouds. For sulfate aerosols, the particle size near the accumulation mode is ~ 0.1 microns and produces a contrast effect between I340 and I388 caused by the wavelength dependent Mie scattering. For absorbing aerosols, the AI>0 effect is produced by interference with the l EXP(4) wavelength dependence of Rayleigh scattering from the atmosphere below the aerosol plume. The interference occurs whether or not the aerosol plume has a wavelength dependent absorption.
AI is useful for the basic detection of the presence of aerosols and will be used for the volcanic-ash aircraft warning capability. However, AI combines the effects of optical depth, particle size, single scattering albedo (absorption), and aerosol plume height. As such, it is not as useful as optical depth and single scattering albedo for quantitative calculations of atmospheric effects caused by the presence of aerosols.
The optical depth T and single scattering albedo w can be obtained using pre-computed tables of radiances as a function of T and w (Hsu et al., 1999b) if other aerosol parameters are known or assumed (e.g., plume height, particle size, refractive index). The tables are computed with I340/I388 and I340 as independent variables. A sample table is illustrated in Figure 14 and results for the optical depth are shown in Figure 15. For the results to be accurate, the 8 km x 8 km scene must be nearly cloud free. That is, the computed scene reflectivity must be less than about 15%.
Of the unknown parameters, the calculated optical depth is most sensitive to uncertainties in the plume height for absorbing aerosols (see next paragraph). Non-absorbing aerosol retrievals do not require knowledge of the plume height. The refractive index is estimated from the known aerosol type, dust, smoke, or volcanic ash. Estimates for these values are known from samples obtained from the local regions of origin (e.g., Saharan dust). Finally, the calculated values are weakly sensitive to the assumed particle size distribution (assumed to be lognormal) and mean particle radius. The mean particle radius is taken from typical values measured for either dust or smoke.
Use of AI or T is very useful for tracking dust plumes based on the wind motions in the 2 to 5 km region of the troposphere. The tracking of aerosol plumes using a data assimilation model or GCM containing known wind fields can be used to determine the aerosol-plume height. This is possible because of wind shears in the lower troposphere causing tracers to follow the observed aerosol plume only if the tracer is at the same altitude as the center of the plume. The hourly measurements from EPIC at 8 km spatial resolution will greatly improve this capability compared to a similar analysis successfully used for TOMS at 100 km resolution and once per day. The accuracy of this method is estimated to be 0.5 km and will reduce the error in optical depth determination from ± 30% to ± 10%. There is a large network of sunphotometers present on the ground (AERONET) from which the optical depth can be directly determined and used to provide validation for EPIC aerosol retrievals (Hsu et al., 1999b).
The more frequent hourly observations afforded by Triana-EPIC will enable the diurnal variation of both dust and smoke plumes to be observed for the first time over wide areas. For example, it is well known from ground-based observations that smoke from large fires is a minimum in the mornings and peaks in the late afternoon. Observing this variation over wide areas is important for the estimation of the radiative forcing of aerosols and their contribution to overall radiative energy balance of the Earth (Hsu et al., 1996).
One of the unique features of UV-radiance detection of aerosols is that they can be detected over both land and water. In the absence of ground-based data to locate aerosol plumes, visible wavelength detection of aerosols, such as done by AVHRR and MODIS, is largely confined to detection over water, or certain highly vegetated areas, where the surface reflectivity is low. For the UV wavelengths, detection of smoke and dust can be carried out over any surface, even highly reflective snow and ice (Hsu et al., 1999c). Detection of non-absorbing sulfate aerosols can be carried out over both land and water since the surface reflectivity is always between 2 and 8% (Herman and Celarier, 1997).
EPIC will be the first Earth observing spacecraft instrument to combine measurements in both UV and visible wavelengths. This will enable additional aerosol properties to be determined. Presently, TOMS can only determine two properties, optical depth and single scattering albedo, and must assume the others. By adding the visible channels, the aerosol mean particle size can also be determined. This quantity is important for distinguishing smoke from dust, and for estimates of radiative forcing caused by the presence of aerosols in the atmosphere.
A dramatic example of detecting large plumes over land with UV wavelengths was obtained during the 1998 Mexican fires that covered the southern US and occasionally extended up to Canada. Figure 16 shows a simulated Triana view of aerosol optical depth corresponding to this event. The EPA used the TOMS data shown in the figure to consider possible exemptions from pollution standards. Triana scenes will be even more useful since they will be at higher resolution and at all times of the day. In this case, for example, current satellite observations missed the increase in the smoke plume that is known to have occurred later in the day. The smoke from these fires also caused a direct environmental risk when the air became unhealthy to breathe in parts of Texas and Florida. Other parts of the US are regularly affected by smaller amounts of smoke every year, such as from the annual Canadian boreal fires and California-Oregon fires. The hourly data from EPIC can be used as an environmental warning system for regions threatened by such smoke plumes. As with TOMS aerosol data, the EPA (Environmental Protection Agency) is expected to be interested in the higher time and spatial resolution EPIC data.
EPIC aerosol data will be made available to the FAA (volcanic ash), US Park Service (smoke), EPA (smoke and dust), and others for their operational needs. The detection of volcanic ash is particularly important in the Northern Hemisphere Pacific rim region where there is frequent volcanic activity and a high density of aircraft routes. Volcanic ash plumes at 3 to 15 km have caused major damage to aircraft and in extreme circumstances could cause accidents.
