Click
Here
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, D08207, doi:10.1029/2006JD007767, 2007
for
Full
Article
Numerical simulation of the October 2002 dust event
in Australia
Yaping Shao,1 John F. Leys,2 Grant H. McTainsh,3 and Kenn Tews3
Received 7 July 2006; revised 17 October 2006; accepted 21 November 2006; published 24 April 2007.
[1] In comparison to the major dust sources in the Northern Hemisphere, Australia is a
relatively minor contributor to the global dust budget. However, severe dust storms do
occur in Australia, especially in drought years. In this study, we simulate the
22–23 October 2002 dust storm using an integrated dust model, which is probably the
most severe dust storm in Australia in at least the past 40 years. The model results
are compared with synoptic visibility data and satellite images and for several
stations, with high-volume sampler measurements. The model simulations are then used to
estimate dust load, emission, and deposition, both for over the continent and for over
the ocean. The main dust sources and sinks are identified. Dust sources include the desert
areas in northern South Australia, the grazing lands in western New South Wales
(NSW), and the farm lands in NSW, Victoria, and Western Australia, as well as areas in
Queensland and Northern Territory. The desert areas appear to be the strongest source.
The maximum dust emission is around 2000 mg m2 s1, and the maximum net dust
emission is around 500 mg m2 s1. The total amount of dust eroded from the
Australian continent during this dust event is around 95.8 Mt, of which 93.67 Mt is
deposited on the continent and 2.13 Mt in the ocean. The maximum total dust load over
the simulation domain is around 5 Mt. The magnitude of this Australian dust storm
corresponds to a northeast Asian dust storm of moderate size.
Citation: Shao, Y., J. F. Leys, G. H. McTainsh, and K. Tews (2007), Numerical simulation of the October 2002 dust event in
Australia, J. Geophys. Res., 112, D08207, doi:10.1029/2006JD007767.
1. Introduction
[2] Australia is a minor dust source to the global atmosphere in comparison to North Africa, the Middle East and
northeast Asia [Prospero et al., 2002]. The numerical study
of Tanaka and Chiba [2006] suggests that the annual dust
emission from Australia is about 106 Tg in contrast to
1087 Tg from North Africa and 575 Tg from Asia (including Arabian Peninsula, Central Asia and China). The
maximum dust event frequency is around 5 yr1, which
occurs in central Australia [McTainsh and Pilblado, 1987].
However, severe dust storms do occur under certain atmospheric and land surface conditions, especially in drought
years. Past severe events include the 8 February 1983
Melbourne, the 1 December 1987 Brisbane, the 8 February
1996 central Australia dust storms [Raupach et al., 1994;
Knight et al., 1995; Shao and Leslie, 1997]. According to
McTainsh et al. [2005], the 22– 23 October 2002 dust storm
1
Institute of Geophysics and Meteorology, University of Cologne,
Cologne, Germany.
2
Department of Natural Resources, Gunnedah, New South Wales,
Australia.
3
Faculty of Environmental Sciences, Griffith University, Brisbane,
Queensland, Australia.
Copyright 2007 by the American Geophysical Union.
0148-0227/07/2006JD007767$09.00
in eastern Australia was the largest in at least the past
40 years: It covered the largest area of the continent and
carried the largest dust load.
[3] McTainsh et al. [2005] and Chan et al. [2005] studied
the 22– 23 October 2002 dust event in detail and estimated
the dust load from dust concentration derived from synoptic
visibility data using an empirical relationship. They found
that the total dust load associated with the dust storm falls
between 3.35 and 4.85 Mt. However, questions such as
the distribution and intensity of dust emission and dust
deposition, dust storm structure, etc., could not be satisfactorily answered through synoptic analysis alone. We are
thus motivated to carry out a numerical simulation of the
22– 23 October 2002 dust event to provide estimates of
these quantities.
[4] Since the late 1980s, integrated models for global,
regional and local dust problems have been constructed,
which couple modules for atmospheric, land surface and
aeolian processes with land surface parameter databases, so
that the main processes and environmental factors for dust
storm development are accounted for. Early attempts on
dust modeling were made by Westphal et al. [1988], Gillette
and Hanson [1989], and Joussaume [1990]. Examples of
global dust models include the studies of Zender et al.
[2003], Ginoux et al. [2004], and Tanaka and Chiba [2006].
Examples of regional dust models include the studies of
Wang et al. [2000], Nickovic et al. [2001], Uno et al. [2001],
and Liu et al. [2003]. Seino et al. [2005] simulated dust
D08207
1 of 13
SHAO ET AL.: DUST EVENT IN AUSTRALIA
D08207
D08207
storms in the Tarim Basin using a mesoscale model, and
Uno et al. [2005] applied the multinesting technique to
modeling dust events in the Tarim Basin. Shao et al. [2003]
applied an integrated dust models to the real-time prediction
of Asian dust storms. However, most of the recent modeling
works are concerned with North African and northeast
Asian dust storms, not Australian dust storms. Several
numerical studies on Australian dust storms were carried
out in the late 1990s [Shao and Leslie, 1997].
[5] In this paper, we present a numerical simulation of the
22– 23 October 2002 Australian dust event. The objectives
of the study are (1) to estimate the amount of dust emission,
load and deposition, so that Australian dust storms can be
put into perspective with respect to those in other areas of
the globe; and (2) to investigate the synoptic features of
Australian dust storms, including dust storm structure and
dust motion in the atmosphere.
Q is the saltation flux of ds; for a given ds, Q can be
calculated using the Owen [1964] model; g is acceleration
due to gravity; sp = pm(dt)/pf (dt) with pm(di) and pf (di)
being the minimally and fully dispersed particle size distributions, respectively. The Owen [1964] model predicts Q
for given ds as follows:
2. Dust Model
where rb is soil bulk density and p is soil plastic pressure.
The rate of dust emission for the ith particle size group is
determined by
[6] The dust model used in this study is Computational
Environmental Modeling System 5 (CEMSYS5), which
consists of an atmospheric model, schemes for land surface
processes, wind erosion, dust transport and deposition, and
a geographic information system (GIS) database. The atmospheric model of CEMSYS5 is a limited area model
[Leslie and Wightwick, 1995]. The original version of model
has been modified considerably and now has advanced
treatments for atmospheric dynamic and physic processes,
including advection, convection, turbulent diffusion, radiation, clouds and the atmospheric boundary layer. The land
surface scheme used in the study is Atmosphere and Land
Surface Interaction Scheme (ALSIS [Irannejad and Shao,
1998]). ALSIS simulates soil moisture and temperature in
the unsaturated zone and land surface energy, mass and
momentum fluxes.
[7] Dust is divided into I particle size groups. The wind
erosion scheme of Shao [2001, 2004] is used to estimate
wind erosion threshold friction velocity, saltation flux and
dust emission rates for different particle size groups. In this
paper, we have set I = 6, although the model formally
allows an arbitrary choice of I. The representative particle
diameters for the six groups are 2, 11, 22, 52, 90 and
125 mm. The scheme takes into consideration of three dust
emission mechanisms (aerodynamic entrainment, saltation
bombardment and aggregates disintegration) and predicts
that
~ ðdi ; ds Þ ¼ cy hfi ð1 g Þ þ gsp ð1 þ sm Þ gQ
F
u2*
ð1Þ
~ is the dust emission rate for the ith size group with
where F
size di generated by the saltation of particles of size ds, cy is
a dimensionless coefficient, hfi is the mass fraction of dust in
the parent soil for the ith size group (if in a soil, for example
pure sand, dust of size di is not present, i.e., hfi = 0, then dust
emission of size di is not possible) and g is a weighting
function which satisfies
g¼
1
0
u* ! u*t
u* ! 1
ð2Þ
8
!
u2
>
>
< co r u 3 1 *t
u2
g *
Q¼
*
>
>
:0
u > u*t
*
ð3Þ
u* u*t
with co being a dimensionless coefficient of order 1. The
bombardment efficiency sm is estimated by
rffiffiffiffiffi
r
rb
sm ¼ 12u2 b 1 þ 14u*
* p
p
^ ðdi Þ ¼
F
Z
d2
~ ðdi ; d Þdd
F
ð4Þ
ð5Þ
d1
where d1 and d2 are the lower and upper limit of saltation
particle size. The total dust emission rate is
F¼
i¼I
X
^ ðdi Þ
F
i¼1
Shao [2004] compared the predicted and measured F for
several soil texture and surface conditions, and found that cy
falls between 105 and 5 105 and p between 1000 and
50000 Pa.
[8] The threshold friction velocity for the surface u*t is
expressed as
u t ¼ u t0 ðds Þfl ðlÞfw ðwÞfsc ðsc Þfcr ðcr Þ::
*
*
ð6Þ
where u*t0(ds) is the threshold friction velocity for sand
particle of size ds in the idealized situation when soil is dry,
bare and free of salt and crust. In this study, u*t0(ds) is
calculated using the expression given by Shao and Lu
[2000]. The multiplicators fl, fw, fsc and fcr are correction
functions for surface roughness elements, soil moisture, salt
concentration and crust, respectively. The correction function fl follows Raupach [1992] and Raupach et al. [1993]
and fw follows Fécan et al. [1999]. At this stage, we know
little about salt concentration and soil surface crust and have
therefore set fsc and fcr to one.
[9] Dust transport through advection, (subgrid) convection and diffusion has been considered. The evolution of
dust concentration obeys the mass conservation equation,
which is numerically solved simultaneously with the atmospheric model equations. The particle eddy diffusivities in
the conservation equation are calculated following Csanady
[1963]. The treatment of subgrid convective dust transport
built upon the Kain and Fritsch [1993] scheme is as
described by Jung et al. [2005]. The dry deposition velocity
2 of 13
SHAO ET AL.: DUST EVENT IN AUSTRALIA
D08207
D08207
Table 1. A Comparison of Model-Simulated and Observed Dust Concentration for the October 2002 Australian Dust Eventa
Name
Buronga
Sydney
Brisbane
Gladstone
Mackay
Location
142.2E,
151.2E,
153.1E,
151.3E,
149.2E,
34.2S
33.8S
27.4S
23.9S
21.1S
Time (October 2002) UTC
0000
0028
0225
0058
0000
0300
0700
1800
2200
19
21
22
23
24
23
23
23
23
to
to
to
to
to
to
to
to
to
0000
0215
0042
0026
0000
0900
2300
0200
0500
21
22
23
24
28
23
23
24
24
Observed, mg m3
32
75
446
114
28
130
264
277
684
(266,
(841,
(537,
(899,
0500 23)
1100 23)
2100 23)
0100 24)
Model, mg m3
51 (0000 20 to 0000 21)
45
344
154
56 (0000 24 to 0000 25)
20 (24, 0500 23)
559 (776, 1500 23)
306 (365, 2100 23)
138 (163, 0000 24)
a
For Buronga, average TSP data, while for Sydney, Brisbane, Gladstone and Mackay, average PM10 data, as reported by Chan et al. [2005] are used for
comparison. For the four coastal cities, the measured peak value and the time (in UTC) of its occurrence are listed in parentheses.
is parameterized following Raupach et al. [2001], and the
(below cloud) wet deposition is as described by Jung and
Shao [2006].
[10] The area of simulation is (110°E, 45°S) to (170°E,
5°S) with a spatial resolution of 50 km. The s coordinate
system is used and the atmosphere is divided into 25 vertical layers with smaller increments near the surface. The
National Centers for Environmental Prediction – National
Center for Atmospheric Research (NCEP-NCAR) reanalysis data are used for specifying the initial and boundary
conditions for the atmospheric model of CEMSYS5. The
boundary conditions are updated every 6 hours. CEMSYS5
is run for October 2002 by making sequential 24 hour runs,
i.e., at the beginning of each 24 hour run, the atmospheric
variables (temperature, wind, humidity, etc.) are initialized
using the NCEP-NCAR reanalysis, while the dust variables
are initialized using the predictions of the previous day. The
atmospheric, land surface and dust fields are updated every
two minutes, but all outputs are hourly averages.
[11] The surface is screened and divided into the categories of nonerodible and erodible lands. Dust emission is
calculated only for the latter category and in this case, the
land surface data are used to calculate threshold friction
velocity (u*t), sand drift and dust emission. The calculation
of u*t requires roughness (mainly vegetation) frontal area
index l and soil moisture as inputs. The latter is obtained
from the ALSIS simulation for the top 50 mm soil layer. In
this study, l is derived from a combination of satellite
Normalized Differential Vegetation Index (NDVI) data and
vegetation-type data. Empirical relationships are used to
estimate l from advanced very high resolution radiometer
(AVHRR) NDVI grid data.
[12] The soil- and vegetation-type data are essentially
those of National Soil Resources Institute [1980] and
Australian Surveying and Land Information Group
(AUSLIG) [1990]. The spatial resolution of the data is
5 5 km. Australian soils are classified into 28 soil classes,
e.g., shallow permeable sandy soil, deep massive earths,
cracking clay soils with low permeability, etc. For each soil
class, there is a qualitative description of the soil properties
and landforms. The 28 soil classes are regrouped into
11 USDA (United States Department of Agriculture) soil
texture classes ranging from sand to clay [Shao et al., 1997].
A number of soil samples have been collected in Australia
and their pm(d) and pf(d) are determined through laboratory
analysis [McTainsh et al., 1997].
[13] The Australian vegetation is divided into 35 classes
according to height, density and number of canopy layers.
Among the 35 classes, the most extensive cover, 31.5% of
the continent, is tall shrublands in sparse open form,
followed by low woodlands and low open woodlands,
27%, and other medium and short vegetation, 22%. On
the basis of the vegetation data, estimates are made for
quantities such as vegetation height, albedo and minimum
stomatal resistance.
[14] The horizontal resolution of the dust fields of concentration, emission, etc., is determined by those of the
atmospheric model and the GIS data. In this study, the
atmospheric model resolution is 50 km. The GIS data sets,
with various resolutions as described above, are formally
unified at a resolution of 5 km by applying spatial interpolation to the lower resolution data and averaging to the
higher resolution data. To increase the accuracy in the
calculation of dust emission, the mosaic subgrid closure is
applied to the dust scheme. For each atmospheric model
cell, the land surface cells of the same soil type and similar
vegetation cover are grouped into a tile. Typically an
atmospheric model cell contains up to 10 tiles. The dust
emission rate for each tile is calculated and the atmospheric
model cell is calculated as the tile-area-weighted average of
the rates for the tiles.
3. Comparison With Dust Measurements
3.1. In Situ Measurements
[15] In situ measurements of dust concentration or total
suspended particulates (TSP) were made by a high-volume
sampler during the event in one of the dust source regions at
Buronga in southwest New South Wales (NSW). Data on
downwind dust concentrations (PM10) were available for
the coastal cities of Sydney, Brisbane, Gladstone, and
Mackay [Chan et al., 2005]. All measurements are daily
averages. A comparison of the model estimates and the TSP
and PM10 measurements are shown in Table 1.
[16] For Buronga, the observed and simulated dust concentrations are consistent, both in magnitude and temporal
variation. At this site, the highest dust concentration
occurred on 22 October 2002. For this day, the measured
average TSP was 446 mg m3 and the model predicted was
344 mg m3. When the dust storm passed through Buronga
between 0930 and 1200 UTC 22 October 2002, the visibility was around 4.8 km and the corresponding dust concentration estimated using equation (10) (see section 3.2) was
around 1200 mg m3. The model simulation suggests that
the high dust concentration region at this time was located
to the north and northwest of Buronga. At Buronga, the
predicted peak dust concentration was 700 mg m3, but
3 of 13
SHAO ET AL.: DUST EVENT IN AUSTRALIA
D08207
D08207
network is patchy, with higher densities in the eastern half
of the continent than in the west. Dust-derived visibility
data only were used, by selecting visibility data with an
accompanying dust code at the time of observation or less
than 3 hours since the previous observation time. Data
availability throughout the event varied depending on the
number of stations making observations at the time. For
example, for the four times of the event discussed in
section 3.3, the percentage of stations recording ranged
from 44% at 2100 UTC 22 October to 94% at 0000 UTC
23 October 2002. These two issues limit the utility of the
visibility data.
[19] Visibility can be converted to dust concentrations
concentration using empirical relationships. As weather
stations are relatively widely distributed, the visibility data
are currently the most suitable data for verifying the
predicted pattern of near-surface dust distribution. Different
empirical relationships for the estimation of dust concentration, C, from visibility, V, have been proposed by Chepil
and Woodruff [1957], Patterson and Gillette [1977], Tews
[1996], and Shao et al. [2003]:
Chepil and Woodruff ½1957
Figure 1. Comparison of model predicted PM10 concentration (in mg m3 at four eastern Australian coastal cities,
including (a) Mackay; (b) Gladstone; (c) Brisbane, and
(d) Sydney.
Patterson and Gillette ½1977
Tews ½1996
3
about 4000 mg m at Deniliquin (about 250 km southeast
of Buronga) and 7000 mg m3 at Broken Hill (about 250 km
northwest of Buronga), but no observations were available
at these locations for comparison.
[17] As the four cities of Sydney, Brisbane, Gladstone and
Mackay are all located at the eastern coast of Australia, dust
observed there was transported from inland areas. In
Figure 1, the predicted and observed hourly PM10 concentration for the four cities are compared. The observations are
as reported by Chan et al. [2005]. In general, the model
reasonably well captured the timing and intensity of the dust
front passage through the four cities. For Brisbane and
Gladstone, the predicted temporal evolution and magnitude
of the PM10 concentration are consistent with the measurements, but for Brisbane, the predicted dust episode lasted
longer than the observed. This is likely an model artifact
caused by the rather coarse model resolution of 50 km. The
PM10 measurement and various other reports indicate that
dust briefly affected Sydney in the morning of 23 October
2002, with PM10 concentration reaching 250 mg m3. The
model correctly predicted the time of the dust arrival, but
under predicted the concentration. For Mackay, the model
under predicted the concentration by a factor of 5, but the
temporal evolution is well reproduced. In fact, if the
predicted PM10 concentration for Mackay is multiplied by
5, then the predictions and observations would agree quite
well. The under prediction for Mackay is probably due to
the under prediction of dust emission in north Queensland.
3.2. Visibility Data
[18] Visibility data were available from 96 weather stations in Australia. The spatial coverage of the station
Shao et al: ½2003
C¼
C ¼ ð7078=V Þ1:25
ð7Þ
C ¼ ð10507=V Þ1:07
ð8Þ
C ¼ ð2032=V Þ0:877
ð9Þ
3802:29V 0:84
V < 3:5
expð0:11V þ 7:62Þ V 3:5
ð10Þ
where V is in [km] and C in [mg m3]. However, dust
concentration estimated in this way has large uncertainties
for three reasons: (1) the data set used for deriving the
empirical equation is small; (2) the dependency of visibility
on dust concentration is affected by dust particle size and
humidity; and (3) visibility measurements are subjective and
often inaccurate. For a given value of V, the estimates of C
using the above four equations may greatly differ. In this
study, equation (10) is used to convert visibility to dust
concentration for comparison with model predictions, the
reason being that it is derived based on a more reliable set of
measurements. Equations (7) and (8) tend to over estimate
dust concentration, while equation (9) tends to under
estimate dust concentration.
[20] Figure 2 shows a comparison of the simulated
and visibility-derived near-surface dust concentration at
four different times. The basic features of the predicted
and observed dust patterns are consistent. At 0600 UTC
22 October 2002, dust storms were developing in South
Australia (SA) and Victoria (Vic). For the next 15 hours,
the model predicted a rapid intensification and northeastward propagation of dust storms. By 0000 UTC 23 October
2002, dust storms were wide spread with the highest dust
concentration occurring in southern Queensland (Qld),
northern SA and northwestern New South Wales (NSW).
In the next 6 hours, the dust storms were moving further
4 of 13
D08207
SHAO ET AL.: DUST EVENT IN AUSTRALIA
Figure 2. A comparison of the simulated and visibility-derived near-surface dust concentrations.
Triangles represent the locations where dust weather was observed.
Figure 3. (a) Simulated dust concentration on 900 hPa at 0400 UTC 23 October 2002. Dots represent
the current weather dust reports for the period between 0000 UTC and 0600 UTC 23 October 2002; dots
in deeper color indicate lower visibility. (b) Same as Figure 3a, but for 850 hPa. (c) Same as Figure 3a,
but for 700 hPa. For visualization, dust concentration on 700 hPa is multiplied by 5. (d) GMS image for
0426 UTC 23 October 2002. (e) NOAA16 image for 0350 UTC 23 October 2002. To facilitate
comparison, dust cloud sections are marked with A, B, C, D, E, and F.
5 of 13
D08207
D08207
SHAO ET AL.: DUST EVENT IN AUSTRALIA
D08207
Figure 4. (a) Near-surface velocity field, dust concentration [mg m3], and temperature [K] for 1800 UTC
22 October 2002; (b) Same as Figure 4a, but for 0000 UTC 23 October 2002.
northeastward, affecting even larger areas, but gradually
weakening.
[21] Some discrepancies exist between the predicted and
observed dust patterns. For example, according to the model
prediction, dust concentration was highest at 2100 UTC
22 October 2002, and dust was wide spread. However, dust
weather was reported only for a small number of stations. A
possible explanation for this discrepancy is that at that time,
the dust-affected areas were mostly desert where few
weather stations exist. Further, at 2100 UTC (0600 LT
23 October (Eastern Standard Time)) routine observations
were made only at 44% of the weather stations. Judging
from the evolution of the dust event, the predicted dust
pattern is reasonable. Another discrepancy is that the model
did not predict the observed dust pattern in the vicinity of
Sydney on 23 October 2002. It is certain that the dust
observed there was transported in the first instance from the
northwest. The dust probably originated from the agricultural areas in inland NSW and Vic, which was advected
toward northeast and then southwest to the Sydney region,
following the clockwise-curved streamlines of the trough. It
is likely that the dust emission from the agricultural areas
in inland NSW and Vic was underestimated due to the
inaccuracies of land surface data for these areas.
3.3. Satellite Images
[22] The dust storm was observed by both the GMS and
the SeaWIFS satellites (Figure 3). From the SeaWIFS image
for 0000 UTC 23 October 2002 [McTainsh et al., 2005], a
dust front can be seen with its maximum extent being about
2400 km long and up to 400 km across. According to
McTainsh et al. the dust ceiling was about 1.5 to 2.5 km in
height. A comparison of the GMS image with the predicted
dust concentration patterns on 900, 850 and 700 hPa
(approximately 1000, 1500 and 3000 m above ground) is
shown in Figure 3. Air column dust load has also been
calculated (not shown) and it is found that the pattern of
dust load is very similar to that shown in Figure 3a. This is
not surprising because in the early stages of the dust storm
development, much of the dust is confined to the lower
layers of the atmosphere. The predicted dust pattern is in
good agreement with the satellite image. Relatively high
dust concentration occurred over Qld and northeast NSW.
The vertical structure of the dust clouds was rather complex.
Near the surface, dust was wide spread. Dust concentration
decreased rapidly with height. At 700 hPa level, only a
narrow dust band along the eastern coast of Australia can be
seen. Unfortunately, the near-surface dust distribution cannot be ambiguously identified from the satellite image, as
both the temperature and the color of the near-surface dust
were probably close to those of the ground surface. The
predicted dust field suggests that the dust front stretched
from the eastern coast of Australia, over Northern Territory
(NT) and well into Western Australia (WA). This is consistent with the near-surface observations. The western part of
the dust front (section C) could not be clearly seen, but a
faint dark band was identifiable. Also, the predicted dust
front section B stretching toward the Gulf of Carpentaria
could not be clearly seen from the GMS image (Figure 3d)
apart from a light colored faint band. However, section B
can be clearly seen from the NOAA16 satellite image for
0350 UTC 23 October 2002. The model prediction suggests
that along section B, the dust clouds were lower than along
sections A and F.
[23] The model also predicted dust over the western WA
(near NW Cape). The Earth Probe TOMS aerosol index for
the time period was high in that area (not shown) and it is
very likely that the predicted dust indeed occurred. Unfor-
6 of 13
D08207
SHAO ET AL.: DUST EVENT IN AUSTRALIA
tunately, several fires were also burning at the time and it is
thus difficult to unequivocally identify dust from the TOMS
aerosol index.
4. Dust Storm Structure
[24] The dust storm was closely related to the northeastward propagation of a cold front. At 0000 UTC 22 October,
a low over the Southern Ocean was centered at (140°E,
50°S) and a high was located just to the south of Perth. The
cold front extended from the low-pressure center to the
Great Australia Bight. Strong post frontal southerly airflow
was directed onto the southern coast of the continent. This
low-pressure system was preceded by a prefrontal trough
which extended from the Eyre Peninsula north to northwest
Australia. As a consequence, moderate to a strong hot dry
north westerly was blowing over parts of SA, Vic and
western NSW. Dust storms were reported in northern Vic
and western NSW. In the next 48 hours, the low was
moving eastward rapidly. This was accompanied by the
rapid northeastward propagation of the cold front and the
high-pressure system behind it. By 1200 UTC 23 October
2002, the entire Australian continent was dominated by the
high-pressure system.
[25] Figure 4 shows the simulated near-surface wind field
and dust concentration for 1800 UTC 22 October 2002 and
0000 UTC 23 October 2002. Comparisons with synoptic
observations confirm that the flow field is well simulated.
At that time, the cold front was an arc-shaped narrow region
stretching from NSW to WA, as can be seen from the dense
temperature contours (for simplicity only five contours are
drawn). Strong southwesterly wind (maximum 12 ms1)
prevailed behind the cold front resulting in strong dust
emission and high dust concentration. The dust was carried
northeastward by the southwesterly to converge to the
frontal area, resulting in high dust concentration near the
front. The dust was then transported northwestward and
southeastward along the front. As a consequence of this
converging (normal to front) and diverging (along front)
flow pattern, the dust storm affected a large area, although
the area of dust emission was much smaller.
[26] In eastern Australia, the dust front was advancing
rapidly northeastward. Within the next four hours, the
dust front advanced more than 350 km. At 0000 UTC
23 October, dust was wide spread, although the maximum
concentration was somewhat lower than that at 1800 UTC
22 October. According to the surface observations, at about
0000 UTC 23 October, the dust front passed over Sydney
reducing visibility at the Sydney airport to 6 km. The model
did not quite correctly predict the very southern section of
the dust front.
[27] The airflow and dust concentration fields on surface,
950 hPa, 850 hPa and 600 hPa are shown in Figure 5
for 2000 UTC 22 October 2002. Close to the surface
(Figures 5a and 5b), the cyclonic flow associated with the
low-pressure system (located to the south of Tasmania)
dominated much of Australia, while the anticyclonic flow
affected mainly WA. The anticyclonic flow became more
prominent in the upper levels, while the cyclonic flow was
confined to southeast Australia. The center of the anticyclone was located over northwest Australia. On 600 hPa,
dust concentration was rather low. This flow pattern enabled
D08207
the dust particles entrained in SA and central Australia to
be transported in three directions. (1) In southeast Australia,
dust particles were carried by the cyclonic flow toward the
cyclone center, as indicated by the relatively high dust
concentration over Tasmania; (2) near the surface, dust
particles were transported from central Australia northwestward along the front toward WA; and (3) at the upper
levels (e.g., 850 and 600 hPa) over the same area, flow was
mainly southwesterly and dust particles were transported
northeastward. These rather interesting features revealed
from the numerical modeling are identifiable from the
satellite images shown in Figure 3.
5. Sources and Sinks
[28] Suppose dust emission rate is FS and dust deposition
rate is FD, then net dust emission can be defined as
FN ¼ FS FD
Both FS and FD are calculated by the model. The simulated
dust emission, i.e., FS averaged over a 24 hour period for 22
and 23 October 2002 are shown in Figures 6a and 6b,
respectively. On 22 October (Figure 6a), the main dust
sources were the desert areas in northern SA (S1), the
grazing lands in western NSW (S2, with mining areas near
Cobar and Broken Hill) and the farming areas in NSW and
Vic (S3). Net dust emission also occurred in WA (S8). S1
was the strongest source, with the maximum dust emission rate exceeding 2000 mg m2 s1. Dust emissions
from S2 and S3 were weaker. Apart from the dust source
in WA (S8), dust emission occurred mainly behind the
cold front where wind speed exceeded 6 m s1. Overall, on
22 October, dust emission dominated. On 23 October
(Figure 6b), dust emission was spread much wider but
weaker in intensity. New dust sources S4, S5 and S6
emerged in NSW, Qld and NT. S4, S5 and S6 correspond to
the Channel Country (near Birdsville), northern NSW floodplains (near Moree) and western Darling Downs. All these
regions are open grasslands or farms with little perennial
vegetation. During droughts, vegetation cover in these
regions diminishes faster than regions with perennial bush
or tree cover. The Channel Country is a well-known source
area while the Moree and Darling Downs are known to be
dusty in dry years. The fine sediments of loams and selfmulching clays in these three areas provide favorable
conditions for dust emission. In NT, to the north of S4, there
existed a large area of relatively weak dust emission (S7).
[29] The simulated net dust emission, i.e., FN, averaged
over a 24 hour period for 22 and 23 October 2002 are
shown in Figures 6c and 6d, respectively. As expected, net
dust emission is significantly smaller than dust emission.
The magnitude of maximum FN is about half that of FS. On
22 October (Figure 6c), regions of significant net dust
deposition (i.e., negative FN) were confined, but to the
north of S1, a region of substantial dust deposition (D1)
can be found. On 23 October (Figure 6d), net dust emission
was much weaker. Surrounding the dust sources to the
north, northeast and northwest were regions of significant
dust deposition, D1, D2 and D3. The latter almost covered
the entire eastern coastal area of Australia. Dust from S8
was deposited in the Indian Ocean.
7 of 13
SHAO ET AL.: DUST EVENT IN AUSTRALIA
D08207
D08207
Figure 5. Streamlines and dust concentration at (a) surface, (b) 950 hPa, (c) 850 hPa, and (d) 600 hPa.
For visualization, dust concentrations on 850 and 600 hPa are multiplied by 3 and 20, respectively.
[30] The total net dust emission IN is computed by
integration of FN over the domain surface,
Z
IN ¼ IS ID ¼
Z
FS ds FD ds
S
Likewise, the total dust load M can be obtained by
integration of dust load over the domain. Figure 7a shows
the time series of IS and ID (in Mt h1) and Figure 7b shows
the time series of IN (in Mt h1) and M (in Mt) for the 4 day
time period of 22 –25 October 2002. IS and ID were of
similar order of magnitude. The maximum of IS and ID
occurred at around 0600 –0700 UTC 23 October 2002. This
result is expected because at daytime the atmosphere is
unstable and the momentum transfer from the upper layers
of the atmosphere to the surface is stronger. At the same
time, soil moisture in the very topsoil layers is lower due to
stronger evaporation and solar radiation. Positive net dust
emission occurred over a period of 32 hours between 0100
UTC 22 October 2002 and 0800 UTC 23 October 2002
(Figure 7b). After 0800 UTC 23 October 2002, net dust
emission was negative, implying that dust deposition
8 of 13
D08207
SHAO ET AL.: DUST EVENT IN AUSTRALIA
D08207
Figure 6. (a) Averaged dust emission rate in mg m2 s1 over the period of 0000 – 2400 UTC
22 October 2002. The 294 K temperature contours at 0600 UTC (thin solid line), 1200 UTC (thick solid
line), 1800 UTC (dotted line), and 2400 UTC (dashed line) of 22 October 2002 are drawn to show the
approximate location of the cold front. Main dust sources are marked with S1, S2, etc. Wind vectors
with speed exceeding 6 m s1 are drawn. (b) Same as Figure 6a, but for 23 October 2002. (c) Same as
Figure 6a, but for net dust emission. (d) Same as Figure 6b, but for net dust emission.
exceeded dust emission. During the dust episode, peak net
dust emission occurred in the (local) daytime of 22 and
23 October 2002. In the night, net dust emission was
substantially weaker. The maximum net dust emission from
Australia was found to be around 0.4 Mt h1 which
occurred at around 0600 – 0800 UTC 22 October and 0600
UTC 23 October 2002. At 0600 UTC 23 October, the
intensity of dust emission was not strong, but the area of
dust emission was large. Because of the positive net dust
emission, the total dust load was increasing until the
maximum was reached at 0800 UTC 23 October. From that
time, M decreased. The magnitude of the maximum total
dust load is comparable with the estimate of McTainsh et al.
[2005]. Their estimated total dust load falls between 3.35
and 4.85 Mt. In their study, dust in WA was not included.
We have also calculated the total net dust emission and total
dust load with WA excluded (Figure 7a). Then, the modelestimated maximum total dust load is 5 Mt. This estimate is
remarkably close to that of McTainsh et al. [2005], although
the methods used in the two studies are completely
different.
[31] Deposition dominated the later stages of the dust
episode. Strong net deposition occurred between 1100 and
2300 UTC 23 October 2002 (2000 LT 23 October and
0800 LT 24 October, night), with maximum exceeding
0.2 Mt h1. This was accompanied by a rapid decrease in dust
load. In the 2 days to follow, net deposition continued although
at a lower rate and dust load continued to decline. Dust
load was below 2 Mt by 0000 UTC 26 October 2002, much
of the dust was floating in the atmosphere over the ocean.
[32] The numerical results suggest that the total net
dust emission during the 0100 UTC to 2200 –0800 UTC
23 October 2002 Australian dust event is about 6 Mt, of
which 1 Mt from WA and 5 Mt from the rest of Australia.
These estimates are consistent with the maximum dust load
which occurred at 0800 UTC 23 October. The 6 Mt dust is
9 of 13
D08207
SHAO ET AL.: DUST EVENT IN AUSTRALIA
Figure 7. (a) Time series of total (domain surface) dust
emission and deposition, both in Mt h1, starting from 0000
UTC 22 October 2002. (b) Time series of total net dust
emission in Mt h1 and total (domain volume) dust load
in Mt.
R
the residual of 66.7 Mt dust entrainment
(i.e., ISdt) and of
R
60.7 Mt dust deposition (i.e., IDdt).
[33] A portion of dust is deposited in the ocean. Dust
emitted from the eastern part of Australia, including SA,
Vic, NSW, Qld, and NT was deposited mainly in the Pacific
to the east of Australia and the Gulf of Carpentaria. Dust
emitted from WA was deposited in the Indian Ocean to the
northwest of Australia. Figure 8 shows the distribution of
dust load and deposition averaged over 0000 –2400 UTC
23 October 2002. On 22 October, dust load and deposition
over the oceans surrounding Australia were very low (not
shown). On 23 October, a significant dust load over the
Pacific Ocean can be seen along the entire northeastern
coast of Australia, as well as over the Indian Ocean. The
highest dust load exceeded 1000 mg m2 off the coast of
Brisbane, the NW cape and the Gulf of Carpentaria
(Figure 8a). High dust depositions also occurred in these
three areas, as seen in Figure 8b.
[34] The time series of total dust load over the ocean and
total dust deposition in the ocean are plotted in Figure 9.
Over the ocean, dust load was increasing in the first 2 days
of the dust episode, the occurrence of maximum dust load
occurred was around 0000 UTC 24 October, about 12 hours
later than that of maximum dust load over land. This delay
in time is expected, as dust was transported from over land
to over ocean. The maximum dust deposition in the ocean
also occurred at 0000 UTC 24 October, reaching almost
0.04 Mt h1. In the subsequent days, deposition in the
ocean continued.
[35] The dust balance over the period of 22– 25 October
2002 is summarized in Table 2. Note that the dust loads are
the predicted values for 2400 UTC 25 October 2002, while
dust emission and depositions are the totals over the 4 day
period. About 97.8% [(92.87 + 0.8)/95.8] of the dust
D08207
emission was deposited to the continent. Over the 4 day
period, 1.12 Mt dust was deposited in ocean. By 2400 UTC
25, 1.01 Mt dust remained in suspension over the ocean,
which we assume will eventually be deposited in the ocean.
Thus the total amount of dust lost to the ocean during the
dust episode is 2.13 Mt, which is 2.2% of total dust
emission.
[36] In northeast Asia, severe dust storms frequently
occur in spring. A number of numerical simulations on
northeast Asian dust storms have been carried out in recent
years [e.g., Shao et al., 2003; Han et al., 2004]. In terms of
the orders of magnitudes of dust load, emission and deposition, the estimates of the two studies are consistent. We
use the results of Shao et al. [2003] to compare the Asian
and Australian dust storms, because the dust model used in
that study is almost identical as the one used here. Between
1 March and 30 April 2002, seven successive dust events
occurred in northeast Asia. Plotted in Figure 10a are the
total dust emission, deposition and load for the 22 –
25 October 2002 Australian dust event and in Figure 10b
those for the 28– 30 March and 5 – 9 April 2002 Asian dust
events. The 5 – 9 April 2002 Asian dust event was the
Figure 8. (a) Average dust load over the ocean (mg m2).
(b) Average dust deposition in the ocean (mg m2 s1). The
average is over 0000 – 2400 UTC 23 October 2002.
10 of 13
D08207
SHAO ET AL.: DUST EVENT IN AUSTRALIA
Figure 9. Time series of dust load (Mt) and deposition
over the ocean (Mt h1) for the period of 0000 UTC
22 October to 2400 UTC 25 October 2002.
largest reported in that year and among the largest in Asia in
recent years. The 28– 30 March 2002 Asian dust event was
a relatively weak event. The Australian dust storm is about
1.5 times that of the 28 –30 March Asian dust storm: The
former has a total dust load of 5.8 Mt (net dust emission
5.8 Mt), while the latter 4 Mt (net dust emission 3.5 Mt).
The Australian dust storm also has larger dust emission and
deposition. However, it is about 1/2 to 1/3 of the 5 – 9 April
Asian dust storm which has a total dust load of 16 Mt (net
dust emission 14 Mt). Although the magnitude of dust
emission and deposition are similar for both events, the
Australian event lasted about 32 hours, much shorter than
the 72 hours for the Asian event.
6. Discussion and Conclusions
[37] In comparison to the major dust sources in the
Northern Hemisphere, such as the Sahara, the Middle East,
and northeast Asia, Australia is a relatively minor contributor to the global dust budget [Tanaka and Chiba, 2006].
However, severe dust storms do occur in Australia, especially during drought years. According to McTainsh et al.
[2005], the 22– 23 October 2002 dust storm is the most
severe dust storm in Australia in at least the past 40 years. In
this study, we have simulated this dust event using an
integrated dust model.
[38] There is a lack of quantitative data for the verification of the model results. To our knowledge, only a few
high-volume sampler measurements of dust concentration
exist for this dust event, with only one in the dust source
area. We have compared the simulated dust concentration
with the TSP and PM10 data for the locations where
measurements are available. The model results are in
reasonable agreement with the observations. We have also
compared the simulated dust concentration with the surface
visibility data and the satellite images, and it can be
concluded that the model well reproduced the spatial pattern
and temporal evolution of the dust field.
[39] The climatic background and synoptic features of
this dust event are now clear. Prior to the dust event, the
eastern part of Australia experienced a severe drought and
periods of high temperature anomalies. As a consequence,
D08207
the land surface was dry and lacked vegetation cover. The
synoptic system that generated this dust event was a fast
moving cold front accompanied by a low level trough.
Strong wind (with velocity exceeding 12 m s1) occurred
behind the cold front, which entrained large quantities of
dust particles into the atmosphere. The dust emitted from
the regions behind the cold front converged to the frontal
area and then diverged southeastward and northwestward
along the front. As a consequence, high dust concentration
was found near the front. In the upper levels, dust particles
were mainly found in the trough, forming a narrow dust
band stretching several thousand kilometers from the southeast coast to the desert in northwest Australia. This dust
band was clearly visible from the satellite images.
[40] The simulations suggest that on 22 October 2002, the
main dust sources were the desert areas in northern SA, the
grazing lands in western NSW and the farming areas in NSW
and Vic. Dust emission also occurred in WA. The desert
areas appeared to be the strongest source. The maximum
net dust emission rate was around 500 mg m2 s1, which
is not excessively large in comparison to the dust emission
rate in other areas of the world, for example the Gobi
Desert in northeast Asia. On 23 October, dust emission was
much wider spread but weaker in intensity. In addition to
the above mentioned dust sources on 22 October, dust
sources emerged in NSW, Qld and NT. The NSW dust
sources were farming areas. The net dust emission in
Australia was positive for the period of 0000 UTC 22 to
1100 UTC 23 October 2002. After 1100 UTC, net dust
emission was negative. It is found that the daytime dust
emission was stronger than the nighttime dust emission.
This is expected, because the near-surface wind is generally
stronger and the soil is drier during the day than in the
night. The maximum total dust load over the simulation
domain has been found to be around 5 Mt. This estimate is
remarkably close to the 4.85 Mt estimate of McTainsh et al.
[2005], although totally different methods are used in the
two studies. The total amount of dust eroded from the
Australian continent during this dust event is around
95.8 Mt, of which 93.67 Mt is deposited back to the
continent and 2.13 Mt is deposited in the ocean.
[41] It is likely that the model over estimated the total
dust emission from the Australian continent, because the
predicted dust emission in some areas behind the cold front
lasted longer than expected. For example, dust emission
from the desert areas in SA lasted for much of the 2 day
Table 2. Dust Loads, as Predicted for 2400 UTC 25 October
2002, and Total Dust Emission and Depositions Over the Period of
0000 UTC 22 October to 2400 UTC 25 October 2002a
Value
Total load (max)
Land load (max)
Ocean load (max)
Emission
Total deposition
Land deposition
Ocean deposition
1.81 (5.79)
0.80 (5.40)
1.01 (1.31)
95.80
93.99
92.87
1.12
a
All quantities are in Mt. Total Load represents the load over the
simulation domain, land load represents the load over land, etc. The
maximum dust loads that occurred in the period of 0000 UTC 22 October to
2400 UTC 25 October 2002 are given in parentheses.
11 of 13
SHAO ET AL.: DUST EVENT IN AUSTRALIA
D08207
D08207
Figure 10. Comparison of dust load, emission and deposition of (a) the Australian dust storm with
those of (b) two northeast Asian dust storms. In Figure 10b, BST (Beijing Standard Time) is used.
period. This is because the dust emission model assumes
that the amount of dust for emission is unlimited. However,
in nature the amount of erodible dust is limited and once it
is depleted from the surface, dust emission will cease. In
other words, the aeolian process in reality is source limited.
This is a rather difficult problem for dust models to handle,
because the amount of dust available for dust emission
depends on the history of wind erosion and the weathering
process. In the current version of the dust scheme, the
source limiting effect is not included. Despite this limitation, the model produces satisfactory results and elucidates
where the dust source areas are, provides information on
dust emission, load and deposition rates. Finally, the modeling approach allows us to compare this dust storm with
others reported elsewhere in the world.
[42] Acknowledgments. We thank Andrew Chan of Griffith University for providing the PM10 data for model validation. NDVI data are kindly
provided by Jenny Lovell, Peter Briggs, and Vanessa Chewings of CSIRO
Earth Observation Centre.
References
Australian Surveying and Land Information Group (AUSLIG) (1990),
Atlas of Australian Resources, vol. 6, Vegetation, Aust. Surv. and Land
Inf. Group, Dep. of Admin. Serv., Canberra.
Chan, Y. C., G. H. McTainsh, J. F. Leys, H. McGowan, and K. Tews
(2005), Influence of the 23 October 2002 dust storm on the air quality
of four Australian cities, Water, Air Soil Pollut., 164, 329 – 348.
Chepil, W., and N. Woodruff (1957), Sedimentary characteristics of dust
storms. II. Visibility and dust concentration, Am. J. Sci., 255, 104 –
114.
Csanady, G. T. (1963), Turbulent diffusion of heavy particles in the atmosphere, J. Atmos. Sci., 20, 201 – 208.
Fécan, F., B. Marticorena, and G. Bergametti (1999), Parameterization of
the increase of the aeolian erosion threshold wind friction velocity due to
soil moisture for arid and semi-arid areas, Ann. Geophys., 17, 149 – 157.
Gillette, D. A., and K. J. Hanson (1989), Spatial and temporal variability of
dust production caused by wind erosion in the United States, J. Geophys.
Res., 94, 2197 – 2206.
Ginoux, P., J. Prospero, O. Torres, and M. Chin (2004), Long-term simulation of global dust distribution with the GOCART model: Correlation
with the North Atlantic Oscillation, Environ. Model. Software, 113 – 128.
Han, Z., H. Ueda, K. Matsuda, R. Zhang, K. Arao, Y. Kanai, and
H. Hasome (2004), Model study on particle size segregation and deposition during Asian dust events in March 2002, J. Geophys. Res., 109,
D19205, doi:10.1029/2004JD004920.
Irannejad, P., and Y. Shao (1998), Description and validation of the Atmosphere-Land-Surface Interaction Scheme (ALSIS) with HAPEX and
Cabauw data, Global Planet. Change, 19, 87 – 114.
Joussaume, S. (1990), Three-dimensional simulation of the atmospheric
cycle of desert dust particles using a general circulation model, J. Geophys. Res., 95, 1909 – 1941.
Jung, E. J., and Y. Shao (2006), An intercomparison of four wet deposition
schemes for dust transport model, Global Planet. Changes, 52, 248 – 260.
Jung, E., Y. Shao, and T. Sakai (2005), A study on the effects of convective
transport on regional-scale Asian dust storms in 2002, J. Geophys. Res.,
110, D20201, doi:10.1029/2005JD005808.
Kain, J., and J. Fritsch (1993), Convective Parameterization for Meso-scale
Models: The Kain-Fritsch Scheme, Meteorol. Monogr., vol. 24, pp. 165 –
170, Am. Meteorol. Soc., Boston, Mass.
Knight, A. W., G. H. McTainsh, and R. W. Simpson (1995), Sediment loads
in an Australian dust storm: Implications for present and past dust processes, Catena, 24, 195 – 213.
Leslie, L. M., and G. R. Wightwick (1995), A new limited-area numerical
weather prediction model for operations and research: Formulation and
assessment, Mon. Weather Rev., 123, 1759 – 1775.
Liu, M., D. L. Westphal, S. Wang, A. Shimizu, N. Sugimoto, J. Zhou, and
Y. Chen (2003), A high-resolution numerical study of the Asian dust
storms of April 2001, J. Geophys. Res., 108(D23), 8653, doi:10.1029/
2002JD003178.
McTainsh, G. H., and J. R. Pitblado (1987), Dust storms and related phenomena measured from meteorological records in Australia, Earth Surf.
Processes Landforms, 12, 415 – 424.
McTainsh, G. H., A. W. Lynch, and R. Hales (1997), Particle-size analysis
of aeolian dusts, soils and sediments in very small quantities using a
Coulter Multisizer, Earth Surf. Processes Landforms, 22, 1207 – 1216.
McTainsh, G. H., Y. C. Chan, H. McGowan, J. F. Leys, and K. Tews
(2005), The 23rd October 2002 dust storm in eastern Australia: Characteristics and meteorological conditions, Atmos. Environ., 39, 1227 – 1236.
National Soil Resources Institute (1980), NATMAP, Atlas of Australian
Resources, vol. 3, Soils and Land Use, Div. of Nat. Map., Canberra.
Nickovic, S., S. Kallos, A. Papadopoulos, and O. Kakaliagou (2001),
A model for prediction of desert dust cycle in the atmosphere, J. Geophys.
Res., 106, 18,113 – 18,129.
12 of 13
D08207
SHAO ET AL.: DUST EVENT IN AUSTRALIA
Owen, R. P. (1964), Saltation of uniform grains in air, J. Fluid Mech., 20,
225 – 242.
Patterson, E. M., and D. A. Gillette (1977), Measurements of visibility vs.
mass concentration for airborne soil particles, Atmos. Environ., 11, 193 –
196.
Prospero, J. M., P. Ginoux, O. Torres, S. E. Nicholson, and T. E. Gill
(2002), Environmental characterization of global sources of atmospheric
soil dust identified with the NIMBUS 7 Total Ozone Mapping Spectrometer (TOMS) absorbing aerosol product, Rev. Geophys., 40(1), 1002,
doi:10.1029/2000RG000095.
Raupach, M. R. (1992), Drag and drag partition on rough surfaces, Boundary Layer Meteorol., 60, 375 – 395.
Raupach, M. R., D. A. Gillette, and J. F. Leys (1993), The effect of roughness elements on wind erosion thresholds, J. Geophys. Res., 98, 3023 –
3029.
Raupach, M. R., G. H. McTainsh, and J. F. Leys (1994), Estimates of dust
mass in recent major Australian dust storms, Aust. J. Soil Water Conserv.,
7, 22 – 24.
Raupach, M. R., P. R. Briggs, N. Ahmad, and V. E. Edge (2001), Endosulfan transport II: Modeling airborne dispersal and deposition by spray
and vapor, J. Environ. Qual., 30, 729 – 740.
Seino, N., H. Sasaki, A. Yamamoto, M. Mikami, H. Zhou, and F. Zeng
(2005), Numerical simulation of meso-scale circulation in the Tarim
Basin associated with dust events, J. Meteorol. Soc. Jpn., Ser. A, 83,
205 – 218.
Shao, Y. (2001), A model for mineral dust emission, J. Geophys. Res., 106,
20,239 – 20,254.
Shao, Y. (2004), Simplification of a dust emission scheme and comparison
with data, J. Geophys. Res., 109, D10202, doi:10.1029/2003JD004372.
Shao, Y., and L. M. Leslie (1997), Wind erosion prediction over the Australian continent, J. Geophys. Res., 102, 30,091 – 30,105.
Shao, Y., and H. Lu (2000), A simple expression for wind erosion threshold
friction velocity, J. Geophys. Res., 105, 22,437 – 22,443.
Shao, Y., L. M. Leslie, R. K. Munro, P. Irannejad, W. F. Lyons, R. Morison,
D. Short, and M. S. Wood (1997), Soil moisture prediction over the
Australian Continent, Meteorol. Atmos. Phys., 63, 195 – 215.
D08207
Shao, Y., et al. (2003), Northeast Asian dust storms: Real-time numerical
prediction and validation, J. Geophys. Res., 108(D22), 4691,
doi:10.1029/2003JD003667.
Tanaka, T. Y., and M. Chiba (2006), A numerical study of the contributions
of dust source regions to the global dust budgets, Global Planet.
Changes, 52, 88 – 104.
Tews, E. K. (1996), Wind erosion rates from meteorological records in
eastern Australia 1960-92, B.Sc. thesis, Griffith Univ., Nathan, Queensland, Australia.
Uno, I., H. Amano, S. Emori, K. Kinoshita, I. Matsui, and N. Sugimoto
(2001), Trans-Pacific yellow sand transport observed in April 1998:
A numerical simulation, J. Geophys. Res., 106, 18 331 – 18 344.
Uno, I., K. Harada, S. Satake, Y. Hara, and Z. Wang (2005), Meteorological
characteristics and dust distribution of the Tarim Basin simulated by the
nesting RAMS/CFORS dust model, J. Meteorol. Soc. Jpn., Ser. A, 83,
219 – 239.
Wang, Z., H. Ueda, and M. Huang (2000), A deflation module for use in
modeling long-range transport of yellow sand over east Asia, J. Geophys.
Res., 105, 26,947 – 26,957.
Westphal, D. L., O. B. Toon, and T. N. Carson (1988), A case study of
mobilization and transport of Saharan dust, J. Atmos. Sci., 45, 2145 –
2175.
Zender, C. S., H. Bian, and D. Newman (2003), Mineral Dust Entrainment
and Deposition (DEAD) model: Description and 1990s dust climatology,
J. Geophys. Res., 108(D14), 4416, doi:10.1029/2002JD002775.
J. F. Leys, Department of Natural Resources, Gunnedah, NSW 2380,
Australia.
G. H. McTainsh and K. Tews, Faculty of Environmental Sciences,
Griffith University, Brisbane, Queensland 4111, Australia.
Y. Shao, Institute of Geophysics and Meteorology, University of
Cologne, Zülpicher Str. 49, D-50674 Cologne, Germany. (yshao@meteo.
uni-koeln.de)
13 of 13