JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D13, 8497, doi:10.

1029/2002JD002563, 2003

Micropulse lidar observations of tropospheric aerosols

over northeastern South Africa during the ARREX
and SAFARI 2000 dry season experiments
James R. Campbell,1 Ellsworth J. Welton,2 James D. Spinhirne,3 Qiang Ji,1 Si-Chee Tsay,3
Stuart J. Piketh,4 Marguerite Barenbrug,4 and Brent N. Holben3
Received 23 May 2002; revised 7 November 2002; accepted 30 January 2003; published 9 April 2003.

[1] During the Aerosol Recirculation and Rainfall Experiment (ARREX 1999) and

Southern African Regional Science Initiative (SAFARI 2000) dry season experiments, a
micropulse lidar (523 nm) instrument was operated at the Skukuza Airport in northeastern
South Africa. The lidar was colocated with a diverse array of passive radiometric
equipment. For SAFARI 2000, a daytime time series of layer mean aerosol optical
properties, including layer mean extinction-to-backscatter ratios and vertical extinction
cross-section profiles are derived from the synthesis of the lidar data and aerosol optical
depths from available AERONET Sun photometer data. Combined with derived spectral
Angstrom exponents, normalized broadband flux measurements, and calculated air mass
back-trajectories, the temporal evolution of the surface aerosol layer optical properties is
analyzed for climatological trends. For dense biomass smoke events the extinction-tobackscatter ratio is between 50 and 90 sr, and corresponding spectral Angstrom exponent
values are between 1.50 and 2.00. Observations of an advecting smoke event during
SAFARI 2000 are shown. The smoke was embedded within two distinct stratified
thermodynamic layers causing the particulate mass to advect over the instrument array in
an incoherent manner on the afternoon of 1 September 2000. Significant surface
broadband flux forcing of over
50 W/m2 was measured in this event. The evolution of
the vertical aerosol extinction profile is profiled using the lidar data. Finally, observations
of persistent elevated aerosol layers during ARREX 1999 are presented and discussed.
Back-trajectory analyses combined with lidar and Sun photometer measurements indicate
the likelihood for these aerosols being the result of long-range particulate transport from
INDEX TERMS: 0305 Atmospheric Composition and
the southern and central South America.
Structure: Aerosols and particles (0345, 4801); 0345 Atmospheric Composition and Structure: Pollution
urban and regional (0305); 0360 Atmospheric Composition and Structure: Transmission and scattering of
radiation; 0394 Atmospheric Composition and Structure: Instruments and techniques; KEYWORDS: lidar,
aerosols, biomass burning, Sun photometry, lidar ratio, SAFARI 2000
Citation: Campbell, J. R., E. J. Welton, J. D. Spinhirne, Q. Ji, S.-C. Tsay, S. J. Piketh, M. Barenbrug, and B. N. Holben, Micropulse
lidar observations of tropospheric aerosols over northeastern South Africa during the ARREX and SAFARI 2000 dry season
experiments, J. Geophys. Res., 108(D13), 8497, doi:10.1029/2002JD002563, 2003.

1. Introduction
[2] The climatic impact of biogenic, pyrogenic and anthropogenic surface emissions across the southern African subcontinent (described here as the region south of the 15
S
latitude) has been the focus of much recent field research. The
area is marked by significant industrial and domestic sulfate
and carbon release. A dense network of generator stations
1

Science Systems and Applications, Inc., Greenbelt, Maryland, USA.

Goddard Earth Sciences and Technology Center, University of
Maryland, Baltimore County, Maryland, USA.
3
NASA Goddard Space Flight Center, Greenbelt, Maryland, USA.
4
University of The Witwatersrand, Gauteng, South Africa.
2

Copyright 2003 by the American Geophysical Union.

0148-0227/03/2002JD002563$09.00

SAF

situated along the Highveld of northeastern South Africa,

responsible for most electrical power to the region, and the
predominance of biomass burning, both as a domestic fuel
source and in the form of savannah fires, are the most notable
contributors [Piketh et al., 1999]. Beginning with the 1992
South African Regional Science Initiative (SAFARI) field
campaign, and continuing through numerous projects,
including the 1999 Aerosol Recirculation and Rainfall
Experiment (ARREX) [Terblanche et al., 2000] and SAFARI
2000, many diverse data sets have been collected in the
region through coordinated surface, airborne and satellite
remote sensing and in situ sampling. The goal of these efforts
is the accurate characterization of regional aerosol evolution,
transport and eventual deposition, and an improved understanding of the regional climatic and biological repercussions
of its presence in the atmosphere [Swap et al., 2002].

33 - 1

SAF

33 - 2

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

[3] This article discusses observations made through

colocated active and passive remote sensing at the Skukuza
Airport in the Krueger National Park of northeastern South
Africa during the ARREX and SAFARI 2000 dry season
components. A micropulse lidar (MPL), described by Spinhirne [1993], is the primary instrument of focus. Lowpowered, eye-safe, and autonomously operated, the ruggedized MPL instrument is quite amenable to remote field
operating conditions [Spinhirne et al., 1995]. The MPL is a
single-channel (523 nm), elastic backscatter lidar. Despite
being low-powered, the MPL has been shown to be sensitive
to nearly all forms of tropospheric cloud and aerosol
[Spinhirne, 1993]. MPL instruments have been used in
numerous recent aerosol-related field experiments including
the Indian Ocean Experiment (INDOEX) [Welton et al.,
2002a], the Aerosol Characterization Experiment-2 (ACE-2)
[Welton et al., 2000; Powell et al., 2000], ACE-Asia [Welton
et al., 2001a], and the Chesapeake Lighthouse and Aerosol
Measurements for Satellites experiment (CLAMS) [Smith et
al., 2001].
[4] A single-channel lidar is somewhat limited in standalone capability, particularly in light of more robust lidar
systems with depolarization and multiple channel capacities
used during SAFARI 2000 [McGill et al., 2003], and
previous African field campaigns [e.g., Fuelberg et al.,
1996; Browell et al., 1996]. The advantage of the MPL
instrument is in its practicality (ease in remote deployment),
and ability to make continuous full-time measurements. In
addition to the lidar a diverse suite of passive radiometric
instrumentation was operated simultaneously at the airport
site. A CIMEL Sun photometer also operated less than two
kilometers away in the Skukuza base camp. Integration of
these data sets, as well as corollary information derived
from upper air back-trajectories, reveals much information
on the optical characteristics of the incident aerosols, and
well as qualitative information on the nature of the aerosols
being observed.
[5] In this presentation we examine the temporal evolution of local dry season aerosol optical parameters. Particular attention is given to the layer mean extinction-tobackscatter ratio (S), derived by combining lidar particulate
backscatter and Sun photometer aerosol optical depth measurements. MPL instruments serve an ancillary purpose in
collecting ground-validation data sets for the National
Aeronautics and Space Administration (NASA) Earth
Observing System (EOS) Geoscience Laser Altimeter System (GLAS) project. Proposed GLAS data processing
algorithms specify a priori knowledge of this parameter
[Palm et al., 2001]. However, while numerous investigations have increased our understanding of this parameter
with global consequence [e.g., Takamura et al., 1994;
Anderson et al., 2000; Peppler et al., 2000; Welton et al.,
2000; Ansmann et al., 2000; Ferrare et al., 2001; Voss et
al., 2001; Welton et al., 2001a], and modeling studies have
supplemented their work [e.g., Ackermann, 1998], practical
characterization of the regional variation of S is still
incomplete. The findings presented here, combined with
similar measurements made by the cloud physics lidar
[McGill et al., 2003] aboard the NASA ER-2 aircraft [King
et al., 2003] will be used in this cause. Variation in S and
spectral Angstrom exponent calculations, combined with
analysis of upper air back-trajectories allow us to investigate

distinct smoke and haze conditions observed during the

SAFARI 2000 experiment relative to otherwise ordinary
conditions.
[ 6 ] Two case studies are also examined. The first
describes measurements of an evolving smoke plume from
nearby biomass fires observed advecting over the instrument array. This case is unique as the smoke was observed
to have evolved within separate boundaries of two vertically
and thermodynamically distinct layers. Combined with surface flux measurements, this case is fully characterized with
mind kept to the potential for validation of modeling studies
on downwind smoke plumes [e.g., Trentmann et al., 2002]
and regional smoke transport. The second study describes
the unexpected observation of persistent elevated aerosol
layers (throughout the troposphere) during ARREX. The
available evidence indicates a strong likelihood that they be
the result of long-range transport of particles from a South
American continental source via strong westerly upper level
advection. The exact nature of the aerosols is unknown,
though potential sources are discussed. The prolonged
existence of significant elevated aerosol layers would prove
bothersome to passive satellite algorithm retrievals that
constrain observed aerosol characteristics to the surface
layer. Additionally, long-range aerosol transport modeling
studies require regional characterization of the vertical
aerosol structure to interpret model output. With no previous evidence for intercontinental aerosol transport
between South America and southern Africa, these observations are potentially very significant. Note that crossvalidation studies involving Skukuza MPL data sets and
measurements from airborne instrumentation during
SAFARI 2000 can be found in papers by Schmid et al.
[2003] and McGill et al. [2003].

2. Instrument Setup and Data Processing

[7] A common research site was used during both
SAFARI 2000 and ARREX. The site encompassed the
southwestern corner of the Skukuza Airport (24
590S,
31
350E, 315 m ASL) in the Krueger National Park. A
passive radiometric instrument array was constructed in a
field across from two aviation hangers. This suite of instruments included a multifilter rotating shadowband radiometer, scanning microwave radiometer, and multiple
devices measuring broadband solar (diffuse and direct)
and infrared radiation, as well as specific target wavelengths. The MPL was operated at the hangers, varying in
placement slightly by year, approximately 100 m from the
passive array. A permanent Aerosol Robotic Network
(AERONET) CIMEL Sun photometer site [Holben et al.,
1998] was in operation less than 2 km from the airport in the
Skukuza base camp.
[8] Following Campbell et al. [2002], the MPL instrument used during ARREX featured a V2.0 optical configuration. The instrument was housed in dual insulated
containers, allowing for environmentally sheltered outdoor
operation, placed between the hangers to limit direct solar
heating. The containers were mated top (optical transceiver)
and bottom (power supplies, computer and scalar unit), with
collard ducts in between the two to allow for shared wiring
and circulation. A broadband-coated, optically flat window
was mounted along the top face of the upper box nearest to

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

the transceiver. Despite much effort, the design of these

containers failed to properly regulate and stabilize the internal thermal operating environment. Welton et al. [2002a]
discuss MPL data sets where the thermal stability of the
instrument has been compromised, and the difficulties in
processing them. Quantitative analysis of the ARREX data
sets was severely limited as a result. The MPL used during
SAFARI 2000 featured a V2.1 optical design. The instrument
was operated inside the northernmost hanger. Though this
shelter was not ideally sealed or insulated, many of the
thermally induced difficulties experienced the previous season were overcome. A window mount was constructed on top
of the hanger to accommodate an optical-quality transmitting
window. A ceiling-mounted fan placed in the ceiling hatch
limited window condensation.
[9] During both experiments, very little instrument downtime was experienced (>90% collection efficiencies). For
ARREX, data were recorded at a 30 m vertical resolution
setting. A lower setting (75 m) was used the following year
for SAFARI 2000. Both data sets were recorded in sixtysecond averaged profiles, and stored to local disk (roughly
150,000 pulse integrations corresponding to the 2500 Hz
laser pulse rate). Instrument calibrations were accomplished
using techniques outlined by Campbell et al. [2002]. The
airport complex is situated on a significant butte. Unobstructed horizontal sightlines were available in most directions. Calibration of the instrument near-range overlap
function (done using a horizontally homogenous target
scattering layer) was relatively easy, and done frequently
to insure data quality.
[10] Normalization of raw MPL data requires accounting
for ambient background light, instrument-specific correction
terms, and solving for the system calibration coefficient
through the lidar equation. Campbell et al. [2002] discuss
the algorithm and assorted techniques used to develop the
raw data to the point of uncalibrated backscatter (termed
normalized relative backscatter, or NRB). Welton and
Campbell [2002] outline the methods for calculating uncertainties for this algorithm and each of the instrument
correction terms. Furthermore, Welton et al. [2002] describe
an algorithm for solving the absolute calibration coefficient
(C), as well as indirectly derived optical parameters (i.e.,
backscatter and extinction cross-section profiles, optical
depth profiles and layer mean extinction-to-backscatter
ratios), including uncertainties for all terms. These algorithms and techniques are relied upon for this work.
[11] In our analyses, we use 30-min averages of NRB
profiles centered on an independent measurement of the
total column aerosol optical depth. Solving for C (briefly
described below) relies on an estimate of column particulate
transmission rates. For both experiments CIMEL Sun photometer data, available through the AERONET online database at http://aeronet.gsfc.nasa.gov, was used as the
reference source. Measurements from this instrument were
used with an uncertainty taken to be 0.01 corresponding to
their level 2.0 product [Holben et al., 1998]. Campbell et al.
[2000] describe the variability of C over a diurnal cycle
using an MPL. To insure accurate data processing with the
lowest quantifiable uncertainties, we restrict our analyses
for this presentation to daytime periods where CIMEL
measurements were available. Ambient background light
decreases signal-to-noise rates significantly during the day-

SAF

33 - 3

time, therefore requiring significant profile averaging for

analysis.
[12] The algorithm described by Welton et al. [2000]
employs a user-specified range above the surface aerosol
layer, and free of particulate backscatter, from which to
solve for C. After this calculation, the algorithm then
searches for the top height of the aerosol layer in order to
properly run a backward Fernald numerical inversion [Fernald, 1984] to solve the two unknowns in the lidar equation
(particulate backscatter coefficient, and particulate transmission). The result is a layer mean value of S, and profiles
of extinction and backscatter coefficients. In order to
automate this algorithm, we document in Appendix A the
basic approach to first solving for the top of the aerosol
layer, so as to choose a range for solving for C as near to the
instrument as possible. This has a significant effect on
decreasing the uncertainty in C that, in turn, improves the
uncertainty in those subsequently derived parameters
dependent upon it.

3. SAFARI 2000 Optical Characteristics

[13] In this section we examine optical properties derived
from combined lidar and Sun photometer data collected
during SAFARI 2000. We supplement these findings with
air mass back-trajectories calculated using the National
Oceanic and Atmospheric Administration (NOAA) Hybrid
Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) Model to infer the nature of the aerosols investigated
relative to local source regions and predominant flow.
HYSPLIT model calculations can be interactively accessed
via the NOAA Air Resources Laboratory Web site at http://
www.arl.noaa.gov/ready/hysplit4.html. During SAFARI
2000, the MPL was operated from 18 August through 22
September. Low clouds spoiled much of the last week of
this period, so we focus here on the period through 13
September inclusive. Corresponding to each day, three-day
back-trajectories were calculated from 0000 UTC at five
atmospheric levels (surface, 850, 700, 600 and 500 hPa). It
was found that, in general, the surface to 850 hPa layers had
similar transport patterns, while in the upper atmosphere
circulation also correlated well, though different from that at
the surface. We differentiate between the two layers in our
discussion.
[14] Figure 1 is a collection of results encompassing the
entire SAFARI 2000 period including: daily classifications
of derived back-trajectories for the surface and upper layers
(Figure 1a), algorithm output (S with error bars, and top of
the surface aerosol layer, Figures 1b and 1e, respectively)
and the spectral Angstrom exponent (AE, Figure 1c) and
523 nm aerosol optical depth (hereto forth referred to as
simply AOD, while implicitly understood to be at the MPL
wavelength, Figure 1d) derived from available CIMEL data.
For the back-trajectories, the surface and upper level classifications were broken out into five and four regimes
respectively. Surface cases were divided into easterly (1),
recirculated (referring cases where the airmass originated
over the continent, was blown off over nearby waters, and
recirculated back to Skukuza, (2), westerly (3), northerly (4)
and southerly (5). Upper level cases were divided into westnorthwesterly (1), jet stream westerly (2), recirculated northerly (3), and recirculated easterly (4). AOD values were

SAF

33 - 4

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 1. A summary of observations during the SAFARI 2000 experiment, 18 August to 13 September
(fractional days 231 to 257 inclusive): back-trajectory classifications for lower (surface to 850 hPa) and
upper levels (700 to 500 hPa), as described in the text (a), derived values of S for surface aerosol layer
with uncertainty bars (b), AERONET CIMEL 523 nm aerosol optical depths (c), and Angstrom
exponents (d), and the derived aerosol layer top height used to initiate algorithm inversion techniques (e).

derived using a second-order polynomial fit to the natural

logarithm of AOD versus wavelength where at least four of
the seven available CIMEL channels (between 340 and
1020 nm) reported data. AE values were similarly derived,
though using a traditional power law curve-fit [e.g., Angstrom, 1964]. Owing to some fluctuations in MPL instrument thermal/optical stability, only those observations
where the instrument temperature was less than 27.8 C
are shown (chosen on the basis of prior experience with the
subject instrument).
[15] As noted, derived values of S represent layer mean
averages, as limited by the single-channel lidar instrument
and Fernald inversion methodology. Algorithm noise,
mostly due to cloud contamination of both the lidar profiles
and ingested CIMEL AOD spectral profiles, is also expected
to be present in our calculations. Error bars calculated for
each case represent predominantly those quantifiable uncer-

tainties arising through photon-counting statistics, both for

the measured signal and for the instrument correction terms
as referenced above. Therefore there is a floating uncertainty
in our measurements that represents both the vertical inhomogeneity of S in the vertical column and algorithm noise
for which we cannot account for. However, separate papers
by McGill et al. [2003] and Schmid et al. [2003] show the
MPL calculations (including profiles of extinction coefficient as well as S) to be in good overall agreement with other
instruments running concurrently near or at Skukuza at
various points during the experiment.
[16] Previous researchers have shown that the composition of the aerosol common to the region is highly complex,
with many factors influencing it at any given time and place
[e.g., Piketh et al., 1999]. With regards to Figure 1, this
seems acutely apparent. Stable synoptic conditions dominated the latter part of August, into early September allow-

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 2. For the period 15 August through 15 September,

the average aerosol optical depth versus average Angstrom
exponent from available CIMEL Sun photometer data
during 1998 2001 for the Skukuza Airport site.

ing for many data points. To establish some perspective,

Figure 2 shows averages of AOD and spectral AE (with
standard deviations) for the period 15 August through 15
September inclusive during the years 1998 to 2001, as
derived from the AERONET database for Skukuza. With
respect to AE, and to a lesser degree AOD, the 2000 season
was somewhat unusual. Increases in both parameters from
the previous two years indicates a downward shift in
particle size distributions in local aerosols, and is consistent
with an increase in overall particulate loading with respect
to the prevalence of local biomass burning [Reid et al.,
1998; Remer et al., 1996].
[17] From 18 to 22 August (fractional days 231 235)
surface flow was consistently from the east, while upper
level flow varied cyclically. Easterly surface flow represents
onshore flow from the western Indian Ocean, leading to
increased near-ground relative humidities. Aerosol humidification effects should be greatest in this pattern as convective mixing of the boundary layer enhances growth of
particles each afternoon near the top of the lowest thermal
inversion level [e.g., Berg et al., 1998]. Since the observations reported here coincide mostly with synoptically stable,
clear-sky conditions, thermal expansion and convective
mixing of the boundary layer was likely simultaneously
occurring. Evidence of resulting shift in particle size distribution can be seen in the diurnal increase of AE during
each afternoon of the experiment (Figure 1d). The prevalence of regional biomass burning as a fuel source for
cooking may also have had some influence, as local smoke
concentrations regularly increased through the day. As AOD
can be seen in Figure 1c to have decreased during most
afternoons though, this effect was likely much smaller. The
aerosol layer top height peaked near 5.00 km on 21 August
(day 234), with corresponding peaks in AOD and AE. AE
values approached and briefly exceeded 2.00, and were
among the highest observed during the experiment. These
measurements coincided with an upper level trajectory shift
to west northwesterly. Such a pattern would increase the
influence of continental transport from upwind biomass

SAF

33 - 5

burning sources in Namibia, Botswana and Zambia. General

trends in S calculations are in agreement with these observations. From 18 August, where upper level flow was
easterly, then northerly, values were between 40 and 60 sr.
Beginning on 20 August, when upper level trajectories
became more westerly, much higher values, near and above
80 sr, were derived.
[18] Upper level trajectories from 23 to 28 August were
westerly. This pattern decreased the influence of continental
transport, increasing the impact of urban and industrial
pollution sources from along the Highveld and in central
South Africa. AOD values peaked on 25 August, though
AE values did not reach the maximum values derived in
previous west northwesterly upper level conditions (peaking
near 1.90 on 24 August). Both parameters decreased notably through 28 August. S during this six-day period
remained consistent with the previous smoke period, with
values occurring between 40 and 80 sr. Relative variance in
AE measurements here indicate some change in the aerosol
composition, as hinted to by the change in upper level
trajectories. However, relative persistence in S signifies that
the overall absorptive character of the aerosols had not
changed significantly.
[19] The three days, 1 3 September (days 245 247),
were heavily influenced by smoke (to be examined further
below). Beginning on 31 August, upper level trajectories
shifted back to west northwesterly, where they would stay
through the experiment (but for a slight deviation on 3
September). Nearer the surface, trajectories turned to westerly. The surface aerosol layer swelled to near 5.00 km, AE
values peaked again near 2.0, AOD reached an experiment
maximum over 0.70, and S remained between 40 and 80 sr.
This period ended abruptly as a synoptic disturbance moved
through the area inducing low and midlevel cloudiness for a
number of days. Lower level trajectories varied between
northerly (post-frontal) and westerly during this time. By 9
September (day 253) AOD values dropped to 0.20, and the
aerosol layer top height was near 3.00 km. More notably,
AE values plunged to near 1.00, and S values to roughly 40
sr. A more maritime-based aerosol composition was likely
in place, as low-level trajectories were in a recirculating
pattern.
[20] Figures 3a and 3b display the relative frequency of
the individual air mass back-trajectories during the experiment for the upper and lower levels respectively. Trajectories at lower levels showed easterly and recirculated origin
over 60% of the time. At upper levels, westerly and west
northwesterly trajectories occurred nearly 90% of the time.
In Figure 4a is a comparison of S to corresponding AE
value for each of the data points shown in Figure 1. A
similar comparison of AOD versus AE values is shown in
Figure 4b. As discussed above, changes in AE showed
appreciable correlation with back-trajectory shifts, and is
therefore considered here to be a much more reliable
indicator of changes to local aerosol composition. AE is
directly related to the slope of a Junge-type size distribution.
S is dependent upon the aerosol phase function and the
single scatter albedo, and therefore is sensitive not only to
changes in particle size, but also shape and refractive index.
And, as aerosol species also have varying hydroscopic
properties, particle size, shape, and refractive index are all
dependent upon relative humidity to some degree. Given the

SAF

33 - 6

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 3. Corresponding to Figure 1, a breakdown of

percentage frequencies of air mass back-trajectories for
lower (a) and upper levels (b). From the surface to 850 hPa,
trajectory classifications were broken into five categories:
northerly, westerly, easterly, southerly and recirculated.
From 700 to 500 hPa, trajectory classifications were broken
into four categories: west-northwesterly, jet stream westerly,
recirculated northerly, and recirculated easterly.
unique, yet chaotic regional mix of industrial and biomass
pollutants, and the proximity of a major water source,
overall variance in S should be quite high, and this is
apparent in Figure 4a.
[21] Use of back-trajectory analyses allows the opportunity to better characterize the nature of the incident aerosols.
Figures 5a 5d expand upon Figure 4a by breaking down
the available data points into the four most commonly
occurring lower/upper level trajectory scenarios: easterly/
westerly (22%, Figure 5a), recirculated/west northwesterly

(22%, Figure 5b), westerly/west northwesterly (15%, Figure

5c) and easterly/west northwesterly (11%, Figure 5d). The
first of these scenarios (Figure 5a), was noted to have
occurred between 23 and 28 August, and responsible for a
greater influence of urban and industrial haze at upper
levels. A linear regression calculation showed no apparent
correlation, though AE values were otherwise between 1.30
and 1.90, and S varied between 50 and 90 sr. The second
scenario (Figure 5b), occurring near the end of the experiment, produced the lowest number of data points in this
breakdown. With recirculation prevalent at lower levels, a
greater influence of maritime conditions on the aerosols
yielded AE values between 0.80 and 1.60, while S varied
between 30 and 70 sr. The most successful linear correlation
coefficient was found in this scenario (R = 0.79). The third
scenario (Figure 5c) coincided with the strong smoke event
at the beginning of September. AE values were between
1.50 and 2.00, and S was between 50 and 90 sr. The second
highest correlation coefficient was calculated in this scenario (R = 0.31). The final scenario (Figure 5b), responsible
for the first significant biomass event at the beginning of the
experiment yielded AE values between 1.70 and 2.00, and S
values between 50 and 80 sr. This scenario correlates well
with the biomass conditions noted in Figure 5b. As a whole,
however, these findings suggest a low correlation between S
and AE even when broken down by air mass origin. This
suggests significant variability in the aerosol refractive
index (S dependent upon through the single-scatter albedo
term) as a function of particle size distribution (both
dependent upon, S through the scattering phase function).
If the former term were held constant, we would expect to
see some sort of relationship otherwise. This again speaks to
the overall complexity of the aerosols, despite air mass
origin, at any given point in time.
[22] We briefly consider a less intensive analysis of the S
results derived from this experiment. For aerosols, the
absorption as quantified by the wavelength-dependent single scatter albedo is the major determining factor for a given
AOD of the aerosol induced heating rate and surface shortwave flux forcing. Both S and the ratio of broadband
shortwave diffuse to direct flux depend on aerosol absorption. For S, it is known that increasing the imaginary part of
the refractive index, and thus absorption, leads to an
increase in this parameter. This shows up most dramatically
in Mie calculations for spherical particles. For a given size
distribution and real part of the refractive index, S can
change by an order of magnitude. In past studies [e.g.,
Spinhirne et al., 1980] attempts have been made to estimate
the imaginary part of the refractive index of aerosol from S
measurements. Although there are large uncertainties introduced by particle nonsphericity, composition and unknown
size distribution, results do fit a realistic model. Sky brightness is expected to decrease as aerosol absorption increases.
A dark aerosol of lower single scattering albedo should
lower the diffuse shortwave flux. Although there is also a
scattering phase function effect, the primary impact of
increasing the imaginary part of the refractive index in
calculations is to lower the diffuse to direct ratio due to
absorption. Surface albedo is also a major influence on the
diffuse to direct ratio and must be known for a quantitative
assessment. A higher surface albedo would magnify the
aerosol absorption influence.

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

SAF

33 - 7

Figure 4. Corresponding to Figure 1, a comparison of Angstrom exponent values calculated from

available AERONET CIMEL Sun photometer measurements vs. derived S values (a), and AERONET
CIMEL 523 nm aerosol optical depths (b). Linear regression functions are shown, including relevant
statistics such as the correlation coefficient.
[23] In Figure 6 is shown a comparison of normalized
diffuse to direct shortwave flux ratio measurements versus
the corresponding path normalized CIMEL AOD (multiplied by airmass) for three distinct grouping of S (less than
55, between 55 and 75 and greater than 75 sr). The expected
dependence of a lower S with a higher diffuse to direct
shortwave radiance ratio is clearly evident on the basis of
linear regression fits to the available data points. High
correlation coefficients were found with each fit. Linear
functions were chosen on the basis of corresponding model
calculations (not shown) using single-scatter albedo as the
independent variable within these same axis parameters. For
values of single-scatter albedo commonly associated with
biomass burning aerosols (approx. 0.80) [e.g., Reid et al.,
1998], a roughly linear relationship was observed in the
simulation results. Increasing nonlinearity was found with
increasing divergence from these values. The data in Figure
6 are in agreement with the model, with variability in the
derived slope values a reflection of the true nonlinearity of
the relationship. As we have already concluded there to be
significant complexity with respect to S, this agreement is
likely tenuous (though still encouraging) given the effect of
the scattering phase function over nearly thirty days of field
observations.

4. Case Studies
[24] Two notable events were identified from the SAFARI
2000 and ARREX data sets, and are discussed in this
section as case studies. First, we examine the evolution of
a fresh smoke plume over the Skukuza site as observed by
the surface instrument array. Second, observation of persistent elevated aerosols during ARREX led to an investigation

into their potential nature and source. The lack of corroborating observations the following year makes these observations quite noteworthy.
4.1. Advecting Smoke Layer: 1 September 2000
[25] Our first case study deals with observations made on
1 September 2000 (fractional day 245), when smoke aerosols from recently begun fires upwind were observed
advecting over the airport site. (Note that times are reported
in this section in fractional day UTC. Local time at Skukuza
was +0200 hours relative to UTC.) Figure 7 is a threechannel (0.66 mm, 0.55 mm, and 0.47 mm) early morning
composite image of level 1B Moderate Resolution Infrared
Spectrometer (MODIS) data taken aboard the NASA EOS
Terra satellite (approximately 245.37). The Skukuza Airport
(24.97
S, 31.58
E) is denoted in the image by a white dot.
The satellite pass was centered well to the west of Skukuza
(note the edge of the data swath on the right side of the
image), so the resolution in the area of interest has been
compromised to some degree. However, numerous fires and
their resulting plumes existed in and around the vicinity of
Skukuza. The highest concentration of fires was confined to
a 3
by 3
grid directly north of the airport site. No fires
were evident the previous morning in the available MODIS
imagery.
[26] In Figure 8, MPL NRB data are displayed for the full
day. The surface aerosol layer was confined to the first 4.00
km AGL. Two daytime balloonsonde profiles are shown in
Figures 9a 9b. The top of the layer was capped by a slight
temperature inversion near 580 hPa in the early morning
(Figure 9a, 245.21) and near 560 hPa during the midafternoon (Figure 9b, 245.50). Measurements from a broadband
solar radiometer at the airport are shown versus correspond-

SAF

33 - 8

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 5. Corresponding to Figures 3 and 4, comparisons of Angstrom exponent values calculated from
available AERONET CIMEL Sun photometer measurements vs. derived S values for the four most
common lower/upper level back-trajectory classifications: easterly/westerly (a), recirculated/westnorthwesterly (b), westerly/west-northwesterly (c), and easterly/west-northwesterly. Linear regression
functions are shown for each figure, including relevant statistics such as the correlation coefficient.
ing airmass (defined as the inverse cosine of the solar
zenith angle) in Figure 10. These data are broken into
morning and afternoon observations, relative to the solar
angle rising and lowering. Figure 11 is an expansion of
Figure 1 for the subject day, with MPL algorithm output
supplemented by corresponding CIMEL measurements.
Finally, in Figures 12a 12c are three profiles of the
extinction cross section as derived from the MPL measurements for the early morning (245.23), midafternoon
(245.50), and late afternoon (245.60) respectively.
[27] What was unique about this event was that the
advancing smoke plume, as witnessed at Skukuza, developed within two thermally separate layers. Aside from the
upper inversion cap, a strong near-ground inversion was
present in both radiosonde profiles (Figure 9). During the
morning it was near 1.00 km. By afternoon, thermal
expansion of the layer caused this boundary to lift to near

1.50 km. A clear delineation of the layers throughout the

day is evident in both the NRB image (Figure 8), and the
derived extinction profiles (Figures 12a 12c).
[28] From the NRB data, in the early hours of the day
aerosols in the upper layer were well mixed. Some wisp-like
structure (a sign of the oncoming event) was detected near
245.10. In the lower layer, a denser aerosol layer was
detected. Though the first available sounding profile
(Figure 9a) came a few hours later, it is likely that the
extremely stable thermal structure measured in the lower
layer at that time prevailed earlier as well. With the sunrise
(near 245.20) an evolution of the layer to the more convectively unstable one seen in the afternoon (Figure 9b)
commenced. This can be seen in the NRB data as the aerosol
convectively mixed to near the inversion cap height with
time (similar to as suggested for a humidified aerosol
scenario in the previous section). Gradually this layer dis-

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 6. A comparison of the measured diffuse to direct

broadband solar flux ratio versus path-normalized AERONET CIMEL 523 nm aerosol optical depth for three groups
of S; less than 55 (plusses), between 55 and 75 (asterisks)
and greater than 75 (open circles). Linear regression
functions are shown, including relevant statistics such as
the correlation coefficient.

sipated by 245.40. A snapshot of the evolving structure is

seen in the first extinction profile (Figure 12a) where a peak
value over 0.30 km
1 is seen in the lower layer, while the
upper layer contained values under 0.10 km
1. AE values
beginning after sunrise (Figure 11) were near 1.50.
[29] By midafternoon (245.50), AE values increased, as
did the AOD. The MODIS image, though earlier, showed
that the corridor within the Lowveld was completely covered with varying degrees of smoke. However, dense
concentrations were not yet at Skukuza. The NRB data at
this time indicated that the smoke was almost exclusively
confined to the upper layer. A peak extinction coefficient
from the corresponding profile near 0.20 km
1 was detected
at 2.50 km (Figure 12b). There was no indication yet of any
significant increase in particulate matter (i.e., smoke) near
the surface. This is presumably due to higher wind speeds
above the near-ground inversion causing increased particulate advection in the upper layer (wind data was not
available from the balloonsonde data).
[30] By late afternoon (245.60), smoke embedded in the
lower layer reached the airport., and the AOD increased
rapidly to near 0.70. The corresponding extinction profile
(Figure 12c) shows the remarkable change in the structure
of this layer. S, highly variable earlier in the day, converged
at this time to near 60 sr. This observation is seemingly
counter-intuitive given the high absorptive character of
fresh smoke particles [e.g., Reid et al., 1998; Martins et
al., 1998]. It signals a counter-balancing effect from the
scattering phase-function term in S owing to an appreciable
decrease in particle size distribution. (AE increased accordingly at this time to near 1.70). This event marked one of the
densest smoke events with respect to AOD observed during
the experiment.
[31] As discussed above, lower level back-trajectories
were westerly and those in the upper level were west

SAF

33 - 9

northwesterly. Stabilization in S at this time suggests that

the vertical structure and composition of the aerosols was
temporarily persistent. As such, it makes sense that the
location of the upwind burning was in somewhat close
proximity, as the MODIS imagery validates. Turbulent
mixing over the lifetime of the plume had yet to have much
effect in entraining dissimilar air, thereby inducing pockets
of inhomogeneity, which would cause this parameter to
fluctuate. The change in broadband forcing at the ground
caused by the smoke was approximately
50 W/m2, when
compared to corresponding airmass values from earlier in
the day (Figure 10). As smoke was observed in the upper
layer during this earlier time, we estimate that the overall
forcing of the smoke relative to the otherwise common
background aerosol was actually somewhat higher than this
amount.
[32] Tyson et al. [1995, 1996] and Garstang et al. [1996]
discuss climatological air transport patterns over southern
Africa in the lower troposphere, and corresponding aerosol
recirculation in conjunction with the convective boundary
layer. They also identify climatological trends in observed
thermal inversion boundaries. Of note, Tyson et al. [1996]
use downward looking lidar measurements made during
SAFARI-92 to correlate aerosol concentrations with significant vertical thermodynamic structures. They found comparable discontinuities in aerosol concentration with each
incident layer, though they did not document a case such as
this.
[33] Transport-modeling efforts involving smoke in such
a unique thermal environment face a daunting task. The
available data limits us from deriving any potential optical
disparities between the two layers. However, we can safely
assume that as the case further unfolded with time, a
continued delineation of the layers led to two uniquely
modified aerosol types. What is clear is that a significant
amount of smoke had reached the upper thermal layer, and
that it was advecting quickly upstream from that trapped in
the lower layer. Models will need accurate characterization
of the efficiency at which a plume may punch through an
initial inversion (i.e., buoyancy), not to mention proper
initialization of the ambient thermal environment itself.
Finally, the evolution of the aerosol composition would
need to be accurately depicted in response to entrainment
and modification.
4.2. Elevated Aerosols During ARREX
[34] Though the MPL data from ARREX were not
suitable for thorough algorithm processing, our second case
study involves a portion of these data that contrasted with
results derived from the 2000 season. Shown in Figure 13a
are MPL NRB data from 7 September 1999. Cloud and
heavy aerosol loading was present during the first twothirds of the day. By early evening the upper troposphere
could be identified and much aerosol structure was evident.
This day represents one of the more significant examples
amidst a near two-week period (between 24 August and 9
September) where detectable aerosol was measured well
above the common near-surface stable layers. No such cases
of upper tropospheric aerosols could be found in the
SAFARI 2000 data sets.
[35] It is not readily obvious what the aerosols consist of.
Interpretation of the available lidar and passive data alone is

SAF

33 - 10

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 7. MODIS level 1B geolocated RGB composite imagery from bands 1, 4 and 3 (.66 mm, 0.55
mm, and 0.47 mm) for southeastern Africa from 0900 to 0905 UTC, 1 September 2000. Longitude/latitude
designations are detailed along the border of the image, and the location of the Skukuza Airport site
(24.97
S, 31.58
E) is labeled with a white star. The southeastern continental boundary is denoted on the
right side of the image by a white border (though there are no data). An escarpment, separating the
Highveld (west) and Lowveld (east) regions, can be identified as the darker winding terrain border
beginning from the bottom of the figure near 28.00
E. Smoke plumes, and diffuse smoke aerosols are
evident throughout most of the northern portion of the Lowveld. Hydrometeor clouds are evident in the
southeastern sector, in close proximity to the Indian Ocean.

ambiguous. In situ aircraft measurements were made during

the period, but the data were not conclusive. Ozone concentrations were found to be above normal background
levels, which are indicative of biomass burn remnants.
However, these flights were understandably coordinated to
sample near-surface aerosols, and did not extensively penetrate the elevated layers. The possibility of downstream
advection from fires in southern Zambia and neighboring
areas exists [e.g., Swap et al., 1996], though the lack of

corroborating observations the following year raises significant uncertainty in this scenario.
[36] Though no sounding data were available from Skukuza for early September, the apparent structure in Figure
13a suggests particle stratification along thermal boundaries. This is consistent with a prolonged presence, whereby
newly introduced particulate would likely show some
degree of turbulent (mixing) structure. Seven-day air mass
back-trajectories calculated for two upper tropospheric

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

SAF

33 - 11

Figure 8. MPL normalized relative backscatter (counts*km2/uj*us) data from 1 September 2000 at the
Skukuza Airport site.

levels (6 km and 9 km) beginning from 0000 UTC 7

September at Skukuza are shown as Figure 13b. The
trajectories show no air mass interaction with the southern
African continent before reaching Skukuza, effectively
eliminating the possibility of regional biomass influence.
The origins of the air mass can be traced back from between
days 3 and 5 to southern South America, where it appears

these aerosols must have originated. Previous researchers

have described aerosol transport events of South American
biomass burning matter into the southern Atlantic basin
[Anderson et al., 1996; Browell et al., 1996; Kent et al.,
1998]. Further advection to the African continent is plausible. Examination of AERONET CIMEL data from sites in
central South America showed high AOD and AE values

Figure 9. Balloonsonde profiles of temperature versus height above ground (km) from 1 September
2000 at approximately 0500 UTC (fractional day 245.2076) (a), and 1200 UTC (245.4993) (b) launched
from the Skukuza Airport site.

SAF

33 - 12

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 10. Broadband solar irradiance measurements

versus airmass for 1 September 2000. Circles/solid
represents measurements as Sun was rising, and square/
dotted represents measurements as Sun was setting.
Radiative forcing of afternoon smoke layer can be
interpreted by comparing corresponding airmass measurements (i.e., early morning and late afternoon). The forcing
of this event is estimated to be greater than
50 W/m2.

from sites in western Brazil indicating an early start to the

regional burning season. However, investigation of the
available data from Skukuza (below) casts uncertainty on
this seemingly reasonable scenario.
[37] This event was unique with regards to upper level
transport. In Figures 14a and 14b respectively are seven-day
air mass back-trajectories from 0000 UTC at Skukuza for
the ARREX (sample cases only, 17 days) and SAFARI
2000 (entire study, 27 days) periods. An intermediate upper
tropospheric starting level (7.50 km) was used. Transport
trajectories during ARREX were more persistent, and displaced southward relative to those calculated for SAFARI
2000. Air mass encounters with the South American continent during ARREX were more common than SAFARI
2000 (85% to 75%). They nearly all occurred within 3 5
days before reaching Skukuza (not implicitly shown, though
greater wind speeds can be inferred from Figures 14a and
14b), while between 30
and 35
S, compared to between 4
and 7 days and within 15
and 35
S the following year. The
southward displacement of the ARREX trajectories is in
disagreement with previous studies of South American
biomass burning particulate transport patterns into the
southern Atlantic basin [Nobre et al., 1998]. However, they
do suggest that the upper level environment was ideally
conducive to a persistent transport event, whatever the
source. Such persistence in the predominant flow over a
significant timescale is likely anomalous, which raises the
question as to the dynamic likelihood of any such intercontinental transport event occurring. Of course, this also
depends on the ultimate nature of the event, and the
efficiency at generating a significant amount of elevated
particulate matter, which could be advected great distances
under such ideal conditions. Perhaps more interesting to
contemplate is just how far these aerosols ultimately traveled downstream beyond southern Africa.

[38] MPL optical algorithm retrievals from these data

were limited, but offered some intriguing evidence. Successful retrievals from 7 September found that the contribution to total column AOD of the elevated layers
amounted to no more than approximately 0.05 0.03 at
523 nm (not shown). CIMEL optical depths from this day at
the same wavelength were approximately 0.22. AE calculations from 17 August to 10 September 1999 are shown in
Figure 15, along with the seasonal average values from
1999 and 2000, as derived in Figure 2. A sharp decrease in
this parameter corresponds with 7 September, with values
less than 1.30. Such low AE observations are not congruent
with measurements during biomass burning events in 2000,
although they do not deviate significantly. However, considering the lower contribution of the unknown aerosols to
total column optical depth, relative to the common haze
layers near the surface (approximately 25% on the subject
day), lower AE values indicate a secondary influence on
column scattering by larger particles that bias the parameter
to the lower value. From Figure 14, the lowest AE values
(0.8 and below) were observed on 25 August (MPL data not
shown). Elevated aerosols on that day were observed to near
10 km above ground, and their contribution to column AOD
was much higher (estimated near 50%). These data suggest
that the elevated aerosols consisted of relatively large
particles, which is again inconsistent with those found
during biomass burning episodes. Particle growth over time,
perhaps through cloud interaction [ODowd et al., 1999]
given the great distances from the original source region is
conceivably possible. To result in AE values less than 1.0,
however, would have required a considerable mean particle
size evolution.
[39] Aside from biomass burning, two alternate mechanisms have been identified as possible sources for the
aerosols. The first is based on the existence of two deserts
in southern South America, the Atacama (along coastal
Chile between 18
and 30
S) and Patagonia (between
roughly 73
64
W and 55
40
S shared between Chile
and Argentina). Grousset et al. [1992] and Iriondo [2000]
previously found evidence for intercontinental transport of
Patagonia dust matter to the Antarctic Peninsula. Current
global dust transport modeling efforts, such as the Georgia
Tech/Goddard Global Ozone Chemistry Aerosol Radiation
and Transport (GOCART) model consider both of these
deserts as source generating regions [Ginoux et al., 2001].
Surface winds in the Patagonia region during the subject
time are at a seasonal peak, coinciding with the influence of
wintertime extratropical cyclones [Lassig et al., 1999] (a
literary search on reference to similar phenomena for the
Atacama Desert found no corresponding studies). Though
AE values taken at face value are not consistent with those
of dust [e.g., Tanre et al., 2001], our previous reasoning of a
low bias in AE caused by scattering owing to larger
particles in the elevated layers makes dust a plausible
candidate.
[40] The second scenario involves a phreatic eruption
cycle of the Guagua Pichincha volcano (0
100S, 78
360W)
in Ecuador from June to September 1999 [Legrand et al.,
2002]. Multiple events were observed during this cycle,
with the most significant occurring on 24 August. Matter
was projected over 6 km above sea level during that
eruption. Volcanic aerosol transport has been the focus of

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

SAF

33 - 13

Figure 11. Corresponding to Figure 2 but expanded for 1 September 2000, without back-trajectory
classification table.

much recent research as a consequence of the major Mt.

Pinatubo eruption in June 1991. The optical characteristics
of aerosols associated with that event, as well as other less
significant eruptions, have been well documented [e.g.,
Russell et al., 1996; Schmid et al., 1997; Watson and
Oppenheimer, 2001]. The proximity to the equator of this
volcano, and subsequent lack of upper level steering flow at
that latitude (so as to be entrained into the trajectories
shown in Figure 14), and the large northerly displacement
relative to the calculated trajectory tracks lessens the likelihood for this event being the responsible source. However,
the timing and persistence of the eruptive activity, and the
ability of volcanic particle intrusions to generate aerosol
episodes significant on global scales are intriguing. Additionally, particle size distributions common to volcanic
aerosols are larger with respect to biomass aerosols and
could produce the observed AE bias previously noted.
[41] Though we cannot offer a definitive classification of
the nature of these aerosols, we are at least very confident
that their source was the South American continent. The

striking contrast in vertical structure and temporal extent of

these data to the more robust observations made the
following year are at worst still noteworthy however. The
statically stable nature of the southern African lower-troposphere should inhibit the prolonged protrusion of surfaceelevated particulate matter well above the height of the
uppermost inversion layers (500 hPa). However, the
extended presence of any particles, even with relatively
low optical significance, would have some quantifiable
effect on tropospheric column heating rates, thereby indirectly influencing the local meteorology and directly affecting surface radiative characteristics. As similar conditions
were not witnessed during SAFARI 2000, those regional
transport-modeling efforts, and radiative studies buoyed by
data sets collected during that experiment should pay careful consideration to these findings. If by chance this
scenario were actually a residual of local biomass burning,
we can conclude on the basis of the case study discussed in
the previous section that it would have required an extreme
event to generate the significant buoyancies capable of

SAF

33 - 14

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 12. MPL-derived extinction cross-section profiles with uncertainty bars versus height above
ground (km) for 30-min data periods on 1 September 2000; fractional day 245.2281 (a), 245.5011 (b) and
245.6071 (c). S for each case is noted.
penetrating the near-surface inversion layers over an
extended period.

5. Conclusions
[42] In this paper we describe observations made during
August and September for 1999 and 2000 at the Skukuza
Airport in northeastern South Africa in concurrence with the
ARREX and SAFARI 2000 field campaigns. The primary
instrument of focus is a micropulse lidar instrument (MPL)
supplemented with observations from colocated passive
radiometric equipment and derived air mass back-trajectories. The purpose of these measurements is to enhance our
understanding of the optical properties of regional aerosols,
which are heavily influenced by industrial sulfate emissions
and biomass burning. Specifically, regional characterization
of the extinction-to-backscatter ratio (S) is of particular
importance to satellite-based lidar programs as calibration
reference points. We have shown that during the dry season
months, there is great variability in this parameter as a
function of air mass origin.
[43] Under dense smoke/haze conditions, S was found to
vary between 50 and 90 sr. In west northwesterly upper
level flow, which increases the influence of continental
transport from prevalent upwind biomass burning locales,
the spectral Angstrom exponent (AE) was found to be
between 1.60 and 2.00. As air mass origins shifted at upper
levels to more westerly, thereby inducing greater advection
of urban and industrial pollutants, AE was found to reach a
maximum only approaching 1.90 with common values
much being much lower. In conditions affected by the
influence of maritime aerosol modification and recirculation, S values dropped to between 40 and 70 sr, while AE

values were between 0.80 and 1.60 in agreement with

previous work [e.g., Welton et al., 2000; Voss et al.,
2001]. The chaotic nature of the regional aerosol, based
on air mass origin, and the inherently sensitive nature of the
S owing to dependencies on various aerosol parameters
make this value difficult to simply distinguish within the
bounds of proposed seasonal look-up tables established to
initiate satellite-based single-channel lidar aerosol property
retrievals. AE values correlated much more clearly with
changes in air mass origin in this study.
[44] Two case studies are detailed. The first describes
the onset (morning/afternoon of 1 September 2000) of
advecting smoke from nearby fires observed by the lidar
and passive instrument array within the separate bounds of
two vertically distinct thermal layers. We show the evolution of this event both in vertical profiles of particulate
extinction cross section, S, and with respect to surface
broadband flux and Sun photometer measurements. Significant surface shortwave forcing at the surface, greater
than
50 W/m2, was measured in this event. 523 nm
aerosol optical depths exceeded 0.70 and AE values
approached 1.80. The stable nature of the southern African
lower troposphere and the prominence of multiple thermal
inversion layers make these observations unique in the
scope of global smoke/particulate transport. Additionally
the efficiency at which rising smoke plumes are able to
penetrate successive regional inversion layers, if and where
present, will have a direct effect on local vertical heating
rate profiles and surface flux rates. In this case, MPL
observations show that the peak extinction cross sections
in each layer were both roughly 0.30 km
1. It can be
concluded that a significant amount of the smoke matter
reached the upper layer in this case.

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

SAF

Figure 13. MPL normalized relative backscatter (counts*km2/uj*us) from 7 September 1999 at the
Skukuza Airport site (top), and seven-day NOAA HYSPLIT back-trajectory calculations for 6.00 and
9.00 km beginning at 0000 UTC from Skukuza (bottom).

33 - 15

SAF

33 - 16

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Figure 14. Seven-day air mass back-trajectories from 0000 UTC at 7.50 km from Skukuza for ARREX
(24 August to 9 September 1999) (a) and SAFARI 2000 (18 August to 13 September 2000) (b).
[45] Finally, we present MPL observations of persistent
elevated aerosols during ARREX. Analysis of daily sevenday back-trajectories from the period indicates that they
likely were the result of intercontinental transport between
southern South America and southern Africa. While previous researchers have shown the existence of biomass
burning aerosols advecting from central South America into
the southern Atlantic Ocean basin, ancillary passive measurements made at Skukuza are not in agreement with such a
scenario being responsible. Two alternate hypotheses are
presented: dust transport from the Patagonia and/or Atacama Deserts desert areas of southern South America, or
elevated matter from the Guagua Pichincha volcano in
Ecuador. The latter experienced significant eruptive activity
from June to September 1999. These observations are

unique in that upper tropospheric aerosols were not

observed the following year during SAFARI 2000. Comparison with similar seven-day back-trajectories during the
2000 experiment showed anomalies in air mass interaction
both in latitudinal displacement and length of day from
Skukuza. Upper level winds were found to be more favorable for a sustained particulate transport event in ARREX
than during SAFARI 2000. Should it be proven in subsequent analyses that these aerosols were the result of dust
transport, previous observations of intercontinental transport
from these deserts exists only to the Antarctic Peninsula
[e.g., Grousset et al., 1992; Iriondo, 2000]. Our findings,
though ultimately not wholly conclusive, are potentially
very noteworthy. Optical depth profiles derived using the
lidar and Sun photometer data from the most significant

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

SAF

33 - 17

Figure 15. Angstrom exponent calculations from available AERONET CIMEL Sun photometer data
for the period 17 August to 10 September 1999. Overlaid in the image are seasonal mean values for 1999
(solid) and 2000 (dashed) as described in Figure 2. 7 September 1999, as referenced in the text,
corresponds to fractional day 250.
periods of the event show that the elevated aerosols contributed approximately 0.05 0.03 at 523 nm, which is
significant with regards to affecting column heating rates.
[46] Deployment of the MPL instruments for these
experiments was coordinated through the recently formed
NASA Earth Observing System (EOS) Micropulse Lidar
Network project (MPLNET) [Welton et al., 2001b]. The
MPL-Net program was designed to establish a global network of MPL systems of long-term climatological cloud
and aerosol research. In the coming years, instruments will
be deployed at preexisting AERONET field sites to supplement colocated Sun photometer measurements with profiles
of vertical structure. Additionally, field campaigns will be
supported as deemed relevant. Data sets described in this
paper are available online, and potential users are encouraged to visit the project website to make specific inquiries.
The MPL-Net World Wide Web site can be accessed at
http://mplnet.gsfc.nasa.gov/.

Appendix A
[47] In this section we describe an automated algorithm
used to find the top of the surface aerosol layer from an
individual (as stored operationally) or group-averaged MPL
profile. When analyzing a single profile, or including it in a
group average, a simple signal threshold filter is applied to
every profile to avoid cloud contamination (D. D. Turner,
personal communication, 1998). The search begins by finding the first range bin from the instrument where the ratio of
NRB signal to NRB signal uncertainty is less than 2.00.
Roughly speaking, this relationship is signal-to-noise ratio
(SNR). However, SNR typically applies to the similar ratio
but for raw signal. Given that there are instrument correction
terms and related uncertainties mixed into this new value it

would be inappropriate to title it as such. However, the

general correlation is still valid. This corresponding range is
considered the top of the resolvable vertical column (i.e., the
range to which there is reliable signal in the profile). As a
value of 1.00 indicates the point of absolute uncertainty with
regards to interpreting a single bin, a value of 2.00 is used to
make the interpretation stringent.
[48] For all range bins from the instrument to the noise
bin a value for C is solved for using the derived AOD (to
solve the two-way particle transmission path), and a modeled molecular scattering profile (as described by Welton et
al. [2002]). Next, the resolution of the profile is degraded to
150 m and C is recalculated for each bin (with lower
uncertainty due to averaging) up to the noise bin. Clearly
this methodology is flawed, as C will not be correct/
applicable for every bin. However assuming that the entire
aerosol layer is contained within a finite range in the area
from the ground to the noise bin, C will converge to a
correct solution with range. In other words, C will be biased
artificially high within the aerosol layer, but for the range
past it, where backscatter is almost entirely molecular, the
calculation will be accurate. Therefore we use the minimum
value found in the degraded resolution profile (lower
uncertainty) and use it as a base from which to normalize
the individual bin values, such that
cr Cr=C*r*

A1

where C(r) is the C calculated at normal resolution, and

C*(r*) is the base normalization value at the lower resolution
(r*). The uncertainty of this relationship is given by


@cr cr*SQRT @Cr=Cr^ 2 @C*r*=C*r*^ 2
A2

SAF

33 - 18

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

In a noise-free environment, the first bin from the ground to

the noise bin where c is equal to 1 could be considered
free of aerosol and entirely composed of molecular
components. However, in considering equation (2) we are
restrained by the uncertainty of the signal and instead use the
first bin where
cr < 1 @cr

A3

to be the top of the aerosol layer. A 1.1 km range of bins

beginning 300 m above the top of the aerosol layer (to offset
small errors in the layer top search) is then designated as the
calibration zone for solving C [Welton and Campbell, 2002].
[49] This somewhat simple technique does not differentiate between individual aerosol layers, rather it finds
the top of the highest aerosol layer as determined by the first
range bin containing signal derived from only molecularscattering, as constrained by uncertainty. Two scenarios can
occur which undermine the methodology. First, if separate
elevated layers are present above the surface layer, such that
a well-defined clear slot is evident in the signal profile,
the algorithm will not search for their presence. Instead, it
will solve for C assuming that the CIMEL AOD measurement was due entirely to transmission in the surface layer.
Second, if the noise bin were to be found before reaching
the top of the surface layer (e.g., dense haze), the algorithm
would default to that point as representing the top of the
layer. If this range were not within 300 m of being accurate,
the algorithm would incorrectly choose the calibration zone.
Both of these circumstances produce clearly erroneous
values of C that are easily filtered out.
[50 ] Acknowledgments. The National Aeronautics and Space
Administration (NASA) Earth Observing System program, and the NASA
Sensor Intercomparison and Merger fund for Biological and Interdisciplinary Oceanic Studies support MPL research at the Goddard Space Flight
Center. The authors wish to thank Dennis L. Hlavka, William D. Hart,
Matthew McGill, Timothy D. Berkoff, and V. Stan Scott (NASA GSFC) for
helpful insights and dedicated work on the MPL program, and Kristy Ross
and Christoph Bollig (U. of Witwatersrand) for their assistance with the
instruments in the field. We also thank Mark Gray (NASA GSFC) of the
MODIS Atmosphere Team, who provided the composite image from 1
September 2000, and Lisa Harold (NASA GSFC), Wendy Job (Department
of Geography, Archaeology and Environmental Studies, Wits University),
and the Cougar Forum for their various forms of technical assistance. We
gratefully acknowledge the comments of two anonymous reviewers, which
significantly enhanced the original draft of this manuscript.

References
Ackermann, J., The extinction-to-backscatter ratio of tropospheric aerosol:
A numerical study, J. Atmos. Oceanic Technol., 15, 1043 1050, 1998.
Anderson, B. E., et al., Aerosols from biomass burning over the tropical
South Atlantic region: Distributions and impacts, J. Geophys. Res., 101,
24,117 24,137, 1996.
Anderson, T. L., S. J. Masonis, D. S. Covert, R. J. Charlson, and M. J. Rood,
In situ measurement of the aerosol extinction-to-backscatter ratio at a
polluted continental site, J. Geophys. Res., 105, 26,907 26,915, 2000.
Angstrom, A., The parameters of atmospheric turbidity, Tellus, 16, 64 75,
1964.
Ansmann, A., D. Althausen, U. Wadinger, K. Franke, D. Muller, F. Wagner,
and J. Heintzenberg, Vertical profiling of the Indian aerosol plume with
six-wavelength lidar during INDOEX: A first case study, Geophys. Res.
Lett., 27(7), 963 966, 2000.
Berg, O. H., E. Swietlicki, and R. Krejci, Hygroscopic growth of aerosol
particles in the marine boundary layer over the Pacific and Southern
Oceans during the First Aerosol Characterization Experiment (ACE 1),
J. Geophys. Res., 103, 16,535 16,545, 1998.
Browell, E. V., et al., Ozone and aerosol distributions and air mass characteristics over the South Atlantic Basin during the burning season,
J. Geophys. Res., 101, 24,043 24,068, 1996.

Campbell, J. R., D. L. Hlavka, J. D. Spinhirne, R. Ferrare, and D. D. Turner,

Automated aerosol retrieval algorithms for ARM micro pulse lidars, in
Preprints of Symposium on Lidar Atmospheric Monitoring, pp. 71 74,
American Meteorol. Soc., Boston, Mass., 2000.
Campbell, J. R., D. L. Hlavka, E. J. Welton, C. J. Flynn, D. D. Turner, J. D.
Spinhirne, V. S. Scott, and I. H. Hwang, Full-time, eye-safe cloud and
aerosol lidar observation at Atmospheric Radiation Measurement Program sites: Instruments and data processing, J. Atmos. Oceanic Technol.,
19, 431 442, 2002.
Fernald, F. G., Analysis of atmospheric lidar observations: Some comments, Appl. Opt., 23, 652 653, 1984.
Ferrare, R. A., D. D. Turner, L. Heilman Brasseur, W. F. Feltz, O. Dubovik,
and T. P. Tooman, Raman lidar measurements of the aerosol extinctionto-backscatter ratio over the Southern Great Plains, J. Geophys. Res., 106,
20,333 20,347, 2001.
Fuelberg, H. E., J. D. VanAusdall, E. V. Browell, and S. P. Longmore,
Meteorological conditions associated with vertical distributions of aerosols
off the west coast of Africa, J. Geophys. Res., 101, 24,105 24,115, 1996.
Garstang, M., P. D. Tyson, R. Swap, M. Edwards, P. Kallberg, and J. A.
Lindesay, Horizontal and vertical transport of air over southern Africa,
J. Geophys. Res., 101, 23,721 23,736, 1996.
Ginoux, P., M. Chin, I. Tegen, J. M. Prospero, B. Holben, O. Dubovik, and
S.-J. Lin, Sources and distributions of dust aerosols simulated with the
GOCART model, J. Geophys. Res., 106, 20,255 20,273, 2001.
Grousset, F. E., P. E. Biscaye, M. Revel, J. R. Petit, K. Pye, S. Joussaume,
and J. Jouzel, Antarctic ice-core dust at 19 k.y.B.P.: Isotopic constraints
on origins, Earth Planet. Sci. Lett., 111, 175 182, 1992.
Holben, B. N., et al., AERONETA federated instrument network and
data archive for aerosol characterization, Remote Sens. Environ., 66,
1 16, 1998.
Iriondo, M., Patagonian dust in Antarctica, Quat. Int., 68, 83 86, 2000.
Kent, G. S., C. R. Trepte, K. M. Skeens, and D. M. Winker, LITE and
SAGE II measurements of aerosols in the southern hemisphere upper
troposphere, J. Geophys. Res., 103, 19,111 19,127, 1998.
King, M. D., S. Platnick, C. C. Moeller, H. E. Revercomb, and S. A. Chu,
Remote sensing of smoke, land and clouds from the NASA ER-2 during
SAFARI 2000, J. Geophys. Res., 108, doi:10.1029/2002JD003207, in
press, 2003.
Lassig, J. L., M. G. Cogliati, M. A. Bastanski, and C. Palese, Wind characteristics in Neuquen, North Patagonia, Argentina, J. Wind Eng. Ind.
Aerodyn., 79(1 2), 183 199, 1999.
Legrand, D., A. Calahorrano, B. Guillier, L. Rivera, M. Ruiz, D. Villagomez, and H. Yepes, Stress tensor analysis of the 1998 1999 tectonic
swarm of northern Quito related to the volcanic swarm of Guagua Pichincha volcano, Ecuador, Tectonophysics, 344, 15 36, 2002.
Martins, J. V., P. V. Hobbs, R. E. Weiss, and P. Artaxo, Sphericity and
morphology of smoke particles from biomass burning in Brazil, J. Geophys. Res., 103, 32,051 32,057, 1998.
McGill, M., D. L. Hlavka, W. D. Hart, E. J. Welton, and J. R. Campbell,
Airborne lidar measurements of aerosol optical properties during SAFARI
2000, J. Geophys. Res., 108, doi:10.1029/2002JD002370, in press, 2003.
Nobre, C. A., L. F. Mattos, C. P. Dereczynski, T. A. Tarasova, and I. V.
Trosnikov, Overview of atmospheric conditions during the Smoke,
Clouds and Radiation-Brazil (SCAR-B) field experiment, J. Geophys.
Res., 103, 31,809 31,820, 1998.
ODowd, C. D., J. A. Lowe, and M. H. Smith, Observations and modeling
of aerosol growth in marine stratocumulusCase study, Atmos. Environ.,
33, 3053 3062, 1999.
Palm, S. P., W. D. Hart, D. L. Hlavka, E. J. Welton, and J. D. Spinhirne,
GLAS Atmospheric Data Products: Algorithm Theoretical Basis Document V4.1, 111 pp., NASA Goddard Space Flight Cent., Greenbelt, Md.,
2001.
Peppler, R. A., et al., ARM Southern Great Plains site observations of the
smoke pall associated with the 1998 Central American fires, Bull. Am.
Meteorol. Soc., 81, 2563 2591, 2000.
Piketh, S. J., H. J. Annegarn, and P. D. Tyson, Lower tropospheric aerosol
loadings over South Africa: The relative contribution of Aeolian dust,
industrial emissions, and biomass burning, J. Geophys. Res., 104, 1597
1607, 1999.
Powell, D. M., J. A. Reagan, M. A. Rubio, W. H. Erxleben, and J. D.
Spinhirne, ACE-2 multiple angle micro-pulse lidar observations from
Las Galletas, Tenerife, Canary Islands, Tellus, Ser. B, 52, 651 660, 2000.
Reid, J. S., P. V. Hobbs, R. J. Ferek, D. R. Blake, J. V. Martins, M. R.
Dunlap, and C. Liousse, Physical, chemical, and optical properties of
regional hazes dominated by smoke in Brazil, J. Geophys. Res., 103,
32,059 32,080, 1998.
Remer, L. A., Y. J. Kaufman, B. N. Holben, The size distribution of ambient aerosol particles: Smoke versus Urban/Industrial aerosol, in Biomass Burning and Global Change, edited by J. Levine, pp. 519 530,
Mass Inst. of Technol., Cambridge, Mass., 1996.

CAMPBELL ET AL.: MPL OBSERVATIONS FROM ARREX/SAFARI 2000

Russell, P. B., et al., Global to microscale evolution of the Pinatubo volcanic aerosol derived from diverse measurements and analyses, J. Geophys.
Res., 101, 18,745 18,763, 1996.
Schmid, B., C. Matzler, A. Heimo, and N. Kampfer, Retrieval of the optical
depth and particle size distribution of tropospheric and stratospheric aerosols by means of sun-photometry, IEEE Trans. Geosci. Remote Sens., 35,
172 182, 1997.
Schmid, B., et al., Coordinated airborne, spaceborne, and ground-based
measurements of massive, thick aerosol layers during the dry season in
Southern Africa, J. Geophys. Res., 108, doi:10.1029/2002JD002297, in
press, 2003.
Smith, W. L., et al., The Chesapeake Lighthouse and Aircraft Measurements for Satellites (CLAMS) Experiment, paper presented at the 11th
Satellite Meteorology and Oceanography Conference, Madison, Wisc.,
Am. Meteorol. Soc., Boston, Mass., Oct. 2001.
Spinhirne, J. D., Micro pulse lidar, IEEE Trans. Geosci. Remote Sens., 31,
48 55, 1993.
Spinhirne, J. D., J. A. Reagan, and B. A. Herman, Vertical distribution of
aerosol extinction cross section and inference of aerosol imaginary index
in the troposphere by lidar technique, J. Appl. Meteorol., 19, 426 438,
1980.
Spinhirne, J. D., J. A. R. Rall, and V. S. Scott, Compact eye safe lidar
systems, Rev. Laser Eng., 23, 112 118, 1995.
Swap, R. J., M. Garstang, S. A. Macko, P. D. Tyson, W. Maenhaut,
P. Artaxo, P. Kallberg, and R. Talbot, The long-range transport of southern African aerosols to the tropical south Atlantic, J. Geophys. Res., 101,
23,777 23,791, 1996.
Swap, R. J., et al., The Southern African Regional Science Initiative (SAFARI 2000) dry-season field campaign: An overview, S. Afr. J. Sci., 98,
125 130, 2002.
Takamura, T., Y. Sasano, and T. Hayasaka, Tropospheric aerosol optical
properties derived from lidar, Sun photometer, and optical particle counter measurements, Appl. Opt., 33, 7132 7141, 1994.
Tanre, D., Y. J. Kaufman, B. N. Holben, B. Chatenet, A. Karnieli,
F. Lavenu, L. Blarel, O. Dubovik, L. A. Remer, and A. Smirnov, Climatology of dust aerosol size distribution and optical properties derived
from remotely sensed data in the solar spectrum, J. Geophys. Res.,
106, 18,205 18,217, 2001.
Terblanche, D. E., M. P. Mittermaier, S. J. Piketh, R. T. Bruintjes, and R. P.
Burger, The Aerosol Recirculation and Rainfall Experiment (ARREX):
An initial study on aerosol-cloud interactions over South Africa, S. Afr. J.
Sci., 96, 15 21, 2000.
Trentmann, J., M. O. Andreae, H.-F. Graf, P. V. Hobbs, R. D. Ottmar, and
R. Trautmann, Simulation of a biomass-burning plume: Comparison of

SAF

33 - 19

model results with observations, J. Geophys. Res., 107, 410 414,

2002.
Tyson, P. D., M. Garstang, and R. Swap, Large-scale recirculation of air
over southern Africa, J. Appl. Meteorol., 35, 2218 2236, 1995.
Tyson, P. D., M. Garstang, R. Swap, P. Kallberg, and M. Edwards, An air
transport climatology for subtropical southern Africa, Int. J. Climatol.,
16, 265 291, 1996.
Voss, K. J., E. J. Welton, P. K. Quinn, J. Johnson, A. Thompson, and H.
Gordon, Lidar measurements during Aerosols99, J. Geophys. Res., 106,
20,821 20,832, 2001.
Watson, I. M., and C. Oppenheimer, Photometric observations of Mt. Etnas
different aerosol plumes, Atmos. Environ., 35, 3561 3572, 2001.
Welton, E. J., and J. R. Campbell, Micro-pulse lidar signals: Uncertainty
analysis, J. Atmos. Oceanic Technol., 19, 2089 2094, 2002.
Welton, E. J., et al., Ground-based lidar measurements of aerosols during
ACE-2: Instrument description, results and comparisons with other
ground-based and airborne measurements, Tellus, Ser. B, 52, 635 650,
2000.
Welton, E. J., J. R. Campbell, T. A. Berkoff, J. D. Spinhirne, and P. Ginoux,
Initial results from the micro-pulse lidar network (MPL-Net), paper presented at the Fall Meeting of the American Geophysical Union, San
Francisco, Calif., Dec. 2001a.
Welton, E. J., J. R. Campbell, J. D. Spinhirne, and V. S. Scott, Global
monitoring of clouds and aerosols using a network of micro-pulse lidar
systems, Proc. Int. Soc. Opt. Eng., 4153, 151 158, 2001b.
Welton, E. J., K. J. Voss, P. K. Quinn, P. J. Flatau, K. Markowicz, J. R.
Campbell, J. D. Spinhirne, H. R. Gordon, and J. E. Johnson, Measurements of aerosol vertical profiles and optical properties during INDOEX
1999 using micropulse lidars, J. Geophys. Res., 107, 8019, doi:10.1029/
2000JD000038, 2002.

M. Barenbrug and S. J. Piketh, University of Witwatersrand, East

Campus, Private Bag X3, WITS 2050, South Africa. (margie@crg.bpb.
wits.ac.za; stuart@crg.bpb.wits.ac.za)
J. R. Campbell, J. D. Spinhirne, and E. J. Welton, NASA Goddard Space
Flight Center, Code 912, Greenbelt, MD 20771, USA. (campbell@
virl.gsfc.nasa.gov; jspin@virl.gsfc.nasa.gov; Ellsworth.J.Welton@nasa.
gov)
B. Holben, NASA Goddard Space Flight Center, Code 923, Greenbelt,
MD 20771, USA. (brent@aeronet.gsfc.nasa.gov)
Q. Ji and S.-C. Tsay, NASA Goddard Space Flight Center, Code 913,
Greenbelt, MD 20771, USA. ( ji@climate.gsfc.nasa.gov; tsay@climate.
gsfc.nasa.gov)