Corresponding author: M. D. Zuluaga, Department of Atmospheric Sciences, University of Washington, 408 ATG Bldg., Box 351640, Seattle, WA 98195, USA. (firstname.lastname@example.org)
 It is well known that aerosols impact climate directly through modifying the radiation budget and indirectly through the modification of cloud processes. Yet, in spite of an improved understanding of the various roles that aerosols play in climate, there still exists uncertainty in their spatial and temporal distributions and their relationship to atmospheric dynamic. Here, we use the Aerosol Index from the Total Ozone Mapping Spectrometer and Aerosol Optical Depth from the Moderate Resolution Imaging Spectroradiometer, in conjunction with atmospheric and oceanic satellite observations and reanalysis data sets to investigate aerosol-environment relationships and interactions over the tropical Atlantic Ocean. Spectral and composite analyses of surface temperature, atmospheric wind, geopotential height, outgoing longwave radiation and precipitation, together with the climatology of aerosols, provide insight on how the variables interact. Different modes of variability, especially on intraseasonal time scales, appear as strong modulators of the aerosol distribution. In particular, we investigate how two modes of variability related to westward propagating African Easterly Waves affect the horizontal and vertical structure of the environment and thus the aerosol distribution. The pattern of propagation of aerosol load shows a strong correspondence with the progression of the atmospheric and oceanic synoptic conditions that have mobilized dust over the African continent and advect it over the Atlantic Ocean. We extend previous studies related with dust variability over the Atlantic region by evaluating the performance of the long period satellite aerosol retrievals in determining modes of aerosol variability. Results of this work are described as useful in allowing a more precise understanding of the response of the energy budget, precipitation and atmospheric circulation to changes in aerosol loading.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 A major challenge for climate science is the accurate representation of the water cycle and the exchange of heat between the atmosphere, the ocean, the land surfaces and space. Both the water cycle and global energy distribution are highly coupled to the large-scale atmospheric circulation pattern, which in turn is strongly modulated by the distribution of clouds and rainfall. Clouds and rainfall influence regional circulations by redistributing energy. The feedbacks between clouds and the large-scale circulation, complicated by the impacts of aerosols on radiation budget and cloud properties, represent one of the largest sources of uncertainty for future climate projections [Stephens, 2005]. However, there are large uncertainties in the impacts of aerosols owing to high spatial and temporal aerosol load variability and heterogeneous distribution of aerosol species [Kaufman et al., 2002; Intergovernmental Panel on Climate Change, 2007].
 Aerosols modulate the radiation budget and the hydrological cycle both directly and indirectly and, thus, have the potential of playing an important role in weather and climate variability [Ramanathan et al., 2001]. Aerosols interact directly with radiation through scattering and absorption in the atmospheric column. The forcing effect can either cool or warm the surface and the atmosphere depending on the nature and location of the aerosol load [e.g., Twomey et al., 1984; Charlson et al., 1992; Haywood et al., 1999]. Aerosols modify the abundance (cloud lifetime effect [Albrecht, 1989]) and optical properties (cloud albedo effect [Twomey, 1977]) of clouds indirectly impacting both the longwave and solar radiation streams through altering the size and growth rates of clouds. In addition, significant impacts at shorter timescale and on regional scales have been noticed [e.g., Coakley and Cess, 1985; Miller and Tegen, 1998; Evan et al., 2009]. For example, the reduction in surface solar radiation imposed by aerosols in dust plumes produces a very strong convectively suppressing inversion. The growth of an inversion changes the vertical atmospheric temperature gradient, limiting precipitation throughout stabilization of the air column and reducing surface shortwave radiation. Evidence suggests that this aerosol variability could even inhibit the formation of tropical cyclones and also reduce their intensity [e.g., Dunion and Velden, 2004; Lau and Kim, 2007].
 All of these aerosol effects occur on different temporal and spatial scales. Thus, an assessment of the impact of aerosol variability effects on the energy radiation budget, and the global circulation and climate requires long period measurements of aerosols at both regional and global scales. Through the use of absorbing aerosol detection at ultraviolet wavelengths by the Total Ozone Mapping Spectrometer (TOMS) and Ozone Monitoring Instrument (OMI), flying on several platforms since 1978, it had become possible to compile a long-term aerosol record. Although the TOMS-OMI instruments were designed originally for the remote sensing of ozone, they have been used to monitor absorbing aerosol transport over land and ocean [Herman et al., 1997]. The global features of aerosol distribution using the TOMS-OMI satellite platforms have been used extensively to document aerosol variability and its effects on climate [e.g.,Torres et al., 1995; Prospero et al., 2002; Jeong and Li, 2005; Torres et al., 2007; George et al., 2008]. The TOMS/OMI data has been complemented by the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor that was designed specifically for aerosol retrievals. The information provided from this platform have expanded the near-infrared aerosol detection to include non-absorbing aerosols (e.g., sulfates and pollution) and several others aerosol properties with greater accuracy [Remer et al., 2005]. A description of the global aerosol distribution has been complemented by the launch in 2006 of the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) which is an aerosol lidar profiler able to retrieve columnar aerosol type and load [Winker et al., 2009].
 Dust is by far the most abundant and ubiquitous aerosol type. North Africa is recognized as the largest dust source providing around the 50% of the global annual mass [e.g., Prospero et al., 2002]. Westward intrusion of dust-laden air from the African continent into the tropical North Atlantic Ocean is a frequent phenomenon during the boreal summer [Carlson and Prospero, 1972]. The synoptic to local atmospheric processes involved have been characterized into conceptual models [e.g., Carlson and Prospero, 1972; Karyampudi and Carlson, 1988] describing the meteorological conditions suitable for dust mobilization. In synthesis, the model describes how dust-laden heated air emerges from West Africa within a series of large scale anticyclonic eddies. This dusty air moves above the cool and moist trade-wind inversion layer elevating the dust air mass to 600–800 hPa layer [Carlson and Prospero, 1972; Karyampudi and Carlson, 1988]. The vertical and horizontal structure of this dust plume, usually referred as Saharan Air Layer (SAL), has been extensively studied and described. For example, Karyampudi et al.  used Meteosat satellite overpasses and reanalysis data to describe a detailed synoptic situation for the conditions suitable for dust mobilization in the tropical Atlantic Ocean. This analysis validated many of the characteristic features of the Saharan dust plume conceptual model [i.e., Carlson and Prospero, 1972]. However, the Karyampudi et al. study was based on a single week of data, limiting the characterization of overall temporal nature of the phenomena.
 The objective of this work is investigate how atmospheric processes modify the summer time dust aerosol load over the tropical Atlantic region and how the environment, in turn, is impacted by aerosol variability. Whereas a number of different studies have recognized the role of AEWs in the modulation of African desert dust transport [Jones et al., 2003; Barkan et al., 2004; Washington and Todd, 2005; Knippertz and Todd, 2010; Huang et al., 2010; Jury and Santiago, 2010] we propose a number of significant extensions. We focus on the spectral representation of maximum aerosol load events as retrieved by the satellite data sets. We evaluate the performance of the long series of satellite aerosol retrievals in the determination of the different modes of variability found over the tropical Atlantic region. We also use reanalysis and satellite observations of atmospheric and oceanic variables to show how the two modes of aerosol variability associated with AEW activity affect the horizontal and vertical structure of the environment and how those impacts are represented in the proposed models for aerosol variability over the Atlantic Ocean [e.g., Carlson and Prospero, 1972]. The analysis is complemented with the study of CALIPSO overpasses that help to characterize the vertical extent of the aerosol maxima over the region.
 The paper is organized in the following manner: Section 2 presents a description of the data and introduces the analysis methodology. Section 3 describes and analyzes the horizontal distribution of aerosols in relation to the AEW variability and the impact of aerosols on sea surface temperature (SST), outgoing longwave radiation (OLR) and precipitation. Section 4 extends the analysis to consider the impact of aerosols in the vertical atmospheric distribution of temperature, humidity and wind direction. Section 5 uses CALIPSO data to analyze the vertical distribution of events of extreme aerosol loading. Finally, section 6 provides an overall summary and lists a number of conclusions.
2. Data and Methodology
2.1. Aerosol Data
 Measurements of aerosol loading retrieved from passive satellite-based sensors have been available since the late 1970s. The TOMS Aerosol Index (AI), used in this study, is a measure of the amount of backscattered UV radiation from an atmosphere containing aerosols (i.e., from observations) differing from the backscatter of a pure molecular atmosphere (i.e., from model calculations) [Herman et al., 1997]. The AI is positive for absorbing aerosols such as mineral dust, smoke from biomass burning and volcanic ash. However, the AI cannot detect non-absorbing aerosols such as sea salt and sulfates. The AI is affected by sub-pixel cloud contamination and by the altitude of the aerosol layer. It is incapable of detecting absorbing aerosols below 2 km elevation [Herman et al., 1997]. Although the AI provides a qualitative characteristic of aerosols, it is helpful in identifying dust sources and transport routes [Sokolik et al., 2001]. The TOMS Version 8 data from the Nimbus 7 platform (1979–1993) is the most recently reclassified and recalculated version of daily records. The data was obtained from the NASA Atmospheric Composition web page (http://macuv.gsfc.nasa.gov/) covering the entire tropical band between 45°N and 45°S at 1° × 1.25° latitude-longitude resolution. The AI values less than 0.5 are treated as “missing values” because contamination of the variable by sea-glint (O. Torres, personal communication, 2009).
 In addition, aerosol optical depth (AOD) data from the MODIS sensor aboard the NASA Earth Observing System (EOS) Aqua platform is used to complement the analysis. The AOD is a measure of the transparency of an atmosphere column containing aerosols and can be associated directly with the amount of aerosols in the atmosphere. The retrieval of AOD from the MODIS sensor uses two independent methodologies (i.e., “Dark Target” algorithm for land and oceanic regions, and the “Deep Blue” algorithm for land regions) using seven of the spectral bands between 0.47 and 2.130 μm sensitive to aerosol content in the atmospheric column [Kaufman et al., 1997; Tanré et al., 1997; Hsu et al., 2004]. Compared with TOMS-AI, which is only capable of retrieving absorbing aerosols, MODIS AOD measurements are sensitive to both absorbing and non-absorbing aerosols because of the use of multiple wavelength retrieval. In addition, the MODIS sensor is able to mask out clouds with a greater accuracy than retrievals from TOMS or AVHRR satellites [Remer et al., 2005]. Daily AOD at 0.55 μm from Aqua platform (MYD08_D3 collection 5.1) from July 2002 until December 2010 downloaded from the Atmosphere Archive and Distribution System (LAADS, http://ladsweb.nascom.nasa.gov/) is used in the study. In order to use the most recent and advanced data, we utilized AOD from the “dark target” algorithm for oceanic regions and the “Deep Blue” for land regions covering the entire globe at an equal-angle latitude-longitude grid with a horizontal resolution of 1° × 1°.
 In order to have a vertical view of the structure of aerosol loading, the nadir-pointing lidar system of the CALIPSO platform is used. Profiles of aerosol extinction coefficient at 532 nm are used as a measure of vertical distribution of aerosol. Details on the CALIPSO science products are given byWinker et al. . Aerosol profile data product (Lidar level 2 version 3.1) at different times of the day from June 2006 until July 2009 were acquired from the NASA Atmospheric Science Data Center at http://eosweb.larc.nasa.gov/. The data are provided on a 5 km along-track horizontal and 60 m vertical resolutions up to 200 hPa. Despite the very high vertical resolution, CALIPSO mission provides transects of the vertical distribution of aerosol of approximately 2000–2500 km horizontal interval at midlatitudes making it difficult to capture the continuous spatial structure of the vertical distribution of African dust. For this reason, the CALIPSO profiles were used in this work only for particular case studies.
2.2. ERA40, ERA Interim, SST-OI, OLR and GPCP Data Sets
 Three-dimensional 6-hourly data for zonal, meridional and vertical wind components, along with geopotential height, potential temperature, and specific humidity were obtained from the European Center for Medium-Range Weather Forecasts (ECMWF). Both reanalysis data sets ERA40 [Uppala et al., 2005] and ERA interim [Berrisford et al., 2009] were used to characterize large-scale atmospheric circulation. Specifically, 12:00 local time data were archived to match the satellite overpass (i.e., TOMS-MODIS platforms). Data at 1° × 1° horizontal resolution and at seven atmospheric levels from 1000 hPa to 200 hPa were used. These two reanalysis data sets were used separately to cover the TOMS and MODIS periods from July 1978 to July 1993 for the first set of aerosol products and from January 2002 to December 2010 for the second set.
 This study also uses daily microwave optimally interpolated SST data derived from the TMI sensor on board of the TRMM satellite and processed by Remote Sensing Systems (RSS) [Gentemann et al., 2004]. These data are obtained from the RSS website and correspond to daily SST analysis on a 0.25° latitude-longitude grid (re-interpolated to 1° × 1° to match the grid in MODIS AOD data). The use of microwave SST is favored because its calculation is unaffected by the presence of aerosols and can be retrieved in the presence of clouds [Gentemann et al., 2004]. In conjunction with SST, interpolated daily outgoing longwave radiation (OLR) data at 2.5° × 2.5° global grid resolution [Liebmann and Smith, 1996] were downloaded from the Climate Diagnostics Center (CDC) from the NOAA ESRL web page (http://www.esrl.noaa.gov/). The OLR estimates used in this study come from NOAA polar-orbiting satellites to distinguish areas of deep tropical convection. To complement the information of tropical precipitation, we also use daily rainfall intensity from the Global Merged Precipitation analysis (GPCP) [Huffman et al., 2001] of the Global Energy and Water Cycle experiment (GEWEX, www.gewex.org/gpcp.html). Because of the limited period covered by the SST and rainfall data sets (i.e., from 2002 to 2010), only data collocated over the MODIS Aqua period with a horizontal resolution of 1° × 1° were used.
2.3. Spectral Analysis
 In order to characterize the temporal and spatial evolution of aerosols relative to AEW variability and their relationship with atmospheric variables, a region near to the coast of Africa in the center of the climatological dust corridor was selected (15°N—17°N, 22°W—20°W; box in Figure 1(top)). Daily time series for the boreal summer months of June, July, August and September of averaged TOMS AI (‘AI’, hereafter) and MODIS-Aqua AOD (‘AOD’, hereafter) over the selected region were computed. The time periods used for this analysis are 1979—1993 and 2002—2010, respectively. To obtain a representative sampling of aerosol data, a 3-day running mean over both gridded satellite data sets was applied to minimize the impact of missing data due to sensor swath coverage. Filtered time series were computed using a band-pass Fourier filter retaining time-band periods in 5 to 15 and 10 to 30 days range to match the two modes of variability of AEW over the region [Agudelo et al., 2011]. The time-band day ranges were chosen in order to aggregate time series periodicities in the two shorter time scales that the aerosol time series provides. Several combinations of time-bands were used and the best matching periodicity relative to the two different AEW modes were the 5–15 and 10–30 day bands. To illustrate the technique used to identify the reference days to construct the composites,Figures 2a and 2b show the averaged AOD time series over the reference region and its corresponding Fourier spectra, respectively. A clear annual cycle is observable in both time series and a noticeable significant short time scale periodicities (i.e., less than 30 days and significant compared to a red noise process). Figure 2c shows the filtered time series retaining periods in 5 to 15 day band. Figure 2d shows the corresponding Fourier spectra of the filtered time series. It is found that that the retained periodicities in the 5–15 day band are significant compared to a red noise process. The next step is the selection of positive values greater than +1 standard deviation in the filtered time series. The date of the maximum value in each period is designated as the day 0 (red diamonds: Figure 2c) allowing the computation of composites. Only the June–September summer period was considered.
 To construct the composites, anomalies of gridded AI and AOD data were computed with respect to a 30-day running mean of the daily climatology in each of the analyzed periods. In addition, wind velocity and direction, geopotential height, potential temperature, specific humidity, SST and OLR anomalies were computed for periods matching the aerosol data sets. For the TOMS daily data, 161 days of maximum AI were found in the 5–15 day period and 80 days in the 10–30 day band. In similar way, MODIS indicated 76 and 44 days of maximum AOD in the 5–15 and 10–30 day period, respectively. The composite periods are available as supplemental information at (http://webster.eas.gatech.edu/papers.html).
3. Horizontal Composite Analysis
Figure 3shows a longitude-time plot of TOMS AI and MODIS AOD anomalies for the 5–15 day and 10–30 day period bands. Day 0 in each plot represents a maximum in aerosol loading over the selected study region and composites were created by averaging the AI and AOD anomalies from −8 to 8 days around day 0. In addition, positive geopotential height anomalies at 700 hPa are also plotted (solid contours). Two modes of westward aerosol propagation are evident in both data sets, one with a period near 5–7 days (for the 5–15 day band filter) and another with a period near 9–11 days (for the 10–30 day band filter). There is a good relationship between positive 700 hPa geopotential height anomalies with the positive aerosol anomalies and negative heights with negative aerosol anomalies. These two regimes represent the wave speeds of the two AEW forms crossing the region. The aerosol anomalies inFigure 3, using the AI, show a wave speed of approximately 8 deg/day (5–15 day band, Figure 3 (top left)) and 6 deg/day (10–30 day band, Figure 3 (right top)); speeds very similar to the AOD anomalies (Figure 3, right).
Figures 4 and 5show the extension of the analyses to the longitude-latitude plane. Composites are constructed in the same way as inFigure 3 and show the horizontal distribution of aerosol anomalies using AI (Figure 4) and AOD (Figure 5) analyses for dates with the minimum and maximum values. Wind vectors and geopotential height anomalies at 700 hPa using ERA-40 and ERA-interim are also shown for the same composite times for the TOMS and MODIS periods, respectively. The 700 hPa level was selected because the average transportation of the dust takes place above the humid trade wind air in the 600–800 hPa layer [Carlson and Prospero, 1972; Kaufman el al., 2005] and because the AI is mostly sensitive to aerosols above 2 km. Progressing from the top to bottom in Figure 4, negative AI anomalies are seen over the study region at 4 days before the aerosol loading maximum (for the 5–15 day band) and 6 days before (for the 10–30 day band) with collocated cyclonic winds and negative geopotential height anomalies. As time progresses toward the maximum in AI at day 0, the aerosol, wind and geopotential height anomalies move westward with a notable increase in magnitude. This composite progression displays westward propagation of aerosols with anticyclone wind and positive geopotential height anomalies, consistent with the results of the Carlson and Prospero  model of dust transport over the Atlantic Ocean. At day 0, a discernible anticyclone circulation is apparent in both composites over the coast of West Africa. As the anticyclonic circulation moves westward, the aerosol loading diminishes moving toward AI negative anomalies over the reference region. This occurs after 4 days in the 5–15 day band and 6 days in the 10–30 day band with a considerable reduction in the magnitude of the geopotential height anomalies as well as a reversal in wind anomaly direction. The progression in MODIS AOD anomalies in Figure 5 is remarkably similar to the progression of AI anomalies shown in Figure 4. The timing of the change between negative to positive AOD and geopotential height anomalies matches the timing of change evident in the AI anomalies. These similarities reaffirm the aerosol progression in accord with the synoptic variability and the speed of the two AEW regimes.
 As the dust is transported over the desert surface toward the ocean, the air layer in which the aerosols are embedded is warmer and drier than the normal tropical atmosphere [Carlson and Prospero, 1972]. The aerosol layer helps to produce a very strong suppressive inversion above the moist trade wind air limiting precipitation, humidity and reducing surface shortwave radiation [Miller and Tegen, 1998; Foltz and McPhaden, 2008; Evan et al., 2009]. To calculate the environmental impact produced by the variation of aerosol loading on the surrounding atmosphere and ocean, SST, OLR and GPCP precipitation anomalies were calculated. This analysis was done only for the MODIS period because data availability. Figure 6shows composites of SST for −6, 0 and +4 days; and OLR and GPCP precipitation for −6, 0 and +6 days; days of the maximum in MODIS AOD over the study region. The composites were calculated using the aerosol time series with a Fourier filter in the 10–30 day band. For each composite, a contour representing positive and negative anomalies of AOD is also plotted. A reduction in SST anomalies is found over the region affected by the dust plume (i.e., dust corridor) from −6 days to +4 days of the maximum in AOD over the west coast of Africa. Such a reduction in SST can be attributed as a result of the high solar absorption in the dust layer that reduces shortwave radiation, which would otherwise be absorbed by the upper ocean. The OLR anomalies (used in this case as a proxy for convection) present an increase in values from −6 days to 0 days of the maximum in AOD and the precipitation anomalies show a reduction from −6 to 0 days of the maximum in AOD. We hypothesize that the positive anomalies of OLR in the presence of maxima in aerosol are deviations toward the presence of fewer clouds with low-temperature top levels (i.e., less convection). This reduction in precipitation as AOD anomalies increase would be consistent with a negative radiative forcing at the surface imposed by absorbing aerosols. The forcing is being balanced by a reduction in upward thermal radiation associated with a decreased surface temperature. A reduction of turbulent fluxes of sensible and latent heating into the atmosphere produces a stabilization of the air column within which vertical motion is inhibited thus reducing precipitation consistent with earlier studies [Coakley and Cess, 1985; Miller and Tegen, 1998].
Figure 7provides a summary of the results showing the time series evolution of the composited anomalies averaged over a box centered in the reference region (i.e., 14°N-18°N, 23°W-19°W).Figure 7 (left) shows the evolution of anomalies in the 5–15 and 10–30 day bands based on TOMS AI data. Figure 7 (right) is based on MODIS AOD data set. Similar patterns emerge for all of the anomaly evolution series described earlier. As positive anomalies of aerosol loading increase toward day 0 (maximum aerosol loading), there is a reversal from easterly to westerly in zonal wind direction, a change of sign from negative geopotential height values (i.e., low pressure) to positive values (high pressure) and an increase toward a maximum in wind speed. As aerosol loading increases there is also a decrease toward minimum values of SST and rainfall rate and toward a maximum in OLR (lack of deep convection). Is important to mention that all the averaged time series of the different variables presented in Figure 7show two different time-scale variability related to the two wave speeds of the AEW crossing the region.
4. Vertical Composite Analysis
 In order to analyze the vertical state of the atmosphere during the evolution of aerosol loading, composite analysis of the latitude-height structure of zonal, meridional, and vertical wind speed, geopotential height, potential temperature and humidity anomalies are constructed.Figure 8shows composites from ERA-interim for the MODIS period averaged over the 40°W to 10°W longitude band for day −4 and day 0 relative to the maximum in AOD over the study region (i.e., 15°N-17°N, 22°W-20°W). Composites inFigure 8 were calculated based on AOD in the 5–15 day band and the vertical wind is exaggerated 1000 times to allow a comparison with the horizontal wind. Four days before the maxima in aerosol (Figure 8a) there is a middle troposphere cyclone centered around 20°N with associated negative geopotential anomalies. We hypothesize that the subsidence near 20°N shown in Figure 8a acts to prevent an aerosol build up. There is a minimum in potential temperature in the lower troposphere around 850 hPa and a maximum in the boundary layer specific humidity (Figure 8b) both displaced northward from the minimum in AOD anomalies. At the time of the aerosol maximum loading, the zonal wind reverses toward an anticyclonic circulation and a maximum in geopotential height (Figure 8c). The presence of a strong middle level easterly jet (centered around 14°N in Figure 8c) between the upstream trough and downstream ridge axis of an AEW is a common characteristic of the SAL [Karyampudi et al., 1999]. There is noticeable upward motion between 12°N and 22°N (i.e., the dust corridor) with a collocation of anomalously warm and dry air mass forming a temperature inversion around 875 hPa near the base of the SAL and over a moist boundary layer (Figure 8d). The inversion between the marine mixed layer and the base of the SAL (i.e., 850 hPa) is identified by a rapid increase in potential temperature with height in Figure 8d. The negative anomaly in specific humidity near the base of the SAL indicates a dry and stable atmosphere, inhibiting cloud formation consistent with Karyampudi and Carlson .
 A common feature of the SAL is that intense radiative heating accompanies dust transport over the desert surface as hot and dry air emerges from the west coast of Africa. The emergence of dusty, anomalously warm and dry air is viewed better in the latitudinal composites averaged between 10°N and 20°N shown in Figure 9. In the study region, four days prior to the maximum in AOD anomalies there is middle level zonal wind convergence favoring sinking motion in the lower levels of the troposphere near 18°W in Figure 9a. There is a region of moist and cool temperatures over a diminished aerosol layer (around 30°W-15°W, inFigure 9b). At the same time, over the Saharan desert (5°W-10°E) there is an upper level trough and upward motion accompanying the hot and dry air emerging from the African desert. The center of positive potential temperature anomalies decreasing with height over North Africa (around 0° inFigure 9b) favors instability from the surface to the mid-troposphere creating a favorable environment for the African dust to be transported upward from the Saharan desert. At day 0, the time of maximum aerosol loading, the meridional wind reverses and the zonal wind becomes easterly with hot and dry air that has moved over the region of maximum aerosol loading (Figures 9c and 9d). This warm and dry air is located over humid and cooler air favoring the maintenance of the African dust plume located in the lower to middle troposphere above the moist trade-wind consistent with earlier studies [Carlson and Prospero, 1972; Karyampudi and Carlson, 1988; Knippertz and Todd, 2010]. Similar characteristic motions and atmospheric states, as shown in Figures 8 and 9, are found in the 10–30 day band variability mode using both TOMS and MODIS satellite and reanalysis data sets (not shown). The only noticeable difference was that the temperature and humidity anomalies in the analysis of the longest variability mode possessed a greater longitudinal extension. This is coincident with a greater wave extension within the longer period of variability in similar way as the difference in wave extension in Figure 3.
Figure 10 shows the evolution of the vertical profiles of zonal and vertical wind, potential temperature and humidity anomalies through the composite period within the 5–15 day band variability associated with the AOD anomalies. The composite profiles are constructed by averaging each variable horizontally over the study region, locating the day zero as the day of maximum aerosol loading and computing averages of −6 to +6 days from the dates about this maximum. Figure 10a shows that about 4 days before day zero, low level easterly wind anomalies start to increase in height reaching a maximum near day −1 and forming the middle level easterly jet. The increase in vertical velocity (i.e., increasing negative anomalies) near the day 0 is also favoring conditions resulting in dust mobilization. In addition, there is an increment of positive anomalies of geopotential height from the surface toward the middle troposphere (not shown). Figure 10bshows a warming and drying in the lower to middle troposphere toward day zero corresponding to the dust-laden heated air emerging from West Africa and moving across the region as the AOD anomalies increases (i.e., black line inFigure 10b). Three days later, the situation reverses to positive vertical wind anomalies and a cool and humid atmospheric state as aerosol loading decreases toward its minimum.
5. CALIPSO Vertical Profiles
 The Carlson and Prospero model of the SAL states that after the passage of the heated dust-laden air from the Saharan desert toward the tropical Atlantic Ocean, the mass of air containing dust is elevated toward the 600–800 hPa layer. Using the CALIPSO lidar, profiles of aerosol extinction coefficient at 532 nm (k532, hereafter) were analyzed for the dates of maximum and minimum in aerosol loading within the 5–15 day band variability band as found in the MODIS-Aqua AOD data. It is important to mention again that the CALIPSO sensor retrieves aerosol profiles in a single trajectory during orbit thus limiting horizontal coverage. For this reason, those satellite trajectories in the day of maximum AOD crossing longitudes between 35°W and 15°W closer to the main study region were selected. In general, just one CALIPSO overpass is found per day to correspond to the location of the reference region. A total of 46 CALIPSO sensor overpasses were found during the dates corresponding to the maximum aerosol loading in the period 2006 to 2009. In 40 of those 46 overpasses, an aerosol plume is recognizable up to 400 hPa in elevation. In the majority of these cases a maximum concentration, represented by higher values of k532 is located around 700 hPa. In comparison, the overpasses for four days before the dates of the maximum in aerosol show, in general, smaller values of k532 than in the maximum cases, representing the lower concentration of aerosol dust in the atmosphere prior a high dust event.
Figure 11shows MODIS AOD anomalies for an example of a case where there is a maximum in aerosol loading occurring in June 23, 2009 together with a minimum in aerosol anomalies on June 19, 2009, four days before. The respective ERA-interim horizontal wind and geopotential height anomalies matching the daily AOD are also shown.Figures 11b and 11d show the k532 values from CALIPSO overpasses near the study region for the same dates as the MODIS AOD anomalies. Four days before the maximum in AOD (Figure 11a), there are negative anomalies of AOD and geopotential height that produces cyclonic wind over the region consistent with a low concentration of dust aerosol in the atmospheric column (Figure 11b). During the day of maximum aerosol loading (Figure 11c), the anomalies shift to positive AOD and geopotential height values associated now with an anticyclonic circulation. The CALIPSO profile (Figure 11d) shows a large increase in aerosol loading up to the 600 hPa level with a maximum centered near 700 hPa. This case illustrates where the aerosol plume concentration is located in the vertical during the day of maximum dust over the reference region. The shift between low to high aerosol concentration, based on the timing of maximum AOD within the 5–15 day band, displays the characteristic progression of dust as discussed earlier and characterize the elevation of the dust plume up to middle levels of the troposphere.
6. Summary and Conclusions
 The study provides detailed analyses of how climatological African Easterly Wave variability modifies the dust aerosol burden in the tropical Atlantic region and how the environment was modified by these intrusions. The analysis uses a diverse set of satellite measurements, which have different capabilities and covered different periods in order to minimize differences in aerosol retrieval techniques. In addition, the use of the ECMWF-ERA reanalyses and satellite observations of atmospheric and oceanic variables helped characterize the co-variability of aerosol dust and propose physical processes that determine the structure and temporal variability of the SAL.
 Using a Fourier filter technique, dates of maximum aerosol loading were selected to characterize the time scale of AEW variability over the study region. The technique was applied to both TOMS-AI and MODIS-AOD data sets. Two modes of westward aerosol propagation are apparent in both data sets, one with a period near 5–7 days and another with a period near 9–11 days. These two regimes represent the phase speeds of the two AEW forms crossing the region. Also, good correspondence is found between positive 700 hPa geopotential height anomalies and positive aerosol anomalies with the same periodicity. In a similar fashion, wind vector anomalies show that the direction of circulation changed depends on the timing of the maximum/minimum aerosol anomalies over the region. As positive anomalies of aerosol increase toward the day of maximum loading, there is a reversal from easterly to westerly in zonal wind direction anomalies, a change of sign from negative values (low pressure) to positive values (high pressure) in geopotential height anomalies and an increase toward a maximum in wind speed. While the pattern progression of the propagation of aerosol load has been evaluated before using reanalysis models and satellite observations [e.g.,Karyampudi et al., 1999], we present extensions to these studies specifically by evaluating the performance of the long series of satellite aerosol retrievals into the determination of the modes of aerosol variability in relation to the two forms of AEW, using reanalysis and satellite retrievals of atmospheric and oceanic variables to show how the atmosphere and the ocean are impacted by dust aerosol variability, and documenting how those impacts are represented in the Carlson and Prospero  model for dust transport over the Atlantic Ocean.
 The environmental impact produced by the variation of aerosol loading on the surrounding atmosphere and ocean shows a decrease toward minimum values of SST, and rainfall rate and to a maximum in OLR as aerosol loading increases over the study region. These changes are in concert with a reduction in precipitation, humidity and surface shortwave radiation as a warmer and drier dust plume is transported over the desert surface toward the Atlantic Ocean.
 The vertical state of the atmosphere during the evolution of aerosol loading is also investigated. Four days before the maxima in aerosol is reached in the study area, a middle troposphere cyclone characterizes the atmosphere near the latitude of minimum aerosol loading. We hypothesize that the subsidence, located in the same place as the minimum in aerosol loading, helps to prevent aerosol build up and suppresses dust generation. However, a more detailed dynamical analysis should be made in order to prove this fact. At the same time, the atmosphere is also found to be cool and humid. However, during the day of aerosol maximum, the middle level circulation reverses toward an anticyclonic flow with the development of a middle level easterly jet, characteristics of the favorable conditions for dust transport over the Atlantic Ocean. The previous cool and dry lower- to middle-level atmosphere is then replaced by an anomalously warm and dry air mass forming a temperature inversion around 875 hPa near the base of the SAL. This change in structure occurs with a continuous transition from cyclonic to anticyclonic circulation, a shift between negative to positive geopotential height anomalies, a shift from positive to negative vertical velocity with considerable warming and drying of the lower-to-middle layers of the atmosphere. Moreover, the evolution of these variables occurs with the same temporal scale as the AEW passing thought the region. The vertical extension of the aerosol extinction coefficient provided by the CALIPSO profiles show that during most of the cases where there was a maximum in AOD values over the study region, the aerosol plume reached a maximum concentration around 700 hPa. In contrast, smaller values of aerosol extinction coefficient are found for events during the fourth day prior to the maximum in AOD over the study region.
 The analysis provides an understanding of how the environment is affected by dust aerosol and how the Carlson and Prospero  model of aerosol variability match the temporal and spatial variation of AEWs crossing the region. The results presented here are useful in determining how a statistical scheme, in conjunction with an atmospheric forecast model, can be developed to forecast probabilistic outlooks of regional aerosol dust loading. In addition, the methodology used to investigate the covariability between aerosols and dynamic and thermodynamics variables can be extended to other regions of the globe (e.g., South Africa or the Middle East). However, it would be necessary to differentiate aerosol type in the analysis for regions affected by multiple sources of aerosols. The use of the capabilities of the MODIS's fine fraction in conjunction with the aerosol scene classification product of CALIPSO sensor can be useful in order to differentiate the distribution of aerosol type over diverse regions of the globe.
 This research has been supported from the National Science Foundation's Atmospheric Science division under NSF grant ATM-0531771 (P.J.W.) and by the National Oceanic and Atmospheric Administration Climate program Office under ACC grant GC08–027 (C.D.H.). We would like to thank the NASA Atmospheric Composition Data & Information Services Center, Atmosphere Archive and Distribution System and the Langley Research Center Atmospheric Sciences Data Center for providing the TOMS, MODIS and CALIPSO data, respectively. Also, we would like to thank NOAA NCDC and ESRL for providing the OLR data, RSS and NASA Earth Science Physical Oceanography Program for the SST data and the GEWEX project for the GPCP precipitation. ECMWF ERA-40 and ERA-interim data used in this study was obtained from the ECMWF data server.