Regional Characterization of Continental Aerosols and Comparison with Satellite Data
Prof. Rudolf B. Husar
Center for Air Pollution Impact and Trend Analysis (CAPITA)
St. Louis, MO 63130-4899
Proposed Project Period: 07/01/98-06/30/00
Proposed Project Budget: $239,989
Aerosol NRA Code Y
400 Virginia Ave, SW, Suite 700
Washington, DC 20024
In Response To:
Investigations to Address the Radiative Impact of Aerosols on the Earth’s Climate
February 20, 1998
The objectives of the proposed work are: (1) to identify and delineate the major tropospheric aerosol regions of the world; (2) develop and maintain a distributed web-based regionalized aerosol information catalog; and (3) integrate and fuse satellite data with surface-based visibility and turbidity observations at about a dozen regions of the world. The project will provide a set of verified methods for the re-processing of long-term satellite aerosol data sets.
The existing observations indicate that the tropospheric aerosol pattern is a collection of regions with unique spatio-temporal pattern and aerosol properties for each region. Therefore, the construction of a global aerosol climatology database for the detection of climatic effects requires the characterization of these specific aerosol regions. AVHRR and surface-based visibility observations indicate that the Indian subcontinent has the highest regional-scale turbidity in the world. Wind-blown dust regions are evident over Sahara, Saudi Arabia, and W. China. Persistent seasonal biomass smoke is observed over Sahel, Central Africa, Amazon, Central America, Indonesia and Australia. Regional-scale industrial haze occurs over Eastern N. America, Central Mexico, SE Brazil, Central Europe, China-Japan and southern Urals. The above set constitutes the main regions over which the aerosol characteristics and satellite detection will be evaluated.
As part of the proposed work a Web-based catalog of regional aerosol information will be maintained utilizing existing metadata standards. Efforts will be made to encourage the global aerosol community to register their aerosol related data and reports in the catalog.
The methods of multi-sensory data fusion will be studied and will be applied to the generation of fused regional data sets. Special emphasis will be placed on the GOES, AVHRR, and SeaWiFS data comparison and reconciliation. A key surface-based database used in this study will be the 7000 station Global Synoptic Meteorological Network. Human visual range observations yield a semi-quantitative index of horizontal extinction coefficient. Previous work has shown that when suitably filtered and processed, the extensive visibility data are particularly suitable for aerosol regional pattern and decadal trend analysis. The visibility data will be fused and fortified with vertical spectral turbidity data from sun-photometers as well as with mass concentration/chemical composition data from filter samples to yield columnar extinction and absorption values. These will serve to complement and validate the satellite data.
The work will involve close interaction with the Science Team. The selection of the specific aerosol regions for detailed study and the incorporation of multiple surface and satellite data sets (e.g. AVHRR, SeaWiFS, and GOES) is to be guided by the Science Team. Conversely, the PI will bring to the Science Team 30 years of research experience in atmospheric aerosols. The proposed work will also benefit from an ongoing Cooperative Agreement with EPA - OAQPS that includes regional aerosol pattern and trend analysis over North America and satellite aerosol detection studies.
Objectives and Expected Significance
The objectives of the proposed work are as follows:
The significance of this project is primarily in the integration and fusion of multiple sensory data sets from surface and satellite sensors. This will provide a set of verified methods for routine re-processing of decadal aerosol data sets. The project will also provide the framework, i.e. a cataloging system that will facilitate the exchange and reuse of data sets maintained by numerous organizations. The integrated global aerosol pattern will also facilitate the adaptation of existing surface-based networks for better coverage of relevant regions.
Global Distribution of Tropospheric Aerosols and their Causal Factors
The approximate global distribution of tropospheric aerosols can be estimated from the existing satellite and surface-based observations as illustrated on the front cover of this proposal. It represents the superposition of aerosol data derived from AVHRR over the oceans (Husar et al., 1997) and surface extinction coefficients derived from synoptic visibility observations over land (Husar and Husar, 1998).
The main feature of the composite global aerosol map for July is that (1) the highest values occur in the sub-tropical zone and in the southern hemisphere; (2) the global aerosol is dominated by regional hot spots; (3) the highest optical thickness is observed over the Indian sub-continent, followed by Sahara, Saudi Arabia, and central Africa; (4) the spatial pattern changes markedly during each season (not shown).
Observational evidence indicates that the major aerosol source types are wind-blown dust, biomass burning, sea salt, biogenic sulfur, and sulfates, organics, nitrates and soot from urban/industrial sources. Each aerosol type has a specific formation and removal mechanism, characteristic size distribution, chemical composition and optical properties.
Currently we lack full understanding of the causal factors for the observed tropospheric aerosol pattern. Below is a brief qualitative discussion of these factors. Human activities that are known causes of aerosol production include fossil fuel combustion, metal smelting, as well as the burning of wood and agricultural slash, and clearing of forested land by fires. Emission of sulfur oxides from industrial sources, organics from automotive and industrial sources, as well as soot from various combustion processes are the main aerosol types from anthropogenic sources. As an index of these emission patterns the population density (Figure 1a) and the global sulfur emission density is shown in Figure 1b. The highest population density is over northern India and northeastern China, as well as north Central Europe, Japan, and Indonesia. Other regions of the world have smaller population density hotspots, such as the Washington-Boston corridor, southern California, and Mexico City.
The global sulfur emission density, Figure 1b, is highest over north Central Europe, the industrial regions of eastern North America, as well as over eastern China-Japan. Somewhat lower, but still substantial sulfur emission densities are also estimated for northern India and southern Africa. Recent observational evidence indicates that anthropogenic sulfur emissions from fossil fuel combustion may not be the main source of man-induced global aerosols. Refuse, agricultural and other burning in the developing countries may be of equal or higher significance.
Biomass burning is a major source of air pollutants such as carbon monoxide, hydrocarbons, ozone, and smoke aerosols. In fact, there is increasing evidence that the tropospheric chemistry of the global atmosphere is largely determined by the characteristics of the biomass combustion products (Fishman et al., 1996). Biomass burning from open fires can be detected by satellites (Kaufman et al., 1990), and fire frequency maps have been compiled by a number of organizations. The global fire maps (Figure 1c) indicate that fires are predominant near the equator, and over the continents in the southern hemisphere. The observations also reveal that biomass burning and the resulting emissions are highly seasonal. For example, in January fires are most prevalent over the savanna region of sub-Saharan Africa, Indochina, and western Australia. In August, the fire frequency maps show a prevailing occurrence over southern Amazonia and southern Africa. The transition seasons have their own global fire pattern. Biomass burning is caused both by anthropogenic and natural forces, with indications that in the sub-tropics most fires are ignited by humans. In some regions, such as sub-Saharan Africa the seasonal pattern of biomass combustion is repeated consistently from one year to another. In other regions, such as Indonesian islands, both the magnitude and the locations of biomass aerosol emissions varies greatly in response to drought conditions. For example, over Indonesia the El Nino years of 1982-1983, and 1997-1998, coincided with major fire events, while the in-between years were almost smoke free.
Wind blown dust constitutes a major fraction of the global tropospheric aerosol mass burden. The major known sources of wind-blown dust are sand areas over Sahara region, southern sections of Saudi Arabia, and western China. These regions consistently emit windblown dust, which is carried in the mid-troposphere up to 10,000 km from their source of origin. Smaller-scale wind-blown dust emissions exist throughout the arid regions of the world. The magnitude of the dust emission rates is known to vary by 50% or more from one year to another, depending on the drought and wind conditions. However, the magnitude of the year to year variation is not well established. In fact, it is anticipated that satellite remote sensing over the decadal time scale will illuminate the longer-term variations as done by Moulin et al., 1997 for the Sahara dust over the Atlantic.
Aerosols, particularly light scattering fine particles, have small dry deposition rates, such that in the absence of precipitation they reside in the atmosphere for several weeks and disperse over much of the hemisphere. Cloud scavenging and subsequent precipitation limit the atmospheric lifetime of particles in the lower troposphere from several days to a week. Consequently, the precipitation pattern is a major factor in determining the distance of the aerosol transport from sources to receptors. A useful parameter in this regard is the precipitation rate normalized by the columnar water vapor content. This parameter can be viewed as the water removal rate constant (Figure 1d). A high water removal (or turnover) rate constant corresponds to low atmospheric lifetime. It is argued here that regions with high water turnover rates will remove aerosol most effectively, causing lower concentrations and shorter transport distances. Conversely, regions of low water turnover times, correspond to low aerosol turnover times. Hence, over the blue colored regions of the map in Figure 1d aerosols will reside in the atmosphere for the longest time, accumulate to higher concentrations, and possibly disperse over a longer distance. This may explain the long West African aerosol plumes over the Atlantic just north of the ITCZ.
The relevance of the above discussion of relevant factors is that the global aerosol trends over past decades may have been caused more by non-industrial emissions, rather than by sulfur emissions from fuel combustion.
Subtle Features of Aerosol Radiative Effects on Climate
Traditionally the role of aerosols in radiative climate has been classified into either direct effects of light scattering and absorption, or indirect effects through aerosol induced changes to cloud microphysics, size distribution, absorption, and backscattering. However, the aerosol cloud interaction extends well beyond these "direct" and "indirect" processes. In fact, there appears to be a strong symbiotic relationship between the life cycle and effects of clouds and aerosols. Clouds are known to enhance the formation rate and the formation quantity of aerosols. Clouds and precipitation are also the key mechanisms by which aerosols are removed from the atmosphere. Hence both, the formation and fate of aerosols are influenced by clouds.
Aerosols in turn, influence the life cycle and characteristics of clouds. There is qualitative evidence that under hazy conditions, the formation of clouds is delayed or prevented altogether. The phenomenon has been observed over the continental eastern United States as well as over the Atlantic Ocean adjacent to West Africa. Conversely, in situations of clean nuclei-free atmosphere over the oceans, the addition of aerosols enhances the cloud formation as evidenced by the ship-trail phenomena.
Another radiative aerosol cloud interaction occurs when aerosol layers, such as the Sahara dust or the central African smoke plume is transported over the ocean. During such long-range transport the dust and smoke layers are typically well above the cumulus cloud layers near the ocean surface. SeaWiFS satellite data with high spectral resolution show that the dust appears yellow-brownish in color (Figure 2). Interestingly, the color of low clouds below the haze layer also appears yellow-brown, and have lower albedo than the clouds in aerosol-free atmosphere. Evidently, the white sunlight reflected from the clouds and backscattered too space is being depleted at the blue end of the spectrum. The overall reduction of the cloud brightness (albedo) further indicates that substantial light extinction is taking place as the backscattered radiation passes through the overlying aerosol layer. The latter observations suggest that not only the vertical optical depth but also the vertical distribution of aerosol is also important. It would be desirable if the aerosol data assembled for this project could contribute to the better understanding of these complex cloud-aerosol interactions.
The approach for the proposed work involves three steps, organized into three corresponding tasks:
Task 1) Identification and delineation of the main tropospheric aerosol regions of the world;
Task 2) Development of a well structured web-based shared data and knowledge base for each region;
Task 3) Integration and fusion of surface and satellite-based observations to derive the aerosol characteristics in each region.
Task one involves conceptual development on aerosol regionalization utilizing the input from the global aerosol community with special emphasis on the Science Team. Task two is an effort to contribute a web-based infrastructure for the organization, cataloging and sharing the available aerosol knowledge distributed among the community. The third task is a contribution of new data integration methodologies for the multi-sensory fusion of surface and satellite-based data sets. This last task is also intended to deliver characteristic aerosol pattern as a result of data fusion for each of the aerosol regions.
Task 1. Identification of aerosol regions
The currently available aerosol maps indicate that the global aerosol pattern is composed of a collection of regional aerosol hot spots, ranging in size between 1,000 and 10,000 km. In between these regional hot spots are aerosol regions with low values. This pattern indicates that the global tropospheric aerosol is determined by the characteristics of the individual aerosol regions rather than by global scale phenomena. However, a meaningful definition of aerosol regions and their respective geographic extent is yet to be determined. In fact, it is here proposed that one of the topics on the agenda of the Science Team should include discussion and delineation of aerosol regions. A suggested pattern of aerosol regions, based on oceanic AVHRR data was proposed by Husar et al., (1997) (attached reprint).
An aerosol region consists of the aerosol source region itself, i.e. the industrial region or the sand dunes, as well as the surrounding areas where the aerosol is transported from the sources to receptors. This latter is referred to as a region of influence. The source region itself, can be geographically delineated for primary emissions, such as windblown dust or sea salt. However, secondary aerosols (sulfates, secondary organics) are created one or two days after emission from their precursors, thus the domain of their "source region" is somewhat ambiguous. The surrounding region of influence depends on the wind pattern, as well as the aerosol formation and removal rates, i.e. aerosol atmospheric life-time. Some source regions have overlapping regions of influence with other source regions. In other words, at any given receptor, the aerosol may originate from several source regions which makes the definition of aerosol regions difficult.
From the observations, it is clear that both the magnitude of the aerosol concentration at these regions, as well as the shape and size is strongly seasonal. For example, the aerosol concentration over southern Amazonia shows a strong peak in September, while the sub-Saharan Sahel region peaks in January. Interestingly, the strongest continental aerosol patch over the Indian subcontinent shows only modest seasonality.
The understanding of the global aerosol behavior can be advanced by the study of the major aerosol source regions and their associated regions of influence. A consensus-based development of aerosol regions has tangible benefits that include: (1) organization of aerosol data into regionally structured data sets; (2) regionalization will allow the grouping of investigators and development of interactive communities based on regional interest; (3)regionalization will also facilitate more indepth data analysis and integration. For example diagnostic regional modeling and data assimilation can be conducted at much higher space, time, and chemical resolution than what is feasible on global scale using global models. (4) Finally, a regional approach is also consistent with the concept of nested monitoring networks (Hicks and Brydges, 1994) and nested models in support of environmental management.
The effort involved in this task of the proposed work involves primarily conceptual developments as well as the stimulation and facilitation of discussion by the global aerosol community that will yield a satisfactory subdivision of the global aerosol space into (possibly flexible) regions.
Task 2. Data structuring and web-based sharing for aerosol regions
We propose that the characterization of the individual aerosol regions be conducted in a structured manner following a common template for each aerosol region. The proposed template is given below:
Context for the aerosol region
Bio-geography (topography, biosphere), emissions
Human factors (population, economics, emissions)
Weather and climate (winds, clouds, precipitation)
Chemical weather and climate (SOx, ozone, hydrocarbons)
Aerosol pattern of the region
Spatial (global, regional, local perspectives)
Temporal (secular, seasonal, weekly, synoptic)
Spectral extinction, phase function, albedo (size and chemical distribution).
The purpose of the above structure is to serve as an organization and shelving system for the available data for each region. The top segment contains data relevant to the emissions, transport, and removal of aerosols. An illustration of the major factors that influence the global aerosol pattern is given in the introduction section of this proposal. Throughout the proposed project similar information contributed by the community, will be cataloged for each region.
The bottom half of the structure contains the actual aerosol observations organized by spatial distribution, including vertical profiles, temporal pattern, as well as the spectral, angular, and absorption characteristics of the regional aerosol extinction. Whenever possible, the size distribution and chemical composition information will also be included.
It is envisioned that most of the available data would be organized according to such a template. It is proposed that the above cataloging system would be shared through the World Wide Web. However, most of the data (including reports) would be stored and maintained by their respective data owners or custodians. The aerosol metadata system would be consistent with two existing standards: The DIF format for the Global Change Master Directory (DIF, 1997) and the Metadata Format of the Federal Geographic Data Committee (FGDC) (FGDC, 1997). The existence of the Science Team for the Radiative Impact of Aerosol Projects provides an excellent opportunity to review our proposed metadata classification system for aerosol relevant information. Accordingly, a substantial part of this sub-task will involve such interaction with the Science Team.
The METEOSAT-AVHRR data comparison described in the next section also provides an outstanding example that highlights the benefits of a web-based data and knowledge sharing system. The AVHRR aerosol analysis report (Husar et al., 1997) was placed on the Web well before the formal publication. The web page included links to the processed seasonal AVHRR AOT grids in a standard EOS-DIS format (HDF). While pursuing his Ph.D. at the University of Paris Cyril Moulin downloaded the "web-published" gridded AVHRR data, prepared a gridded METEOSAT data set covering the same time and spatial scale and grid resolution and performed the comparative statistical analysis shown above. The entire procedure was conducted without the active participation of Husar, the custodian of the gridded data set. However, the results turned out to be beneficial to both Moulin and Husar. It is hoped, that the proposed project will further facilitate such sharing and re-use of many analyzed and processed data sets between the entire global aerosol community. In a sense, it is suggested that as the authors and members of the Science Team prepare the publication of their papers, they also "publish" their corresponding high-level (Level 4 or Level 5) data sets also available for unrestricted re-use by the community.
The establishment of such an aerosol data and information system was recommended by the WMO-Global Atmosphere Watch workgroup on the Global Aerosol Data System (GADS), (Husar et al., 1992). A detailed rationale for the layout of a regional aerosol description of the above template is omitted here. A prototype implementation of the system can be found at the web server of the Center for Air Pollution Impact and Trend Analysis (http://CAPITA.wustl.edu/GAIN). In this task, our CAPITA group will draw upon our past experience from maintaining the interactive website for the Ozone Transport and Assessment Group (OTAG) (http://CAPITA.wustl.edu/OTAG).
Task 3. Data fusion and derivation of aerosol characteristics
Tropospheric aerosols are being detected by a multiplicity of space-borne, airborne, and surface-based sensors. Each sensor device provides a unique view into the multidimensional aerosol data space and apply specific sampling strategies. However, none of the sensors can describe the aerosol completely as it is distributed over at least six independent dimensions: space (x,y,z), time, particle diameter, and chemical composition.
Remote sensing instruments tend to detect integral properties of aerosols. The nature of the integral depends on the sensor. For example a downward looking radiation sensor integrates over four major dimensions. Satellite remote sensors detect only integral properties of aerosols. As Rozenberg (1968) has pointed out these integrals extend over four major scales and dimensions of the aerosol data space: (1) Light scattering by individual particle of known shape, size and chemical composition, (2) integral over different particle sizes of specific aerosol composition, (3) summation over the different aerosol composition which can yield light scattering in specific volume elements and finally (4) integral over the aerosol vertical profile. Existing aerosol detectors do not necessarily collect aerosol data in a form that is directly usable for the assessment of the effects. For example, direct radiative effects on climate depend on spectral extinction, the aerosol phase function, and the single scatter albedo integrated over an atmospheric path. On the other hand, the remote aerosol sensor typically detect path radiance (Kaufman et al., 1997).
In order to make existing, largely fragmented and incompatible aerosol data sets more useful, it is necessary to combine, reconcile and transform these into a form relevant to the specific effects of concern. The proposed work intends to investigate the principles of aerosol data fusion and apply those principles to the fusion of regional aerosol datasets. The proposed research will combine data from multiple aerosol sensors to generate a global multidimensional aerosol data set. An illustration of data integration and reconciliation is given below.
Comparison of AVHRR and METEOSAT derived AOT
This summary is a quantitative comparison of aerosol optical depth derived from the METEOSAT geo-stationary satellite sensor (Moulin et al., 1997) and from the AVHRR polar orbiting sensor (Husar et al., 1997). This section illustrates the benefits and problems associated with the use of different aerosol sensors and retrieval methodologies. The geographic area for the comparison and the AOT seasonal contour maps (June 1989-June 1991) for the two data sets are shown in Figure 3. Qualitatively, the two data sets show a good general agreement with respect to the magnitude, location, and shape of the aerosol plume. A statistical comparison of the corresponding seasonal gridded values is also shown in Figure 3. The correlation coefficient ranges between 0.82 in the summer and 0.91 in spring and winter. For many remote sensing instruments such correlation might be considered adequate. However, the seasonal scatter charts reveal that the slope of the two correlation lines changes seasonally. In the winter season the AVHRR is about 25% higher than the METEOSAT (slope is 1.25), while in the summer the slope=0.57. In fact, the scatter chart for the summer season indicates a non-linear relationship with evidence of AVHRR "saturation" at high AOT values.
The two data sets are derived using different sampling techniques. The METEOSAT AOT is mapped daily for the entire region. The AOT values over cloudy areas were interpolated (filled in) based on the values measured at the surrounding cloud-free areas. This procedure is aimed at obtaining a complete aerosol sample but suffers from the uncertainties of interpolation. The AVHRR-derived AOT values are based on averaging of the available cloud-free data samples. Throughout the sampling region about half of the days were cloud-free. The cloud-free fraction is even lower in the southern segment of the region near the cloudy ITCZ belt of West Africa. Hence, the AVHRR samples are biased toward the cloud-free conditions.
The comparison of the spatial pattern also indicates that during the summer in the southern zone between 5-10° N the METEOSAT optical depth is 20-30% higher than the AVHRR values. In the winter the reverse is true; AVHRR values throughout the region are about 25% consistently higher than the corresponding METEOSAT observations. It is also worth noting that in each season there are about a dozen outlier grid points that deserve clarification. These outliers are asymmetric; they tend to occur as excessively higher AVHRR values.
A further difference between the two aerosol maps appear in the summer over the Atlantic. The AVHRR results show an elongated aerosol plume between 10-30° N that reaches the East Coast of South America. The METEOSAT results show a shorter aerosol plume (stronger decay with distance).
At this time, we do not possess an explanation for any of these deviations. However, these deviations constitute clues that either or both of the AOT values are inadequate. The clues derived from the multi-sensory analysis can be combined with other knowledge to pursue the reconciliation. For instance, it is known that the spatial region of interest is dominated by wind-blown dust from Sahara in the summer, while in the winter smoke from biomass burning over the sub-Saharan savanna is also a contributor. The size distribution and optical properties of the two aerosol types are markedly different and maybe responsible for the seasonal deviations.
The unexpectedly high outlier values in the AVHRR data suggest possible cloud contamination Alternatively, it is a consequence of poor sampling statistics due to low number of cloud-free days. The shorter length of the African aerosol plume derived from METEOSAT may possibly be attributable to "edge effect" in the field of view: the geostationary satellite observes the aerosol over South America at a highly slanted angle and obscuration by clouds and the large airmass may be significant.
The main purpose of discussing the AVHRR-METEOSAT comparison in such detail was to illustrate the procedures and reasoning for inter-satellite data comparison. As part of the proposed work the AVHRR-GOES data comparison will be continued and the differences between the two data sets will be reconciled. Also, the limitations and the applicability of the two data sets will be established. Unique features of the GOES data set are the hourly time resolution and the aerosol illumination under different sun angles throughout the day. The hourly GOES data will be used to examine the nature of the aerosol-cloud interaction throughout the diurnal cycle. The changing sun-aerosol-sensor angle will be evaluated for retrieval of the aerosol phase function. Finally, the "twilight zone" will be examined to extract aerosol height information. As the sun sets or rises, aerosol layers are illuminated differently.
As a final output of this work segment, a methodology will be proposed for the processing and integration of AVHRR and GOES data sets. With this methodology delivered and reviewed to the Science Team, the processing of the long-term AVHRR and GOES data sets will be possible by NASA.
Figure 4 illustrates global aerosol maps produced by the TOMS absorbing aerosol sensor (Herman et al., 1997), AVHRR sensor (Husar et al., 1997) and maps of aerosol path radiance derived from the Coastal Zone Color Scanner (CZCS) sensor obtained from the Goddard Distributed Data Archive. Additional aerosol maps are available based on the short-lived POLDER sensor. The data from these sensors should be inter-compared and evaluated in a manner similar to the METEOSTAT-AVHRR comparison. We anticipate that such comparisons will be conducted by other investigators. The CAPITA group will participate in the process as needed.
More recently (January 1988) a high quality aerosol mapping product is available from the polar orbiting SeaWiFS sensor. Among the 8 near visible spectral channels of SeaWiFS are red, green, and blue, such that the spectral data can be combined into true color images as in digital photography. The available true color images are of immense benefit in identifying qualitative features of the aerosols and clouds from the satellite "photographs". For instance, whether the haze appears yellow, blue, or gray, depends on the aerosol size distribution and its absorbing properties (Figure 2).
As part of this proposed work, the SeaWiFS data set will be used extensively to compare the derived AOT values to GOES and other satellite data as well as to surface-based observations. The SeaWiFS data set will also be used to explore the limitation of existing retrieval algorithms and to evaluate new aerosol retrieval algorithms over land. The land-based retrieval algorithms to be evaluated will be based largely on the methods summarized by Kaufman et al. (1997). The exploratory land-based algorithms will be based on the excess path radiance over dark land surfaces; loss of contrast in surface features and discoloration of surface features due to haze. The application of our aerosol retrieval and haze free image reconstruction method as applied to a LANDSAT image is available through a web page (http://CAPITA.wustl.edu/CapitaReports/LANDSAT/Retrieve.html).
Visibility and Turbidity Observations
Observations of visual range by humans at meteorological sites can be used as a semi-quantitative index of the extinction coefficient at the surface. The visibility reports are hourly and exist over the populated regions of the world. Data are collected during both cloudy and cloud-free conditions. However, the surface visibility data are an imprecise surrogate for extinction coefficients particularly during humid and precipitating weather conditions. The visibility data are best suited for exploratory analysis, long-term trend studies, and as aerosol surrogates following calibration with locally collected aerosol samples (Husar and Wilson, 1993) (attached reprint).
In this research, we will utilize the 9000 station network of surface meteorological observations. The data set, including the data quality control and filtering procedures, as well as preliminary results of analysis are described by Husar and Husar, 1998. The seasonal maps of global continental horizontal extinction coefficient based on 7000 valid visibility monitoring sites are shown in Figure 5. The data point to the revealing observation that the highest regional scale haze exists over the Indian subcontinent. Furthermore, the continental haziness is higher over the subtropical belt and the Southern Hemisphere that over the industrial northern latitudes between 30-60° N.
Sun photometers measure the vertical aerosol optical depth at multiple spectral bands, yielding information on aerosol size distribution. Sun photometers have poor spatial coverage, and deliver aerosol data only during cloud-free conditions. They do not reveal the vertical aerosol structure. The absolute calibration (Langely plots) requires substantial effort. Nevertheless, sun photometers are among the most common ground truth instruments for the verification of space-borne aerosol sensors because they provide spectrally resolved vertical optical depth. In the proposed work, we will draw upon the global sun photometer network operated by NASA (Holben et al., 1991).
Data fusion methodology
This research will also examine the concept of environmental data fusion from an informatics point of view. According to Luo and Kay (1992), multi-sensor data fusion refers to any stage in the integration process that involves an actual combination (or fusion) of different sensory information into one representational format. The fusion can take place at different levels of abstractions: signal (raw voltages), pixel (cloud filters), feature (aerosol plume), and symbol (aerosol type).
The topic of sensory data fusion is under intense study in physiology and neural sciences, as well as their artificial counterpart in robotics. This research will draw directional and conceptual guidance as well as specific data fusion techniques in works such as "Data Fusion in Robotics and Machine Intelligence" (Abidi and Gonzalez, 1992).
A key aspect of multi-sensory information fusion is that it incorporates not only the information supplied by the sensors but also knowledge that has been generated previously. For instance, the knowledge that aerosol size distributions are generally bimodal facilitates the proper fusion of angular scattering and light transmission data into a size distribution. In the proposed aerosol data integration work, we will continue exploring the potential utility of the human sensory information processing as an analog and "blueprint" for multi-sensory aerosol signal processing.
Abidi M.A. and Gonzalez R.C. (1992) Data Fusion in Robotics and Machine Intelligence, Academic Press, Boston.
Directory Interchange Format (DIF) Writer's Guide, Version 5.0a, Web URL: http://gcmd.gsfc.nasa.gov/difguide/difman.html, 1997.
The Federal Geographic Data Committee (FGDC), Content Standard for Digital Geospatial Metadata, Version 1.0, Web URL: http://www.fgdc.gov/Metadata/metav1-0.html, 1997
Fishman J., Brackett V.G., Browell E.V., and Grant W.B. (1996) Tropospheric ozone derived from TOMS/SBUV measurements during TRACE A. J. Geophys. Res., 101, D19, 24069-24082.
Hameed S. and Dignon, J. (1992) Global Emissions of nitrogen and sulfur oxides in fossil fuel combination 1970-1986. J. Air Waste Manage. Assoc., 42, 159-163.
Herman J.R., Bhartia P.K., Torres O., Hsu C., Seftor C., and Celarier E. (1997) Global Distribution of UV-absorbing aerosols from Nimbus 7/TOMS data. J. Geophys. Res., 102, D14, 16911-16922.
Hicks B.B. and Brydges T.G. (1994) A strategy for integrated monitoring, Environ. Management, 18,1-12.
Holben B., Eck T.F., Fraser R.S. (1991) Temporal and spatial variability of aerosol optical depth in the Sahel region in relation vegetation remote sensing, Int. J. Remote Sens., 12, 1147-1163.
Husar R.B., Stowe L.L., and Deepak A. (Eds.) (1992) Report of the meeting of experts on global aerosol data system (GADS). World Meteorological Organization, Global Atmosphere Watch No. 73, WMO/TD-No. 438.
Husar, R.B. and Wilson W.E.. (1993) Haze and sulfur emission trends in the Eastern United States. Env. Sci. Technol., 27, 12-16.
Husar, R.B., Prospero J.M., and Stowe L.L. (1997) Characterization of tropospheric aerosols over the oceans with the NOAA Advanced Very High Resolution Radiometer Optical Thickness Operational Product, J. Geophys. Res. 102, D14, 16889-16909.
Husar, R.B. and Husar, J.D. (1998) Global Distribution of Continental Haziness, Web URL:http://capita.wustl.edu/CAPITA/CapitaReports/GLOBVIZ/GLOBVIS1.html
IGPB-DIS,Web URL: http://www.mtv.jrc.it/projects/fire/products.html
Kaufman Y.J., Tucker C.J., Fung I. (1990) Remote sensing of biomass burning in the tropics, J. Geophys. Res., 95, D7, 9927-9939.
Kaufman Y.J., Tanre D., Gordon H.R., Nakajima T., Lenoble J., Frouin R., Grassl H., Herman B.M., King M.D., and Teillet P.M. (1997) Passive remote sensing of tropospheric aerosol and atmospheric correction for the aerosol effect, J. Geophys. Res., 102, D14, 16815-16830.
Luo R.C. and Kay M.G. (1992) Data fusion and Sensor Integration: State of the Art 1990s. In Data Fusion in Robotics and Machine Intelligence (Abidi and Gonzalez, Editors), Academic Press, Boston.
Moulin C. (1997) Transport atmospherique des particules africanes sur la Mediterande et Atlantique: climatologie satellite (METEOSTAT, 1983-94) et relations avec le climat. PhD Thesis, Univ. of Paris, VI, 1997.
Moulin C., Guillard F., Dulac F., and Lambert C.E. (1997) Long-term daily monitoring of Saharan dust load over ocean using Meteosat ISCCP-B2 data. 1. Methodology and preliminary results for 1983-1994 in the Mediterranean. J. Geophys. Res. 102, D14, 16947-16958.
NASA/GISS (1997) Web URL http://www.giss.nasa/gov
Rozenberg G.V. (1968) Optical investigations of atmospheric aerosol. Sov. Phys. Uspekhi 11, 353-380.
RSS (1998) Web URL: http://www.ssmi.com/ssmiBack.html
Figure 1.Global pattern of parameters that are relevant to the global aerosol analysis. a) Population density over the world based on data published by NASA-GISS. b) Global anthropogenic sulfur emission density based on data from Hameed and Dignon (1992). c). Global distribution of fires in January and August, 1993 (IGPB-DIS). d) Global distribution of water turnover rate (precipitation rate/columnar water vapor, 1/hr). Data are based on the SSM/I microwave imager (RSS, 1998)
Figure 2.SeaWIFS images of haze over West Africa for February 18, 1998. Note the brown coloration of dust as well as the clouds.
Figure 3.Quantitative comparison of AVHRR (left-Husar et al., 1997) and METEOSAT (Moulin et al., 1997) aerosol optical thickness adjacent to West Africa performed by (Moulin, 1997).
Figure 4.Global aerosol pattern derived from TOMS, AVHRR and CZCS sensors for July.
Figure 5.Seasonal distribution of horizontal extinction coefficient derived from the 7,000 station Global Synoptic Meteorological Network. For the March, April, May map the station locations and the magnitude of the extinction coefficients are also shown.