You are on page 1of 14

International Journal of Environmental Research (2019) 13:161–174

https://doi.org/10.1007/s41742-018-0164-1

RESEARCH PAPER

Numerical Simulation of a Severe Dust Storm over Ahvaz Using


the HYSPLIT Model
Reza Khalidy1 · Hesam Salmabadi1 · Mohsen Saeedi1 

Received: 14 June 2018 / Revised: 19 November 2018 / Accepted: 15 December 2018 / Published online: 1 January 2019
© University of Tehran 2019

Abstract
In this paper, numerical simulation of a severe dust storm over Ahvaz on 16–18 June 2016 was performed using the HYSPLIT.
To that end, the HYSPLIT model has been configured for different model setups to determine the most desirable suite
of parameters to represent the measured concentrations better. Among several model parameters, meteorological dataset
(GDAS1.0 and GDAS0.5), friction velocity (P10F) and particle numbers (NumPar) were detected as the most influential
parameters on the simulation results. Accordingly, with P10F = 0.6, NumPar = 5000, and GDAS1.0 configurations, the
model has reproduced the most precise results, and the correlation coefficient between the measured and the simulated ­PM10
concentration was as high as 0.9. Since the threshold friction velocity influences the results significantly, it is suggested to
calculate the P10F coefficient for each distinct dust storm meticulously. Once the best model parameters have been obtained,
the HYSPLIT has been run for 33 possible sources separately to estimate the contribution from each sub-region to the levels
of the measured ­PM10 concentration of Ahvaz during 16–18 June 2016. According to the results, central Iraq, northern Syria,
western Iraq, and Al-Hawzieh/Al-Azim wetland account for 71, 19, 6, and 4%, respectively.

Article Highlights
• A severe dust storm in 2016 over Ahvaz city was simulated quantitatively.
• Validation of simulation results was accomplished using ground-based data and satellite observations.
• Determination of the Middle East dust origins contribution in the dust storm was performed.
• Western and central parts of Iraq were detected as the main sources of the dust storm.

Keywords  HYSPLIT · Source apportionment · Middle East dust storms · Ahvaz · Numerical simulation · Satellite
observation

Introduction et al. 2013; Middleton 2017; Rashki et al. 2017) and are det-
rimental to human health and cause a wide of variety of dis-
Numerous dust particles from world’s arid regions are trans- orders including cardiovascular diseases, asthma, eye infec-
ported over long distances during sand and dust storms tions and meningitis (Goudie and Middleton 2006; Khaefi
(SDS) annually (Lee et al. 2010). It has been estimated that et al. 2017; Maleki et al. 2016; Shahsavani et al. 2012).
1000–3000-million tons of dust particles are transported The Middle East has been identified as one of the most
per year globally (Choobari et al. 2014). SDS affect Earth notable dust origins in the world (Ginoux et al. 2012; Pros-
system through alteration in radiation and energy budget, pero et al. 2002). The main factors forming dust storms
biochemical cycles of oceans and climate processes (Huang in the Middle East region are the “Shamal winds” during
summer and also pre-frontal and post-frontal winds during
* Mohsen Saeedi winter (Kaskaoutis et al. 2015). Ahvaz, the capital city of
msaeedi@iust.ac.ir Khuzestan province and located in the southwest of Iran, has
1
been declared as the most polluted city in the world by an
Environment Research Laboratory, School of Civil
average ­PM10 concentration of 372 µg/m3 in 2013 (Maleki
Engineering, Iran University of Science and Technology,
Narmak, Tehran, Iran et al. 2016; Broomandi et al. 2017). Due to the proximity

13
Vol.:(0123456789)

162 International Journal of Environmental Research (2019) 13:161–174

to Saudi Arabia, Kuwait, and Iraq desert lands, the city has outbreaks. Givehchi et al. (2013) simulated four dust epi-
been exposed to severe dust storms in recent years (Shahsa- sodes using the HYSPLIT which occurred in 2009 and 2010
vani et al. 2012). over Tehran. They determined central Iraq and northern and
One of the most severe dust storms Ahvaz has experi- eastern Syria (Tigris and Euphrates zone) as the chief Mid-
enced in recent years had occurred in June 2016. The dust dle East dust sources impacting Tehran P ­ M10 concentration
storm that lingered in the city for 44 h had the average and level. The same methodology was employed for the same
the maximum ­PM10 concentration 987 and 2322 µg/m3, dust episodes to determine the contribution of Middle East
respectively (DOE 2016), and the horizontal visibility less- dust sources on the P­ M10 level of other cities of Iran (Sotou-
ened to less than 500 m during the dust period. deheian et al. 2016).
Several studies have considered different meteorological In this paper, first, the HYSPLIT model was employed
parameters and conditions in dust transport and dispersion. for numerical simulation of the mentioned dust storm with
Most of these studies are confined to trajectory calculations a full description of model parameters. Furthermore, source
(McGowan and Clark 2008; Notaro et  al. 2013; Rashki apportionment of the dust storm was quantitatively deter-
et al. 2017). To visualize a precise spatial and temporal dis- mined. Finally, the model results were further validated with
tribution of dust concentration in the atmosphere, a com- ground-based observations and satellite images.
prehensive simulation considering advection, dispersion,
and deposition is required. The HYSPLIT, developed by
National Oceanic and Atmospheric Administration (NOAA), Methodology
is a Lagrangian model which has been widely used for air
parcel dispersion, transportation, and deposition simula- Dust Episode
tion (Hemond and Fechner 2014; Stein et al. 2015). The
model calculations are hybrid between Lagrangian and PM10 concentration data of Ahvaz urban zone was acquired
Eulerian approaches (Draxler et al. 2001; Stein et al. 2015). from the Department of Environment (DOE) of Khuzestan.
Advection and diffusion calculations are performed in a By analyzing the observed P ­ M10 data from Air Quality
Lagrangian framework, while pollutant air concentration Monitoring (AQM) stations of Ahvaz (called “measured
calculations are computed on the fixed grid using Eulerian data” from now on), June 16–18, 2016 was chosen as the
methodology (Draxler et al. 2001; Stein et al. 2015; Salma- dusty period for modeling. ­PM10 concentration in selected
badi and Saeedi 2018). Many studies have been carried out period recorded in AQM stations of Ahvaz has been shown
using the HYSPLIT for backward trajectory analysis (Cus- in Fig. 1. Owing to lack of AQM station out of Ahvaz urban
pilici et al. 2017; Hernández-Ceballos et al. 2014; Lee et al. area, background ­PM10 concentration of Ahvaz had to be
2010; Notaro et al. 2013; Xin et al. 2016), forward trajectory calculated to separate P­ M10 concentration related to the dust
analysis (Ge et al. 2016; McGowan and Clark 2008; Rashki storm from P ­ M10 concentration due to anthropogenic activi-
et al. 2017; Salmabadi and Saeedi 2018), and aerosol con- ties. To this end, non-dusty days were identified based on the
centration calculations (Ashrafi et al. 2014; Draxler et al. World Meteorological Organization (WMO) hourly report
2001; Escudero et al. 2006; Givehchi et al. 2013). In this over Ahvaz (https​://www.ncei.noaa.gov/), and the back-
respect, Escudero et al. (2006) identified different Saharan ground concentration was determined by averaging normal
desert segments contribution on the ­PM10 level elevation days concentration (non-dusty days) of June 2016. Figure 1
in Spain cities. The results indicated Mauritania, Western shows the fluctuation of ­PM10 concentration over Ahvaz in
Sahara, and Mali are the main origins of the studied dust June 2016 and the dusty days are shown with yellow dots.

Fig. 1  Ahvaz daily P
­ M10 1200
concentration in June 2016. The
study period has been defined 1000
PM10 Concentra on (ug/m3)

with a red-dashed rectangle.


Yellow dots represent the dusty 800
days based on WMO report.
The gap in P
­ M10 concentration 600
data is owing to monitoring
defect at AQM stations 400

200

0
31 May 2016 05 June 2016 10 June 2016 15 June 2016 20 June 2016 25 June 2016 30 June 2016
Date

13
International Journal of Environmental Research (2019) 13:161–174 163

According to the calculations, 139 µg/m3 was determined Model Parameters


as the background concentration which was subtracted from
recorded concentration in AQM station to assess the P ­ M10 Meteorological Input
level due to the dust storm. It should be mentioned that
some errors might occur in this approach, as meteorologi- Previous studies have shown that the HYSPLIT calcu-
cal parameters and the anthropogenic activities are likely to lations are sensitive to input meteorological parameters
be varied on different days. (Gebhart et al. 2005; Hernández-Ceballos et al. 2014; Su
et al. 2015). Several meteorological datasets are applica-
ble for the HYSPLIT model simulations including GDAS,
Model Description NARR, NCAR REANALYSES, EDAS, and NAM (Stein
et al. 2015). Among the mentioned meteorological data-
There are two different algorithms (primary and revised sets, the spatial resolution of GDAS and REANALYSES
algorithms) for dust simulation embedded in the HYSPLIT is worldwide and could be used in this study. Since the
model. The primary algorithm was developed for Iraq, part horizontal resolution of REANALYSES data is overly
of Syria, Kuwait, Oman, United Arab Emirates, and Saudi coarse (2.5°), the dataset is not appropriate for dust storm
Arabia in the Middle East by Draxler et al. (2001), and it was modeling. The Global Data Assimilation System (GDAS),
examined over Kuwait, Iraq, and Saudi Arabia (Draxler et al. known as one of the operating systems developed by the
2001). The ­PM10 dust flux in this algorithm is calculated as National Centers for Environmental Prediction (NCEP),
follows: is an atmospheric model providing 3-hourly meteorologi-
𝜌 ( cal data. There are two forms of GDAS datasets avail-
Q = K U∗ U∗2 − U∗t
2
)
g
A. (1) able for the HYSPLIT calculations, namely GDAS1.0
and GDAS0.5. GDAS1.0 has been archiving data in a
The soil texture coefficient K and the friction threshold weekly format from December 2004 with 1° horizontal
velocity U∗t were calculated for southwest Asia with regard resolution (1° = 100 km). It provides data at 23 different
to the area’s soil texture (Draxler et al. 2001). Since U∗t and vertical levels within a vertical interval starting from sea
K were calculated only for the mentioned regions, another level (1000 hPa) to height about 26 km (20 hPa) (ARL
equation (Eq. 2) is also embedded in the HYSPLIT model 2018). GDAS0.5 has been archiving data in daily format
which is applicable to other areas that are defined in the from November 2007 with 0.5° horizontal resolution. It
HYSPLIT’s land use map as desert land. In this mode, the provides data at 55 vertical levels from 1000 to 13 hPa
flux injections are defined by: (~ 31 km) (ARL 2018). Despite using the same observed
data for rendering GDAS meteorological data, discrepan-
Q = 0.01U∗4 A. (2)
cies (different horizontal and vertical resolution) between
Another algorithm (revised algorithm) was also devel- GDAS1.0 and GDAS0.5 could lead to varied results (Su
oped by Draxler et al. (2010) using MODIS AOD and mete- et al. 2015). Therefore, both forms of GDAS meteorologi-
orological data. The revised algorithm, initially, has been cal datasets were included in this study.
developed for North America and then for the dust origins
all over the world. The dust emissions in the revised algo-
rithm are computed as follows: Backward Trajectory
(3)
( )
Q = KA U∗ − U∗t ,
Considering wind as the primary avenue for a dust storm
where K is the dust density, A the emission area, U∗t the formation, to determine the primary dust storm origins,
threshold friction velocity and U∗ the friction velocity. This wind rose of Ahvaz was plotted during the dust episode,
equation is not dependent on soil’s characteristics. In the and backward trajectory calculation was accomplished
revised algorithm, Draxler et  al. (2010) have identified in different fractions (0.25, 0.5, 0.75, and 1) of planetary
global dust origins and their characteristics (K and U∗t ) with boundary layer height (PBLH) starting from Ahvaz in 3 h
the horizontal resolution of 0.25° for different months of interval and using GDAS1.0 and GDAS0.5 meteorological
the year by using AOD measurements of MODIS satellite datasets. To ensure that all the aerosols had been traced,
and meteorological data. All the values of K and U∗t are the simulation period was selected from 16 June at 00:00
accessible online through the following link: https​://www. to 18 June at 21:00. In the long run, based on the backward
ready.​ noaa.gov/docume​ nts/Tutori​ al/dust/dust_global​ .zip. To airflows calculations and wind rose plot, the potential dust
perform the simulation of this study, the revised algorithm origins of selected dust storm were determined.
was chosen as it has been developed recently and engenders
more precise results (Givehchi et al. 2013).

13

164 International Journal of Environmental Research (2019) 13:161–174

Source Assignment PM10 concentration simulated by the HYSPLIT model was


determined, and the proposed threshold friction velocities
The trajectory analysis is only inclusive of particles path- were all multiplied by this factor.
way and is not applicable for source apportionment solitary.
Hence, further must be executed to determine the contri- Wet and Dry Deposition
bution of each of dust origins to PM10 level at a specific
receptor. As mentioned in the introduction, several dust As the dust storm occurred in a dry season, the wet deposi-
sources have been identified in the Middle East (Cao et al. tion was excluded. The dry deposition rate was determined
2015a; Ginoux et al. 2012; Prospero et al. 2002). Draxler considering 3 µm as the region particle diameter average
et al. (2010) provided global dust sources map for differ- size, the density of 2.5 g/cm3 and a spherical shape (Draxler
ent months based on AOD value obtained from the MODIS et al. 2001).
instrument. The Middle East dust sources map in June has
been plotted in Fig. 2. As evident in Fig. 2, there are numer- Vertical Simulation Height
ous dust origins in the study area.
PM10 average concentration related to the dust event was
Threshold Friction Velocity Calibration calculated on the spatial interval of the Earth’s surface (0
m) and an average of the PBLH during the modeling period
In the HYSPLIT model, threshold friction velocity has in Ahvaz (Givehchi et al. 2013). To calculate the PBLH,
been calculated according to May 2008–April 2009 (Drax- GDAS0.5, GDAS1.0, and the HYSPLIT software were used.
ler et al. 2010) climatological data and may fluctuate over The PBLH was extracted every 3 h between 16 June 00:00
time. Consequently, the available friction velocities may and 18 June 21:00 from meteorological data. Due to mete-
not represent the precise values for 2016 when the studied orological datasets’ resolution, the time interval of 3 h was
dust storm occurred. The P10F coefficient is included in selected (Tables 1, 2).
the model to modify friction velocity by multiplying with
measured friction velocity. Regarding result sensibility to Other Parameters
friction threshold velocity, acquiring the most appropriate
value for P10F is a matter of significance. To that end, differ- Other parameters influencing simulation results are model
ent P10F coefficients were examined in the simulation. The type, number of particles per cycle (NumPar), and maxi-
multiplication factor leading to the most desirable results of mum number of particles. There are three model types

Fig. 2  Middle Eastern dust


sources map in June provided
by Draxler et al. (2010)

13
International Journal of Environmental Research (2019) 13:161–174 165

Table 1  Fluctuation of the PBLH over Ahvaz during 16, 17, and 18 June 2016 using GDAS1.0 data
Hour date 00.00 (m) 03.00 (m) 06.00 (m) 09.00 (m) 12.00 (m) 15.00 (m) 18.00 (m) 21.00 (m) Daily avg. (m)

16 June 150 39 65 918 2478 3229 3738 683 1412.5


17 June 530 33 64 1878 2700 2764 2746 56 1346.3
18 June 90 25 57 1883 2750 2755 2499 94 1269.1
Total avg. ≈ 1340 m

Table 2  Fluctuation of the PBLH over Ahvaz during 16, 17, and 18 June 2016, using GDAS0.5 data
Hour date 00.00 (m) 03.00 (m) 06.00 (m) 09.00 (m) 12.00 (m) 15.00 (m) 18.00 (m) 21.00 (m) Daily avg. (m)

16 June 156 39 33 854 2254 3165 3867 683 1381.3


17 June 434 33 32 2230 2796 2956 2810 56 1418.3
18 June 122 25 25 1819 3038 3011 2883 30 1369.1
Total avg. ≈ 1390 m

available for the dust simulation where only 3D particle type Earth Sciences (GES) Data and Information Services Center
is compatible with the revised algorithm. Since Escudero (DISC) (https​://giova​nni.sci.gsfc.nasa.gov/).
et al. (2006) investigated 500, 5000 and 50,000 number of
particles released per cycle for the primary algorithm, the
same values were set in this study and the best one was Results and Discussion
determined. The last parameter which should be taken into
account is the maximum number of particles in the simula- As the wind is the principal agent for generating dust storms,
tion period. The magnitude of this parameter should be large right apprehension of wind pathways could help to find pri-
enough to ensure no error will occur during the simulation. mary sources of a dust event. Initially, Ahvaz wind rose
Hence, the maximum number of particles was set to 5 mil- was plotted using WRPLOT software for the dusty period.
lion, and no error happened during the simulations. Regarding dominant wind direction arriving in Ahvaz, it
could be concluded that sources of the dust storm are located
Validation on the west of the city. The 48-h backward trajectory was
executed using the HYSPLIT software to figure out explicit
To determine the best suite of parameters for dust simulation wind pathways arriving in Ahvaz during the dusty period.
in the region and identify sources of dust episode, simulation Wind pathways that came in Ahvaz in the dusty period were
results were compared with the observed data using the cor- utterly traced using the HYSPLIT backward calculations and
relation coefficient. For further validation, aerosol index (AI) are shown in Fig. 4. As it is conspicuous, backward calcula-
maps of the ozone monitoring instrument (OMI) sensor and tions results are in accord with the plotted wind rose for the
Deep Blue aerosol optical depth (AOD) products of Moder- study interval, and it is clear that wind pathways traversed
ate Resolution Imaging Spectroradiometer (MODIS) sensor the west regions of Iran, mainly Iraq and Syria. Hence, it
were employed. AI is a parameter which identifies the occur- could be concluded that the origins of the dust storm are
rence of aerosol with a threshold limit of 0.65 (Hsu et al. located in the mentioned regions.
1999). MODIS AOD representing aerosol abundance is a Based on backward calculations and the wind rose plot
significant remote sensing parameter (Namdari et al. 2018). (Figs. 3 and 4), the emissions from dust origins located in
MODIS AOD products have been utilized to identify dust the west and near Ahvaz (including Iraq, Syria, Jordan,
sources as well as track dust plumes in the myriad of studies Lebanon, most of Kuwait, southwest of Iran, and northern
(Beegum et al. 2018; Ginoux et al. 2012; Moridnejad et al. part of Saudi Arabia) were included in the simulations.
2015a; Notaro et al. 2013). Global level-3 Deep Blue AOD The extent of the study area and the potential dust sources
at 550 nm data and global level-3 AI data of June 15–19, are indicated in Fig. 5. 835 dust origin cells were selected
2016 were employed in this study to track the dust mass and out of approximately 2500 cells in the Middle East includ-
validate the source identification of the selected dust storm. ing Iraq, Syria, Kuwait, Saudi Arabia, Yemen, Jordan,
These data (OMI AI and MODIS AOD) load in daily tem- Lebanon, and southwest of Iran. As each cell’s area is
poral resolution and are accessible through Giovanni online 625 km2, the 835 cells represent an area of ~ 520,000 km2.
software developed and maintained by the NASA Goddard For further calculation, the study zone was gridded into

13

166 International Journal of Environmental Research (2019) 13:161–174

correlation coefficients of different scenarios outputs with


observed data are also shown in Fig. 7.
Regarding the results, no variation was observed when
NumPar parameter altered from 500 to 5000. Hence, the
results of scenarios corresponding to NumPar = 500 are not
shown in Figs. 6 and 7. As it is obvious in Fig. 6, altering the
NumPar does not affect the simulation results significantly.
The NumPar is less influential as the P10F value increases.
This is due to the fact that fewer particles are released in
higher friction velocities. Consequently, as the increment
of NumPar slightly reduces the simulation speed, it is sug-
gested that NumPar = 5000 is chosen. NumPar = 5000 was
also found to be the optimum value in previous studies
(Escudero et al. 2006; Givehchi et al. 2013; Sotoudeheian
et al. 2016).
As it is obvious in Fig. 6, the effect of meteorological
dataset alteration is significant. Based on the results, a huge
Fig. 3  Three-day wind rose of Ahvaz during 16, 17, and 18 June 2016 discrepancy is discernible between results obtained from
GDAS1.0 and GDAS0.5 datasets. Although the spatial
resolution of GDAS0.5 is better than that of GDAS1.0, in
33 sub-regions with an area of 200 × 200 km2, and once most scenarios it does not show a desirable performance.
the model parameters were set the contribution of each of The correlation results (Fig. 7) also delineate that in all sce-
these sub-regions was determined. narios, the correlation coefficient between results obtained
The performance of different meteorological datasets, from GDAS1.0 and the measured data is profoundly higher
P10F coefficient variation, and the number of particles than that of the results obtained from GDAS0.5. The maxi-
released per cycle were assessed by defining 54 different mum correlation coefficient in the scenarios with GDAS0.5
scenarios (Table 3), and the most desirable set of param- dataset is not higher than 0.79, while the maximum correla-
eters were selected using correlation coefficient. tion coefficient in the scenarios with GDAS1.0 dataset is as
The results of the all investigated scenarios have been high as 0.9. Generally, the use of GDAS1.0 is recommended
depicted in Fig. 6. The black, gray, and red lines indicate rather than GDAS0.5, due to time efficiency in simulations.
the results of scenario number 1–27, scenario number GDAS1.0 does not require vertical motion calculation as it
28–54, and the observed data in Ahvaz, respectively. The is a built-in parameter for the dataset, while this parameter
has to be calculated for GDAS0.5 data by the model from

Fig. 4  Two-day backward trajectories initiating from the 0.25, 0.5, 0.75, and 1 HPBL fraction of Ahvaz, during the dusty days using a GDAS1.0,
b GDAS0.5

13
International Journal of Environmental Research (2019) 13:161–174 167

Fig. 5  Study zone and the pos-


sible dust origins (black signs)
of June (Draxler et al. 2010).
The blue dot represents the
geographical location of Ahvaz.
Dust origins placed in the yel-
low rectangle are included in
the simulations

the vertical integration of the horizontal velocity divergence GDAS1.0 meteorological dataset and both NumPar = 5000
(Su et al. 2015). and 50,000. Although the maximum concentration occurring
The alteration of P10F will notably change the simulation on 17 June was slightly underestimated, P10F = 0.6 will give
results. As mentioned before, the threshold friction veloc- the most reliable results as the errors are minor at the end of
ity was calculated with regard to long-term synoptic data the simulation period compared with the other scenarios. As
(Draxler et al. 2010), is a function of different parameters Fig. 6 indicates, in P10F = 0.6 scenarios, GDAS0.5 highly
including soil moisture and vegetation cover (Givehchi underestimated ­PM10 concentration over the studied period.
et al. 2013; Sotoudeheian et al. 2016), and may fluctuate When P10F was reduced to 0.55, the model predicted the
over time. The significance of P10F coefficient variation was maximum ­PM10 concentration precisely (with GDAS1.0),
also investigated in previous studies (Givehchi et al. 2013; but the overprediction notably increased at the end of the
Sotoudeheian et al. 2016). For example, in a study conducted simulation period. In P10F = 0.5 scenarios, the maximum
by Givehchi et al. (2013), the P10F = 0.5 was assigned to ­PM10 concentration was slightly overpredicted, and for both
Iraq and Syria in 2009 and 2010. Consequently, in this GDAS1.0 and GDAS0.5, there was a huge overestimation
study different P10F values (0.4, 0.5, 0.55, 0.6, 0.65, 0.7, in the ­PM10 concentration level at the ending date of the
0.8, 0.9, and 1.0) were examined to find the most expedient simulation period. In P10F = 0.4, the results of all scenarios
value. Based on Fig. 6, less dust particles were generated (NumPar = 5000 and 50,000 of both GDAS1.0 and GDAS0.5
when P10F = 1.0 was set. It means that the threshold friction meteorological data) were overestimated. This overesti-
velocity has changed since it was first calculated. There- mation was intensified at the ending dates. As P10F was
fore, the P10F value has been reduced to find the best value. reduced from 0.6 to 0.4, the correlation coefficient for all
In all scenarios with P10F = 0.9, the simulation results are scenarios also decreased and the overestimation in the simu-
highly underestimated. The correlation coefficient for these lation results increased. The mentioned errors and the dis-
scenarios is not greater than 0.32. Although the correlation crepancy between the simulation results and measured data
coefficient of scenarios with P10F = 0.7 and 0.8 increased were due to the model’s profound sensitivity to threshold
greatly, the huge difference between the simulated P ­ M10 friction velocity, and the difficulty in precisely representing
concentration values and measured ­PM10 concentration this parameter in the model (Draxler et al. 2001).
remained. The correlation coefficient results denote that the As the best-fitted parameters were configured
most fitted results are obtained with P10F = 0.6 (R2 = 0.9) for (NumPar = 5000, P10F = 0.6, and GDAS1.0) in the

13

168 International Journal of Environmental Research (2019) 13:161–174

Table 3  54 sets of the Scenario # Meteorological data NumPar P10F Scenario # Meteorological data NumPar P10F
configuration used for model
calibration in this study 1 GDAS 1.0 500 0.4 28 GDAS 0.5 500 0.4
2 GDAS 1.0 500 0.5 29 GDAS 0.5 500 0.5
3 GDAS 1.0 500 0.55 30 GDAS 0.5 500 0.55
4 GDAS 1.0 500 0.6 31 GDAS 0.5 500 0.6
5 GDAS 1.0 500 0.65 32 GDAS 0.5 500 0.65
6 GDAS 1.0 500 0.7 33 GDAS 0.5 500 0.7
7 GDAS 1.0 500 0.8 34 GDAS 0.5 500 0.8
8 GDAS 1.0 500 0.9 35 GDAS 0.5 500 0.9
9 GDAS 1.0 500 1.0 36 GDAS 0.5 500 1.0
10 GDAS 1.0 5000 0.4 37 GDAS 0.5 5000 0.4
11 GDAS 1.0 5000 0.5 38 GDAS 0.5 5000 0.5
12 GDAS 1.0 5000 0.55 39 GDAS 0.5 5000 0.55
13 GDAS 1.0 5000 0.6 40 GDAS 0.5 5000 0.6
14 GDAS 1.0 5000 0.65 41 GDAS 0.5 5000 0.65
15 GDAS 1.0 5000 0.7 42 GDAS 0.5 5000 0.7
16 GDAS 1.0 5000 0.8 43 GDAS 0.5 5000 0.8
17 GDAS 1.0 5000 0.9 44 GDAS 0.5 5000 0.9
18 GDAS 1.0 5000 1.0 45 GDAS 0.5 5000 1.0
19 GDAS 1.0 50,000 0.4 46 GDAS 0.5 50,000 0.4
20 GDAS 1.0 50,000 0.5 47 GDAS 0.5 50,000 0.5
21 GDAS 1.0 50,000 0.55 48 GDAS 0.5 50,000 0.55
22 GDAS 1.0 50,000 0.6 49 GDAS 0.5 50,000 0.6
23 GDAS 1.0 50,000 0.65 50 GDAS 0.5 50,000 0.65
24 GDAS 1.0 50,000 0.7 51 GDAS 0.5 50,000 0.7
25 GDAS 1.0 50,000 0.8 52 GDAS 0.5 50,000 0.8
26 GDAS 1.0 50,000 0.9 53 GDAS 0.5 50,000 0.9
27 GDAS 1.0 50,000 1.0 54 GDAS 0.5 50,000 1.0

HYSPLIT, the model was separately run from determined 24). The other sources which are relatively minor include
sub-regions (Fig. 5) located in Iraq, Syria, Jordan, Cyprus, western Iraq (Sr14, 15), the northern part of Syria (Sr3, 8),
Lebanon, northern part of Saudi Arabia, parts of Kuwait, and Iraq– Khuzestan province boundary zone (Sr25). The
and southwest of Iran to calculate the precise contribution latter encompasses Al-Azim/Al-Hawizeh marshland which
of each region. The contribution of each area on Ahvaz has been desiccated due to dam construction, war, and cli-
­PM10 concentration has been determined in Fig. 8. Results mate change in recent years (Cao et al. 2015b).
delineated Iraq (Sr9, 14, 15, 16, 23, 24, 25 in Fig. 5), Syria With regard to the MODIS true color images, OMI AI
(Sr3, 8 in Fig. 5), and limited parts of Iran (Parts of Sr25 in maps, and MODIS AOD during 15–19 June, there was not
Fig. 5) as the sources of the dust storm with ~ 78, 20, and any considerable dust mass over the study area on June 15
2% contribution, respectively. The results also revealed that (Fig. 11a–c). According to simulation results (Fig. 9a), few
the central and western parts of Iraq (Sr16, 23, 24 in Fig. 5) minor dust plumes were discernible over Syria and Jor-
are the primary dust storm origins and contribute to ~ 65% dan and a significant dust mass over Kuwait, but there was
of the ­PM10 concentration level of Ahvaz during the dusty not any dust storm over central Iraq. True color images of
period. The mentioned regions were also identified as one of MODIS, OMI AI map, and MODIS AOD map of June 16
the main dust sources in the Middle East (Cao et al. 2015a; show that a significant dust plume emerged over Iraq on
Ginoux et al. 2012; Givehchi et al. 2013; Hamidi et al. 2017; this day (Fig. 11d–f). As evident in Fig. 9b, the simulation
Moridnejad et al. 2015a). All these sources are placed in result of June 16 indicates the emergence of this plume as
summer and winter Shamal wind pathways, flowing all year well. The plume expands and moves to the east and, finally,
long (Francis et al. 2017; Yu et al. 2016). Moridnejad et al. reaches Ahvaz (Fig. 11g–i). Based on the hourly P ­ M10 con-
(2015b) specified abandoned agricultural lands, desiccated centration data of Ahvaz (Fig. 10), the dust mass arrival
lakes and wetlands, and desert regions as the main contribu- time is 16 June at about 18:00 (local time) (Fig. 10). The
tors in dust storm generation in the areas above (Sr9, 16, 23, simulation results (Fig. 9c) and the satellite measurements

13
International Journal of Environmental Research (2019) 13:161–174 169

Fig. 6  PM10 level at Ahvaz 2500 P10F=0.4


GDAS1.0 NumPar=5000
P10F=0.5
GDAS1.0 NumPar=50000
derived from the HYSPLIT GDAS0.5 NumPar=5000
1600
simulations versus measured

PM10 Concentration (ug/m3)

PM10 Concentration (ug/m3)


2000 GDAS0.5 NumPar=50000 1400
(red line) ­PM10 of the dust out- Measured Data
1200
break that originated from the 1500
1000
regional dust which occurred on
16–18 June 2016. The simula- 1000
800

tions were carried out altering 600

the number of particles emitted 500


400

per cycle (5000 and 50,000) of 200


simulation, P10F (0.4, 0.5, 0.55, 0 0
0.6, 0.65, 0.7, 0.8, 0.9, 1.0), and 13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16 13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16

meteorological data (GDAS1.0, Date Date


GDAS0.5). The dashed black 1200 P10F=0.55 P10F=0.6
lines represent the results 1000

obtained with GDAS1.0 mete- 1000


PM10 Concentration (ug/m3)

PM10 Concentration (ug/m3)


orological data, and the gray 800
800

lines are associated with results


600
calculated with GDAS0.5 600

400
400

200 200

0 0
13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16 13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16
Date Title
P10F=0.65 1100 P10F=0.7
1000
900
PM10 Concentration (ug/m3)

PM10 Concentration (ug/m3)


800
700

600
500

400
300

200 100

0 13-Jun-16
-100 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16
13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16
Date
Date
P10F=0.8 P10F=0.9
1000 1000
PM10 Concentration 9ug/m3)

PM10 Concentration (ug/m3)

800 800

600 600

400 400

200 200

0 0
13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16 13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16
Date Date
P10F=1.0
1000
PM10 Concentration (ug/m3)

800

600

400

200

0
13-Jun-16 15-Jun-16 17-Jun-16 19-Jun-16 21-Jun-16
Date

13

170 International Journal of Environmental Research (2019) 13:161–174

Fig. 7  The correlation coef-


ficient between simulated and 1
measured ­PM10 concentration.
Black columns correspond 0.9
to GDAS1.0 meteorological
data, and red columns cor- 0.8
respond to GDAS0.5 meteoro-
logical data. The columns with 0.7
transparency are associated
with NumPar = 5000, and the 0.65
opaque columns are related to
NumPar = 50,000 0.6

0.55

0.5

0.4

-0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

NumPar=50000 NumPar=5000 NumPar=50000 NumPar=5000

(Fig. 11g–i) well depict the spatial coverage of the dust was determined that about 835 cells might contribute to dust
storm on June 17. As the MODIS AOD map and true color storm generation in the study period. 54 different scenarios
images of MODIS suggest (Fig. 11j–l), the dust mass had were examined to obtain the most precise results. To that
been lingering over Ahvaz till June 18 and as the plume end, the simulation results were compared with the observed
dispersed the P­ M10 concentration decreased over Ahvaz data. Correlation coefficients between simulation results
(Fig. 10). Finally, a significant proportion of dust particles and observed data indicate that the best results is obtained
vanished after 44 h at about 18:00 on June 18 (Fig. 10). with NumPar = 5000, GDAS1.0 meteorological data, and
P10F = 0.6 configuration. Although the best P10F is calculated
to be 0.6 in this study, it should be meticulously computed for
Conclusion each distinct dust storm as it is the most influential parameter
in the simulation process.
In this study, a severe dust storm in the Middle East that Results indicate that central Iraq accounts for ~ 71% of the
occurred on June 2016 was simulated using the HYSPLIT ­PM10 level of Ahvaz during 16–18 June. The minor sources
model, and the source apportionment of this dust storm was include the northern parts of Syria (19%), the western parts of
studied. Backward trajectory simulations were employed for Iraq (6%), and Al-Azim/Al-Hawizeh marshlands (4%).
primary source identification. According to primary source If the methodology above is applied to a multiplicity of
assignment, out of 2500 Middle Eastern dust source cells, it dust storms affecting the Ahvaz in the recent years, the source
apportionment will be reliable enough to be used as a guide-
Sr25 line for the authorities to curb the dust crisis in the region.
4% Sr3
Sr24
16% 16%

Sr8
3%

Sr9
8%
Sr23
14% Sr14
1%
Sr15
5%

Sr16
33%

Fig. 8  The contribution of identified sub-regions in the dust storm


generation

13
International Journal of Environmental Research (2019) 13:161–174 171

Fig. 9  Evolution of dust plumes originating from Iraq, Syria Jordan, Cyprus, Lebanon, the northern part of Saudi Arabia, parts of Kuwait, and
southwestern Iran on a 15 June, b 16 June 2016, c 17 June 2017, and d 18 June 2016

2500
15 June 16 June 17 June 18 June 19 June
PM10 Concentration(ug/m3)

2000

1500

1000

500

0
3:00:00
6:00:00

0:00:00

6:00:00

0:00:00
3:00:00

9:00:00

3:00:00

9:00:00

3:00:00
6:00:00
9:00:00
12:00:00
15:00:00
18:00:00
21:00:00

12:00:00
15:00:00
18:00:00
21:00:00

12:00:00
15:00:00
18:00:00
21:00:00

12:00:00
15:00:00
18:00:00
21:00:00

12:00:00
15:00:00

21:00:00
18:00:00
0:00:00

9:00:00

3:00:00

9:00:00

6:00:00

0:00:00

6:00:00

0:00:00

Fig. 10  Hourly ­PM10 concentration (μg/m3) recorded at AQM stations of Ahvaz during 15–19 June 2016

13

172 International Journal of Environmental Research (2019) 13:161–174

Fig. 11  Daily map of MODIS AOD, OMI AI, and MODIS true color images over the study area during 15–18 June 2016

13
International Journal of Environmental Research (2019) 13:161–174 173

Acknowledgements  The authors are grateful to the Atmospheric Blue aerosol products. Rev Geophys 50:RG3005. https​://doi.
Research Center (ARC) of Iran University of Science and Technol- org/10.1029/2012R​G0003​88
ogy for its support to this research. The authors also acknowledge the Givehchi R, Arhami M, Tajrishy M (2013) Contribution of the Mid-
NOAA Air Resources Laboratory (ARL) and NCEI for the provision of dle Eastern dust source areas to PM10 levels in urban receptors:
the HYSPLIT transport and dispersion model, and hourly meteorologi- case study of Tehran, Iran. Atmos Environ 75:287–295
cal observation data, respectively. The MODIS and OMI products were Goudie AS, Middleton NJ (2006) Desert dust in the global system.
acquired through GIOVANNI online software of NASA. We also would Springer, Heidelberg
like to express our thanks to Mr. Alireza Azarnia of DOE- Khuzestan Hamidi M, Kavianpour MR, Shao Y (2017) A quantitative evalu-
branch for providing the hourly ­PM10 concentration data. ation of the 3–8 July 2009 Shamal dust storm. Aeolian Res
24:133–143
Hemond HF, Fechner EJ (2014) Chemical fate and transport in the
environment. Acaemic, New York
References Hernández-Ceballos M, Skjøth C, García-Mozo H, Bolívar J, Galán
C (2014) Improvement in the accuracy of back trajectories using
ARL (2018) Global data assimilation system (GDAS1) archive infor- WRF to identify pollen sources in southern Iberian Peninsula.
mation. https​://www.ready​.noaa.gov/gdas1​.php. Accessed June Int J Biometeorol 58:2031–2043
2018 Hsu NC, Herman JR, Torres O, Holben BN, Tanre D, Eck TF,
Ashrafi K, Shafiepour-Motlagh M, Aslemand A, Ghader S (2014) Smirnov A, Chatenet B, Lavenu F (1999) Comparisons of
Dust storm simulation over Iran using HYSPLIT. J Environ the TOMS aerosol index with Sun-photometer aerosol opti-
Health Sci 12:9 cal thickness: results and applications. J Geophys Res Atmos
Beegum SN, Gherboudj I, Chaouch N, Temini M, Ghedira H (2018) 104(D6):6269–6279
Simulation and analysis of synoptic scale dust storms over the Huang XX, Wang TJ, Jiang F, Liao JB, Cai YF, Yin CQ, Zhu JL,
Arabian Peninsula. Atmos Res 199:62–81 Han Y (2013) Studies on a severe dust storm in East Asia and
Broomandi P, Dabir B, Bonakdarpour B, Rashidi Y (2017) Mineral- its impact on the air quality of Nanjing, China. Aerosol Air Qual
ogical and chemical characterization of suspended atmospheric Res 13:179–193
particles in Ahvaz. Int J Environ Res 11:55–62 Kaskaoutis DG, Rashki A, Houssos EE, Mofidi A, Goto D, Bartzo-
Cao H, Amiraslani F, Liu J, Zhou N (2015a) Identification of dust kas A, Francois P, Legrand M (2015) Meteorological aspects
storm source areas in West Asia using multiple environmental associated with dust storms in the Sistan region, southeastern
datasets. Sci Total Environ 502:224–235 Iran. Clim Dyn 45:407–424
Cao H, Liu J, Wang G, Yang G, Luo L (2015b) Identification of sand Khaefi M, Geravandi S, Hassani G, Yari AR, Soltani F, Dobaradaran
and dust storm source areas in Iran. J Arid Land 7:567–578 S, Moogahi S, Mohammadi MJ, Mahboubi M, Alavi N, Farhadi
Choobari OA, Zawar-Reza P, Sturman A (2014) The global distri- M (2017) Association of particulate matter impact on preva-
bution of mineral dust and its impacts on the climate system: a lence of chronic obstructive pulmonary disease in Ahvaz, south-
review. Atmos Res 138:152–165 west Iran during 2009–2013. Aerosol Air Qual Res 17:230–237
Cuspilici A, Monforte P, Ragusa M (2017) Study of Saharan dust Lee Y, Yang X, Wenig M (2010) Transport of dusts from East Asian
influence on PM10 measures in Sicily from 2013 to 2015. Ecol and non-East Asian sources to Hong Kong during dust storm
Indic 76:297–303 related events 1996–2007. Atmos Environ 44:3728–3738
DOE (2016) Department of Environment, Khuzestan Division. http:// Maleki H, Sorooshian A, Goudarzi G, Nikfal A, Baneshi MM (2016)
khzdo​e.ir/rha/. Accessed June 2018 Temporal profile of PM10 and associated health effects in one
Draxler RR, Gillette DA, Kirkpatrick JS, Heller J (2001) Estimating of the most polluted cities of the world (Ahvaz, Iran) between
PM10 air concentrations from dust storms in Iraq, Kuwait and 2009 and 2014. Aeolian Res 22:135–140
Saudi Arabia. Atmos Environ 35:4315–4330 McGowan H, Clark A (2008) Identification of dust transport path-
Draxler RR, Ginoux P, Stein AF (2010) An empirically derived ways from Lake Eyre, Australia using Hysplit. Atmos Environ
emission algorithm for wind-blown dust. J Geophys Res Atmos 42:6915–6925
115:D16212. https​://doi.org/10.1029/2009J​D0131​67 Middleton N (2017) Desert dust hazards: a global review. Aeolian
Escudero M, Stein A, Draxler R, Querol X, Alastuey A, Castillo S, Res 24:53–63
Avila A (2006) Determination of the contribution of northern Moridnejad A, Karimi N, Ariya PA (2015a) A new inventory for
Africa dust source areas to PM10 concentrations over the cen- Middle East dust source points. Environ Monit Assess 187:582
tral Iberian Peninsula using the Hybrid Single-Particle Lagran- Moridnejad A, Karimi N, Ariya PA (2015b) Newly desertified
gian Integrated Trajectory model (HYSPLIT) model. J Geophys regions in Iraq and its surrounding areas: significant novel
Res Atmos 111:D06210. https​://doi.org/10.1029/2005J​D0063​95 sources of global dust particles. J Arid Environ 116:1–10
Francis DBK, Flamant C, Chaboureau J-P, Banks J, Cuesta J, Brind- Namdari S, Karimi N, Sorooshian A, Mohammadi G, Sehat-
ley H, Oolman L (2017) Dust emission and transport over kashani S (2018) Impacts of climate and synoptic fluctuations
Iraq associated with the summer Shamal winds. Aeolian Res on dust storm activity over the Middle East. Atmos Environ
24:15–31 173:265–276
Ge Y, Abuduwaili J, Ma L, Wu N, Liu D (2016) Potential transport Notaro M, Alkolibi F, Fadda E, Bakhrjy F (2013) Trajectory
pathways of dust emanating from the playa of Ebinur Lake, analysis of Saudi Arabian dust storms. J Geophys Res Atmos
Xinjiang, in arid northwest China. Atmos Res 178:196–206 118:6028–6043
Gebhart KA, Schichtel BA, Barna MG (2005) Directional biases in Prospero JM, Ginoux P, Torres O, Nicholson SE, Gill TE (2002)
back trajectories caused by model and input data. J Air Waste Environmental characterization of global sources of atmos-
Manag 55:1649–1662 pheric soil dust identified with the Nimbus 7 Total Ozone
Ginoux P, Prospero JM, Gill TE, Hsu NC, Zhao M (2012) Mapping Spectrometer (TOMS) absorbing aerosol product. Rev
Global-scale attribution of anthropogenic and natural dust Geophys 40(1):1002. https​://doi.org/10.1029/2000R​G0000​95
sources and their emission rates based on MODIS Deep

13

174 International Journal of Environmental Research (2019) 13:161–174

Rashki A, Arjmand M, Kaskaoutis D (2017) Assessment of dust Stein A, Draxler RR, Rolph GD, Stunder BJ, Cohen M, Ngan F
activity and dust-plume pathways over Jazmurian Basin, south- (2015) NOAA’s HYSPLIT atmospheric transport and dispersion
east Iran. Aeolian Res 24:145–160 modeling system. Bull Am Meteorol Soc 96:2059–2077
Salmabadi H, Saeedi M (2018) Determination of the transport routes Su L, Yuan Z, Fung JC, Lau AK (2015) A comparison of HYSPLIT
of and the areas potentially affected by ­SO 2 emanating from backward trajectories generated from two GDAS datasets. Sci
Khatoonabad copper smelter (KCS), Kerman province, Iran Total Environ 506:527–537
using HYSPLIT. Atmos Pollut Res. https​://doi.org/10.1016/j. Xin Y, Wang G, Chen L (2016) Identification of long-range transport
apr.2018.08.008 pathways and potential sources of PM10 in Tibetan Plateau uplift
Shahsavani A et al (2012) Characterization of ionic composition of area: case study of Xining, China in 2014. Aerosol Air Qual Res
TSP and PM 10 during the Middle Eastern Dust (MED) storms 16:1044–1054
in Ahvaz, Iran. Environ Monit Assess 184:6683–6692 Yu Y, Notaro M, Kalashnikova OV, Garay MJ (2016) Climatology of
Sotoudeheian S, Salim R, Arhami M (2016) Impact of Middle summer Shamal wind in the Middle East. J Geophys Res Atmos
Eastern dust sources on PM10 in Iran: highlighting the impact 121:289–305
of Tigris-Euphrates basin sources and Lake Urmia desicca-
tion. J Geophys Res Atmos 121:14018–14034. https​: //doi.
org/10.1002/2016J​D0251​19

13

You might also like