The impact of dust storms on the Arabian Peninsula and the Red Sea

Introduction Conclusions References


Introduction
Mineral dust is the dominant atmospheric aerosol (Buseck and Posfai, 1999).It plays an important role in the Earth's climate system, although the magnitude and even the sign of its radiative effect at the top of the atmosphere (TOA) remain uncertain.Most airborne dust is generated in desert and semidesert areas and transported across local to global scales.
Dust storms lift millions of tons of dust into the atmospheric boundary layer.Such large quantities of dust cause severe air pollution, reduced visibility, airport shutdowns, and increased traffic and aircraft accidents (Morales, 1979;Hagen and Woodruff, 1973;Middleton and Chaudhary, 1988;Dayan et al., 1991;Chung and Yoon, 1996).Other environmental impacts of dust include reduced soil fertility and crop damage, reduced solar radiation on the surface and as a consequence decreased efficiency of solar devices, damaged telecommunications and mechanical systems, increased occurrence of respiratory diseases, and other impacts on human health (Hagen and Woodruff, 1973;Mitchell, 1971;Fryrear, 1981;Squires, 2007;Jauregui, 1989;Liu and Ou, 1990;Chung and Yoon, 1996;Nihlen and Lund, 1995;Longstreth et al., 1995;Bennett et al., 2006;Bennion et al., 2007).Dust deposition to oceans, however, provides nutrients to ocean surface waters and the seabed (Talbot et al., 1986;Swap et al., 1996 andZhu et al., 1997).
The direct and indirect atmospheric radiative impact of dust has implications in global climate change and presently is one of the largest unknowns in climate model predictions.The African and Asian low-latitude deserts, the so-called dust belt, are the major sources of dust for the entire world (Pye, 1987;Prospero et al., 1987, Arimoto et al., 1989;Laurent et al., 2008;Uno et al., 2005).The dust belt includes P. Jish Prakash et al.: The impact of dust storms on the Arabian Peninsula and the Red Sea the Sahara, arid and semiarid regions in Arabia and Central Asia, and the Taklamakan desert and Gobi in East Asia.During severe dust storms, dust from East Asia can reach far beyond the continent, drifting over the Pacific Ocean to the west coast of North America (Husar et al., 2001;Tratt et al., 2001;Mc Kendry et al., 2001).Similarly, Saharan dust crosses the Atlantic and affects both Americas (Prospero et al., 1987;Reid and Maring, 2003;Haywood et al., 2003).
The Sahara is the world's largest dust source (Washington et al., 2003;Shao et al., 2011).The characteristics of Saharan dust were studied experimentally since the 1990s (Jayaraman et al., 1998;Satheesh and Ramanathan, 2000;Haywood and Boucher, 2000).The Puerto Rico dust experiment (PRIDE) (Reid and Maring, 2003) and the Saharan dust experiment (SHADE) (Haywood et al., 2003) focused on Saharan dust during long-range transport.Two comprehensive field campaigns were conducted in 2006 (Heintzenberg, 2009) and 2008 (Ansmann et al., 2011) under the framework of the Saharan mineral dust experiment (SAMUM) to quantify the optical properties and the radiative impact of Saharan dust near source regions and in the aged mixed plume from Saharan dust and biomass burning aerosols transported west off the African continent.The Fennec 2011 campaign (Washington et al., 2012) made significant advances by providing new measurements close to dust sources over the remote Sahara.
It has long been recognized that radiative effects of mineral aerosols are an important driver of the climate (Tegen andFung 1994, 1995;Tegen et al. 1996;Lie et al., 1996;Andreas, 1996).Early attempts at dust modeling using global models were made by Westphal et al. (1987Westphal et al. ( , 1988)), Gillette and Hanson (1989), Joussaume (1990), and Tegen andFung (1994, 1995).The presence of dust affects atmospheric heating rates and thus atmospheric stability and circulation, which together with changes in the surface energy balance affect the hydrologic cycle (Miller and Tegen, 1998;Miller et al., 2004b).Miller and Tegen (1998) conducted one of the first studies of the radiative effects of dust using a global general circulation model (GCM) coupled with a mixed ocean module to show that dust aerosols reduce the global average surface net radiation by nearly 3 W m −2 during summer in the northern hemisphere.Tegen et al. (1996) showed that dust from disturbed soils causes a decrease in the net surface radiation by about 1 W m −2 .Recently, Balkanski et al. (2007) confirmed that the global average dust-related surface net radiative flux change is in the range of −1.11 to 0.92 W m −2 .However, the radiative effect of dust in desert regions during dust outbreaks could be 2 orders of magnitude stronger (Kalenderski et al., 2013).A comprehensive analysis of simulated global dust distributions and their radiative effects was recently conducted under the scope of the aerosol intercomparison (Ae-roCom) project (Textor et al., 2006;Stier et al., 2007).
The studies mentioned thus far mostly focused on western and central North Africa, the Sahel, and Sahara regions.There are relatively fewer measurement and modeling studies focusing on eastern North Africa and the Arabian Peninsula.This region has been severely under-sampled; there are few available in situ observations, and very few international research campaigns have been conducted in this area compared to the Sahara.In the Arabian Peninsula, dust storms and blowing dust are frequent events throughout most of the year.The major dust sources in the Arabian Peninsula include the Tigris and Euphrates River Valleys, the alluvial plain in Iraq and Kuwait, the low-lying flat lands in the east of the peninsula along the Persian Gulf, and the Ad Dahna and the Rub Al Khali deserts (Kutiel and Furman, 2003;Shao, 2008).Goudie and Middleton (2006) provided an extensive discussion of the causes of dust storms in the Middle East.Eastern Syria, northern Jordan, and western Iraq are the source for most of the fine dust particles (less than 50 µm in diameter) found in Arabian dust storms (NOAA, 2002).This dust is transported far into southern Arabia where it may have contributed to loess deposits on parts of the Arabian Shield and even in the Asir Plateau.Sandstorms containing larger particles (150 to 300 µm in diameter) are also frequent throughout Arabia but rarely reach above 15 m altitude.Most sand grains move by saltation and some by surface creep; they rarely move in suspension by storm winds, haboobs, or dust devils.
Dust storms in the southern Arabian Peninsula are more frequent in summer.In the northern Arabian Peninsula they occur mainly in spring.The peak of dust storm activity occurs usually during the daytime, when intense solar heating of the ground generates turbulence and local pressure gradients (Middleton, 1986a, b).Across the Arabian Peninsula, dust activity peaks from May to August (Prospero et al., 2002;Washington et al., 2003;Barkan et al., 2004;Goudie and Middleton, 2006).The United Arab Emirates unified aerosol experiment (UAE 2 ) focused on dust in the coastal and desert regions of the United Arab Emirates in August to September 2004 to evaluate the properties of dust particles that converge in the UAE region from numerous sources and their impact on the radiation budget (Reid et al., 2008).Mohalfi et al. (1998) studied the effect of dust aerosols on the synoptic system and showed that dust aerosol radiative heating strengthens Saudi Arabian heat low.Kalenderski et al. (2013) simulated a winter dust event that occurred in January 2009 over the Arabian Peninsula and the Red Sea to study various dust phenomena.Notaro et al. (2013) investigated the temporal and spatial characteristics of Saudi Arabian dust storms using trajectory analysis.Using MODIS data, they found that the highest aerosol optical depth (AOD) is achieved during dust storms that originate from the Rub Al Khali and Iraqi Deserts.
This study aims to quantify the impact of dust on the Arabian Peninsula and Red Sea, focusing on the severe dust storm observed from 18 to 22 March 2012.This storm likely drew at least some of its dust from Iraq, Iran, Kuwait, and the Arabian Peninsula Rub' al Khali desert (Empty Quarter), sprawling over parts of Saudi Arabia, Yemen, Oman, and UAE.The simulation period was from 00:00 UTC (Universal Time Coordinate) on 5 March to 00:00 UTC on 25 March 2012 with the model spin-up during the first 6 days.Using model output and observations, we provide an improved estimate of the storm's impact on the terrestrial and oceanic environment.
The remainder of this paper is organized as follows.Section 2 discusses the model's configuration and methodology.Section 3 examines the environmental conditions that led to the development of the dust storm, estimates the dust emissions, load, and deposition, discusses the structure of the dust storm, compares model results with available Aerosol Robotic Network (AERONET) and satellite observations, presents the vertical structure of atmosphere and dust distribution, and calculates radiative effect of dust.We offer a summary of the results in Sect. 4.

Methodology
In this study, we combine advanced high-resolution modeling of meteorological and dust processes with analysis of ground-based and satellite observations of aerosols and meteorological fields.

Modeling
The Weather Research Forecast (WRF) system is a mesoscale forecast model with an incorporated data assimilation capability that advances both the understanding and prediction of weather (Skamarock et al., 2008).WRF has been utilized in a variety of research and operational projects, from the scale of convective storms to the scale of continental weather patterns (Michalakes et al., 2005).The chemistry/aerosol module (WRF-Chem) extends WRF by incorporating a chemistry module that interactively simulates emissions of aerosols and gases, their transport, turbulent and convective mixing, and chemical and microphysical transformations of trace gases and aerosols (Grell et al., 2005).
Here, we configured WRF-Chem with the regional acid deposition model 2 (RADM2) photochemical mechanism (Stockwell et al., 1990), the Fast-j photolysis scheme (Wild et al., 2000), the modal aerosol dynamics model for Eu-rope (MADE), and the secondary organic aerosol model (SORGAM) (Ackermann et al., 1998;Schell et al., 2001).We also employed the Goddard global ozone chemistry aerosol radiation and transport (GOCART) dust emission scheme (Ginoux et al., 2001) to calculate the influx of dust into the atmosphere.GOCART simulates dust emissions as a function of surface wind speed, surface erodibility, and surface wetness (Chin et al., 2002).Emission flux, F p , for a specific aerosol size group, p, is expressed as where C is a dimensional constant coefficient of proportionality; S is the space-dependent dimensionless erodibility field taken from Ginoux et al. (2001) with a spatial resolution of 0.25 • × 0.25 • ; U 10 m is the wind speed 10 m above the ground; U t is the threshold velocity of wind erosion, which depends on particle size and surface wetness; and s p is a particular particle size mode mass fraction emitted in the atmosphere.The original GOCART emission scheme was coupled with the 8 bin aerosol model.However, in this study WRF-Chem is configured with the MADE/SORGAM that uses the modal approach with three log-normal distributed modes (aitken, accumulation, and coarse) to represent sulfate, nitrate, ammonium, organic matters, black carbon, and sea salt.Mineral dust is assumed to have only accumulation and coarse modes.The GOCART emission scheme has been modified to couple with the MADE/SORGAM.It calculates a dust mass flux from the surface (see Eq. 1) assuming that, by default, s p = 0.07 for accumulation mode, which for emitted particles has the modal diameter D = 0.6 µm, width σ =2 µm; and s p = 0.93 -for coarse mode with D = 6 µm, σ = 2.2 µm.According to MADE/SORGAM formulation, the modal diameters change in the atmosphere due to microphysical processes but the widths of both distributions remain fixed.The total emission flux is adjusted using constant C in Eq. (1) to fit AERONET observations, as discussed by Zhao et al. (2010) and Kalenderski et al. (2013).In our simulations, we used C = 0.8 mg s 2 m −5 to achieve consistency between the simulated and AERONET aerosol optical thickness (AOD) observed from 11 to 24 March 2012.
To cover the entire area affected by the storm and to account for processes that were responsible for the storm's development, we ran the WRF-Chem model in the spatial domain from 4 • N to 40 • N in longitude and 25 • E to 80 • E in latitude with 10 km horizontal resolution.In our calculations, we therefore used 495 grid points in the west-east direction and 396 grid points in the south-north direction, as well as 40 vertical levels with the top-of-the-model domain at the 10 hPa level.The following physical parameterizations were used to configure the WRF-Chem simulations: the Lin microphysics scheme, the rapid radiative transfer model (RRTMG) for both long-wave (LW) and shortwave (SW) radiation, the Mellor-Yamada-Janjic (MYJ) boundary layer scheme, the Noah land surface model, and the Grell cumulus parameterization.Lateral boundary and initial condi-P.Jish Prakash et al.: The impact of dust storms on the Arabian Peninsula and the Red Sea tions for meteorological fields were provided by the National Centers for Environmental Prediction (NCEP) global analysis (FNL).We used NCEP daily global sea surface temperature (SST) analysis (RTG_SST_HR) to update SST every 6 h.The various model domain configurations and physics options are summarized in Table 1.

Observations
To test and constrain the model simulations, we used reanalysis fields as well as meteorological and aerosol observations available in the Middle East and surrounding areas.

Space-borne aerosol instrumental observations
Aerosol optical depth (AOD) data from MODIS sensors onboard the Terra and Aqua satellites are widely used in aerosol studies (Salomonson et al., 1989).The MODIS instrument provides high radiometric sensitivity (12 bit) in 36 spectral bands ranging in wavelength from 0.4 µm to 14.4 µm.Two bands are imaged at a nominal resolution of 0.25 km at nadir, five bands at 0.50 km, and the remaining 29 bands at 1 km.A ±55 • scanning pattern of the Earth Observing Satellite (EOS) orbit at 705 km altitude achieves a 2330 km swath and provides global coverage every 1 to 2 days.The MODIS daily Level-2 AOD data are produced at the spatial resolution of a 10 km × 10 km (at nadir).There are two MODIS aerosol data products, one containing data collected from the Terra platform and the other containing data collected from the Aqua platform.To maximize the observation coverage, we used AOD retrievals over land and sea derived from the dark target product (Remer et al., 2005(Remer et al., , 2008) ) and the "deep blue" product over bright land surfaces (Hsu et al., 2004(Hsu et al., , 2006)).The dark target ocean and land AOD products were available from both Terra and Aqua, but the deep blue retrievals were only available from Aqua.In this study, we have used daily deep blue (Level 2) data in combination with the standard ocean algorithm for comparison with the simulated aerosol optical properties.
The Spinning Enhanced Visible Infrared Radiometer (SE-VIRI) (Aminou et al., 2002) is a line scanning radiometer currently located on-board the European geostationary meteorological satellite, Meteosat-9; it was previously on Meteosat-8.This instrument provides data in four visible and near infrared channels and eight infrared channels with a resolution of 3 km at nadir.In this study we used the 0.6 µm channel.An advantageous feature of SEVIRI is its ability to image the study area with high temporal resolution of 15 min.This allows SEVIRI to track aerosol events, which offers a great advantage over polar orbiting instruments like MODIS that usually see the particular scene once a day.The SEVIRI's spatial resolution is, however, coarser than that of MODIS, but the aerosol gridded products for the Arabian Peninsula from both instruments are on the 10 km × 10 km grid.

Ground-based aerosol observations
The Aerosol Robotic Network (AERONET) uses Cimel robotic sun photometers and provides observations of AOD on up to eight wavelength channels between 0.340 and 1.640 µm (Holben et al., 1998) and angular distribution of sky radiance at four wavelengths (0.440, 0.675, 0.870, and 1.020 µm).In addition to the direct measurement of AOD, the inversion algorithm retrieves physical and optical properties of aerosols that comprise the aerosol refractive index, column average size distribution, single scattering albedo, and asymmetry parameter.The maximum AERONET uncertainty in AOD retrieval is estimated to be 0.02 with the highest error in the ultraviolet wavelength (Holben et al., 1998;Eck et al., 1999), and the calibrated sky radiance measurements typically have an uncertainty less than 5 % (Holben et al., 1998).Cimel sun photometers are calibrated annually by comparison to an AERONET master instrument.
The AERONET data are at high time resolution but have limited spatial coverage.During the study period, AOD observations were available at Kuwait University (29.32 • N, 47.97 • E), at Mezaira (23.14 • N, 53.77 • E), and at our own site established in February 2012 on the King Abdullah University of Science and Technology (KAUST) campus (22.30• N, 39.10 • E).We chose cloud-screened (Level 1.5) AERONET AOD for our model validation since Level-2 data (cloud-screened and quality-assured data) were not available during the simulation period.The Angstrom power law was used for comparing the simulated optical properties output at 0.60 µm with the AERONET measurements as follows: AOD(0.600)= AOD(0.675)× 0.600 0.675 where α is the Angstrom exponent calculated from the AERONET measurement as α = ln AOD(0.440)AOD(0.675)ln 0.675 0.440 . (3)

Weather station meteorological observations
There are meteorological data from 108 surface weather stations and 12 upper air stations available in the area of interest.For the purposes of this study, we focus on the upper air radiosonde data.The radiosonde (RS) is a balloonborne instrument platform with radio transmitting capabilities.Data from the radiosonde are interpreted at the launching station and entered into a worldwide communications network.The time sampling of RS data is usually twice daily at 00:00 UTC (midnight) and 12:00 UTC (noon).RS profiles contain pressure, temperature, relative humidity, wind speed, and wind direction.Depending on the sensor and balloon used at each location, the vertical resolution of RS data typically has 40 levels between 1000 and 50 hPa.We used  (Lin et al., 1983) PBL: Mellor-Yamada-Janjic (MYJ) TKE scheme (Janjic, 2001) Surface: unified Noah land-surface model (Chen and Dudhia, 2001) Cumulus convection: new Grell scheme (Grell and Devvenyi, 2002) Emission scheme GOCART scheme (Ginoux et al., 2001) Aerosol model MADE/SORGAM (Ackerman et al., 1998;Schell et al., 2001) Photolysis scheme Fast-J (Wild et al., 2000) Gas-phase mechanism RADM2 (Stockwell et al., 1990) Dust SW refractive index 0.003 observations from three RS stations for a comparison with model results.Stations are identified by their World Meteorological Organization (WMO) five digit codes.The three stations over the Arabian Peninsula selected for this study include Jeddah (41024), Abu Dhabi (41217), and Kuwait (40582).The vertical temperature RS profiles were retrieved from the upper air archive at the University of Wyoming (http://weather.uwyo.edu/upperair/sounding.html).

Reanalysis output
To compare meteorological fields and spatial patterns, we use output from the European Centre for Medium-Range Weather Forecast (ECMWF) ERA-Interim (ERA-I) reanalysis (Dee et al., 2011).It is the latest global atmospheric reanalysis produced by ECMWF as a transition between ERA-40 and a future reanalysis project.It provides information on a large variety of surface parameters (3-hourly), describing weather as well as ocean-wave and land-surface conditions and 6-hourly upper-air parameters (37 pressure levels up to 1 hPa) on a 0.7 • × 0.7 • grid.ERA-I uses an improved atmospheric model and a more sophisticated data assimilation method (4D-Var) for atmospheric analysis compared to ERA-40.Information about the current status of ERA-I production and availability of online data can be found at (http://apps.ecmwf.int/datasets/data/interim_full_daily/).

Results
The dust storm from 18 to 22 March 2012 covered an extremely large area and disrupted human activities in Iraq, Iran, Kuwait, Syria, Jordan, Israel, Lebanon, UAE, Qatar, Bahrain, Saudi Arabia, Oman, Yemen, Sudan, Egypt, Afghanistan, and Pakistan.Many airports in these countries were shut down because of dust impact on visibility and machinery.

Meteorological conditions and storm spatial-temporal development
Figure 1 shows geopotential height from WRF-Chem simulations and ERA-Interim reanalysis at 500 hPa at 00:00 UTC from 17 to 19 March 2012.Sea-level pressure for the same days at 00:00 UTC is shown in Fig. 2. The model results compare well with the reanalysis fields in both figures.On 17 March there was a trough in the mid-troposphere at 500 hPa stretching over the eastern Mediterranean Sea, Syria, Iraq, western Iran, and northern Saudi Arabia.This upper air disturbance caused a significant pressure decrease in the lower atmosphere and a low-pressure system consequently developed in eastern Iraq and northwestern Iran with a central sea-level pressure of 1008 hPa.A high pressure is seen over northeastern Afghanistan and Tajikistan.The spatial distribution of 10 m wind bars at 09:00 UTC from 17 to 19 March is shown in Fig. 3. Strong surface winds of 15 m s −1 are observed over northwestern Iraq on 17 March at 09:00 UTC (Fig. 3a).
On 18 March the high-level trough deepened and moved southeastward into the northern Arabian Peninsula covering Iraq and western Iran.A surface high pressure builds up over the eastern Mediterranean Sea.The presence of a low pressure in southern Saudi Arabia, the Gulf of Oman, and central Iran in combination with a high in the eastern Mediterranean Sea enhanced a strong pressure gradient along the Arabian Peninsula.As a result, a strong surface wind of 15 m s −1 developed to the west of the low near the Persian Gulf in southeastern Saudi Arabia and Qatar and to the northeast of UAE (Fig. 3b).This is well above the generally assumed threshold wind speed (5 to 6 m s −1 ) for dust sus- pension (Gillette, 1978).These strong winds forced the dust emission and transported dust to the south and southwest of the Arabian Peninsula.
On 19 March the high-level trough moved towards eastern Iran and western Afghanistan.The low-pressure system in the southern Arabian Peninsula was overtaken by the highpressure system.A low-pressure system that developed in central Pakistan in combination with the high pressure in the northern Arabian Peninsula increased surface winds in southern Iran up to 20 m s −1 (Fig. 3c).One of the most identifiable synoptic features in the Arabian Peninsula during the simulation is the low-pressure system that moved southeastward through the Peninsula on 19 March 2012, suggest-ing that the dust storm was closely related to the south and southwestward propagation of a cold front.Figure 4 shows the simulated near-surface wind field and dust concentration at 08:00 UTC and 15:00 UTC on 18 March 2012.Comparisons with synoptic observations confirm that the flow field is well simulated.The cold front is a line-shaped narrow region stretching from the western coast of Saudi Arabia to eastern UAE, as can be seen from the dense temperature contours (Fig. 4a).Strong northeasterly wind (maximum 15 m s −1 ) prevailed behind the cold front, resulting in strong dust emissions and high dust concentrations.The dust was carried southwestward by the northeasterly wind to converge on the frontal area, resulting in high dust concentrations near the front.The dust was then transported westward over the Red Sea along the front and towards the southern Arabian Peninsula.In southern Saudi Arabia, the dust front rapidly advanced to the south and southwest.Figure 4 shows that within a time period of 7 h, the dust front advanced more than 150 km.At 15:00 UTC on 18 March (Fig. 4b), dust was wide spread, with the same maximum dust concentration as those at 08:00 UTC.
Dust and sand lift both ahead of and behind cold fronts; winds tend to be stronger behind the front than ahead of it (Sissakian et al., 2013).Behind an advancing cold front, southern Iraq and northern Saudi Arabia experienced blowing dust or shamal-like conditions (Hamidi et al., 2013).

The vertically integrated mass balance and optical depth of the dust
Dust emissions are calculated in the model interactively from Eq. (1) using simulated surface winds that drive dust emissions.To ensure that the model could adequately generate dust, we compare the simulated 10 m wind field (U 10 m ) with the output from the ECMWF Interim global reanalysis, ERA-I.The U 10 m is extracted from ERA-I and interpolated to WRF projections for comparison.The simulated and reanalysis fields are well correlated (not shown) with a timeaveraged correlation coefficient of 0.79. Figure 5a shows the erodibility field S used in Eq. ( 1) depicting the major dust emission areas.Figure 5b-d shows simulated dust emissions  • E. At the center of this valley, there is a large, 1087 km 2 , dry salt lake, Hamume Jaz Murian.The majority of dust at the eastern coast of the Persian Gulf was generated from this lake.Dust source regions along the western coast of the Arabian Peninsula were also active.The area of dust generation was relatively compact but maximum dust emissions exceeded 500 µg m −2 s −1 due to strong winds (Fig. 3).The spatial distribution of daily average simulated dust deposition rates for 17, 18, and 19 March 2012 is shown in Fig. 6.There was no precipitation during the simulation pe- riod, and therefore the dominant dust removal mechanism was dry deposition.The deposition velocity is higher for large particles, and therefore coarse dust particles tend to deposit closer to a source than the fine ones.Figure 6 shows that the maximum dust deposition exceeds 90 µg m −2 s −1 , about one-third of maximum dust emissions rates.On 17 March (Fig. 6a) a lot of dust was deposited in the northeastern Persian Gulf, relatively far from the source regions.On 18 March the strongest dust deposition happened in wide surroundings of the major dust sources; significant amount of dust was deposited in the Persian Gulf and the Red Sea (Fig. 6b).On 19 March (Fig. 6c) dust was deposited in the Red Sea, the Gulf of Oman, and the Arabian Sea off the Iran and Pakistan coasts.
Total dust emission/deposition is computed by integrating emission/deposition over the domain area.Figure 7a shows the time series of daily total emission and deposition (in Mt day −1 ) for the simulation period of 11 to 24 March 2012.The daily emissions (F s ) and depositions (F d ) are of the same order of magnitude.The maximum daily emis-sion, F s = 13.10Mt day −1 , and maximum daily deposition, F d = 10.06Mt day −1 , occurred on 18 March.Dust was dispersed and deposited over much wider areas than those where dust was generated.The maximum total hourly dust emission and deposition rates of 0.85 Mt h −1 and 0.52 Mt hr −1 , respectively, occurred in the daytime at about 07:00 UTC on 18 March (not shown).This is expected because the atmosphere in the daytime is unstable, and the momentum transfer from the free troposphere to the surface, which enhanced the surface winds, is more efficient.Dust emissions are substantially weaker in the night when the boundary layer collapses.
The comparison with other similar studies shows that the dust storm of 18-22 March 2012 is in the range of the most powerful dust events considered in the literature (Zhao et al., 2010;Shao et al., 2003).A portion of the dust deposited in the ocean provides nutrients to marine ecosystems and increases the ocean's net productivity.Dust emitted from southern Iraq, Iran, and the eastern Arabian Peninsula was mainly deposited in the Persian Gulf and the Arabian Sea.Dust emitted from the western coast of the Arabian Peninsula  dust deposited on the land was 66.5 Mt and over the ocean 6.5 Mt, out of which 1.2 Mt was deposited on the Red Sea.
Although the magnitude of dust deposition over the oceans is not accurately known, there is evidence that dust deposition could have a significant impact on chemical and biological processes in the oceans (Martin, 1990;Watson et al., 2000;Fan et al. 2006;Sunda and Huntsman, 1997).In the present study, we have calculated the amount of dust deposited over the ocean and particularly in the Red Sea.Aeolian deposition is especially important for the Red Sea because this sea has very little fresh water discharge from the coastal areas that is mostly associated with flash flood events.
To calculate the total frequency and spatial distribution of severe dust events over the Arabian Peninsula and the Red Sea region, we choose to analyze the MODIS satellite images for the 2002-2013 period, which were provided under the scope of natural hazard reports in this area by NASA's rapid response system (http://earthobservatory.nasa.gov/NaturalHazards).For this 12-year period, the number of cases when a dust plume covered more than 20 % of the Middle East area reached 237 or approximately 20 annually.The number of dust storms, which affected the Red Sea, was 71 or 5-6 annually.We counted dust events based on their extent and optical depth.The obtained number of regional storms is consistent with the results of previous studies (e.g., Rezazadeh et al., 2013 andProspero et al., 2002).Assuming this storm frequency and dust deposition of 1.2 Mt per storm, as calculated in this study, we estimate the total annual mineral dust deposition on the Red Sea to be at least 6 Mt.An extended analysis of the intensity and variability of the dust storm activity over the Arabian Peninsula, as well as contribution of dust deposition during fair weather conditions, are required to make this estimate more certain.
Figure 8a shows the simulated spatial pattern of the dust load at 10:00 UTC on 19 March.Dust plumes were seen spreading over the Gulf of Oman, the Red Sea, and the southern coast of the Arabian Peninsula.The highest vertically integrated dust load exceeded 4.5 g m −2 over the Gulf of Oman and southern Saudi Arabia bordering Yemen.The image taken by Aqua/MODIS at 09:50-10:05 UTC (Fig. 8b) shows a significant dust load over the entire southern Arabian Peninsula and its coastline along the Arabian Sea as well as over the Persian Gulf.The simulated spatial distributions of the dust loads in Fig. 8a are quite consistent with the MODIS dust plume image in Fig. 8b at most locations, suggesting that the model correctly predicted the dust event.Both model simulations and satellite imagery clearly show dust aerosols covering not only the land but also very large areas of the oceans.
Figure 9a-c shows a spatial distribution of AOD from WRF-Chem, MODIS, and SEVERI retrievals.Since the number of passes of the MODIS instrument on-board the Aqua satellite over the domain of interest during the time of interest is limited, two 0.55 µm retrievals at 09:50 and 10:05 UTC on 19 March were used for comparisons with the simulated AOD (0.60 µm) at 10:00 UTC.A combination of both standard ocean and deep blue products was used to get the maximum spatial coverage over the simulated domain.The model captures the spatial distributions of AOD well.Both modeled and observed optical fields indicate a high dust plume over the Persian Gulf, Gulf of Oman, Red Sea, and southern Arabian Peninsula.
Figure 9d-f compares hourly simulated and observed AODs at three AERONET sites (KAUST campus, Kuwait University, and Mezaira) during the simulation period.The AERONET 15 min observations were averaged hourly to compare with modeled values.The time series match well with each other most of the time.Even though the magnitude of the simulated AOD at the KAUST campus was lower compared to that of AERONET on 19 March, the simulations were able to reproduce the peak and the temporal dependence quite well.

Vertical mixing of dust within the boundary layer and free troposphere
The vertical distribution of dust aerosols is important for estimating their radiative effect and their impact on circulation and climate (Forster et al., 2007;Claquin et al., 1998; Huang   et al., 2009;Zhu et al., 2007).The evolution and diurnal cycle of the atmospheric boundary layer directly affects the life cycle and vertical redistribution of dust particles (Cuesta et al., 2009).

Vertical structure of a dust layer
Most dust aerosols reside in the atmospheric boundary layer.
The portion of aerosols that moves above the boundary layer to the free troposphere is subject to long-range transport and therefore is of particular interest.N'Tchayi Mbourou et al. (1997), Engelstaedter et al. (2006), andChaboureau et al. (2007) reported that the diurnal cycle of dust in the Sahel and Sahara is the result of daytime boundary-layer dry convection affecting the uplift and transport of dust in the almost 6 km deep boundary layer (Gamo, 1996;Carolina and Martin, 2012).The planetary boundary layer (PBL) can be influenced by different mechanisms associated with the coastal breezes, terrain, and surface properties (Warner and Sheu, 2000).The Asir mountain range in western Arabia runs along the Red Sea coast and is highest in the south where its altitude reaches 3 km.During the day, the surfaces of the mountains heat the air higher up in the atmosphere quicker than the ocean surface can.This leads to enhancing subsidence over the Red Sea and suppresses the growth of daytime PBL (e.g., Kuwagata and Kimura, 1995;Whiteman, 1982;Kimura and Kuwagata, 1995;Bader and McKee, 1983).Upward motion over heated elevated terrain during the day causes the vertical temperature profile to become less stable and thus possibly tends to increase the PBL height over the Asir mountains.
Figure 10 shows the diurnal cycle of simulated vertical profiles of dust concentration, potential temperature, wind vector, and PBL height in the altitude/longitude cross section at 21.7 • N on 19 March 2012.The aerosol concentration reaches 2000 µg m −3 .The thick dust plume stretches from 28.7 • E to 70.2 • E. At 06:00 UTC heating has just started, and the PBL height nearly follows the topography.Over mountain and desert areas, the PBL is shallow.The inland convection increases, and enhanced subsidence over the Red Sea suppresses the growth of the PBL.The sea breeze starts developing at about 07:00 UTC.At 12:00 UTC this process continues but much more slowly as the sea breeze front moves inland and the PBL height grows farther to the east.A well-mixed, deep boundary layer in our simulations develops during the day up to the height of 3.5 km in the afternoon as seen in Fig. 10c.At night, the PBL collapses, resulting in an elevated dust-laden residual boundary layer above the surface inversion (Fig. 10d).
The temporal PBL variability is characterized by the more rapid PBL growth over the high elevations in the western Arabian Peninsula early in the heating cycle.The high PBL then spreads towards the eastern Peninsula.During the night, the nocturnal jets contribute to the dust mobilization close to the surface, while aerosol particles increasingly sediment from the residual layer after the collapse of the PBL.The peak of dust vertical mixing is observed at about 12:00 UTC (Fig. 10c) over the Asir mountains in southwestern Saudi Arabia, where a significant amount of dust penetrates in the free troposphere downwind of the Asir mountain ridge.
To quantify the amount of dust mixed above the boundary layer, we calculate the domain average loading below and above the diurnal maximum PBL height (Fig. 11) that approximates the top of a residual layer.Aerosols remaining in the residual layer will be again entrained in the PBL during the next diurnal cycle; therefore we assume that only aerosols above the residual layer are disconnected from the surface.Because the PBL dynamics are different over land and over the sea, we did these calculations separately for the land and ocean areas within the model domain.Figure 11 shows that during the simulation period, approximately 15 % of dust over land is in the free troposphere.Over the sea, it is often more than 50 %.The storm increased the dust loading over land by approximately 30 % and almost tripled it over the sea.
To assess the simulated vertical structure of the atmosphere we used RS soundings from the Jeddah's King Abdulaziz Airport (21.7 • N, 39.18 • E).The model and RS profiles of the water vapor mixing ratio and potential temperature (Fig. 12a) show that the PBL height on 19 March 2012 at 00:00 and 12:00 UTC is captured well by the model.The simulated PBL height was approximately 3 km in the afternoon as seen in Fig. 10c, which is in good agreement with the position of the inversion layer from Jeddah's soundings (Fig. 12a).The potential temperature at 12:00 UTC (Fig. 12a) is almost constant between the ground and 3 km altitude as a result of strong mixing in the PBL.The RS vapor mixing ratio profile also shows the mixing height to be at 3 km.The vertical profile of the magnitude and direction of the horizontal wind over Jeddah from RS profiles on 19 March (00:00 and 12:00 UTC) is shown in Fig. 12b.The wind direction shows a strong diurnal change in the boundary layer at 1 km altitude.The wind over Jeddah was found to be east/southeast in the lowest 1 km and mostly west/northwest above 3 km.The wind speed exceeds 15 m s −1 near the surface and is more than 50 m s −1 at 12 km.The surface wind at 00:00 UTC is enhanced by the land breeze.This is a very important factor for the transport and deposition of dust over the Red Sea during the nighttime (Fig. 10a, d).

The direct radiative effect of dust
Aerosols directly influence the Earth's radiative budget by absorbing, scattering, and emitting LW radiation (Cautenet et al., 1991;Markowicz et al., 2003;Ackerman and Chung, 1992;Haywood et al., 2005) and scattering and absorbing SW radiation (direct effect) (Tegen et al., 1996;Haywood and Boucher, 2000;Harrison et al., 2001;Haywood et al., 2001;Sokolik et al., 2001), and indirectly by altering cloud droplet size distribution (indirect effect) (Twomey, 1977;Albrecht, 1989).Because there is little cloudiness over deserts, the direct radiative effect (DRE) of dust is of primary interest in this study.The surface SW cooling by dust results in large reductions in latent and sensible heat fluxes from the surface to the atmosphere (Miller et al., 2004b;Shell and Somerville, 2007).This reduces the turbulent energy within the PBL and the downward transport of momentum to the surface, which suppresses the surface wind speed and dust generation (Miller et al., 2004a).In the troposphere, the absorption of both incoming and reflected solar radiation, as well as emission of thermal radiation by dust particles, affects the air temperature and circulation (Alpert et al., 1998;Miller and Tegen, 1998;Sathesh et al., 2007).Depending on their physical and optical properties as well as their chemical composition, aerosols exert a cooling or warming influence on the atmosphere and underlying surface (e.g., Chylek and Wong, 1995;Sokolik and Toon, 1996;Miller and Tegen, 1998;Ahn et al., 2007;Shell and Somerville, 2007;Balkanski et al., 2007).Large uncertainties in the assessment of the radiative impact of mineral dust on regional and global scales have been pointed out in a number of early and recent studies (Sokolik and Toon 1996;Tegen et al., 1996;Carlson and Benjamin, 1980;Sokolik and Golitsyn, 1993;Ackerman, 1997;Liao and Seinfeld, 1998a;Claquin et al., 1998;Stier at al., 2007).
To assess the DRE of dust aerosols we calculate the change of total clear-sky radiative flux F , at the top of the atmosphere (TOA) and at the bottom of the atmosphere (BOA) and in simulations when aerosols are present (F A ) and absent (F C ).Then the DRE can be defined at BOA and TOA as where the arrows indicate the direction of the fluxes: ↓ denotes a downward flux and ↑ an upward flux.This convention is used for both LW and SW radiation.Definitions (Eqs. 4 and 5) imply that negative values of F at the BOA and TOA are associated with an aerosol cooling effect, while positive F is associated with warming.The DRE, as defined here, is different from a conventional instantaneous direct radiative forcing (DRF) because it includes model responses (see, e.g., Stenchikov et al., 1998).However, the clear-sky SW DRE discussed below should be a fairly good approximation of the clear-sky SW DRF.

The shortwave radiative effect
The spatial distributions of daily average clear sky SW, LW, and net (SW+LW) dust DRE at BOA and TOA on 19 March are shown in Fig. 13.The SW DRE is mostly negative at the surface, reaching −134 W m −2 (Fig. 13a).The domainaveraged SW DRE at BOA equals −16 W m −2 and tends to cool the surface.The daily averaged SW DRE at TOA over desert areas is about 10 W m −2 and tends to warm the surface-atmosphere system over the desert (Fig. 13b).However, the domain-average SW DRE at TOA equals −4.25 W m −2 .The warming effect of dust aerosols over desert surfaces is largely explained by two factors: absorption of solar radiation by dust and high reflectance of the underlying surface.Dust absorbs and scatters the SW radiation.Both these effects reduce the downward flux at BOA but tend to compensate for each other at TOA.Therefore, the BOA cooling effect is robust, but the TOA warming is very sensitive to the dust's single scattering albedo and surface reflectivity.Over the ocean, the SW DRE at TOA is negative due to the smaller ocean albedo.Over the ocean, the dust layer increases the outgoing SW radiation since it is a much better reflector than the underlying dark water.The negative values of SW DRE at TOA over the ocean are consistent with the results from Perez et al. (2006) for Mediterranean dust.Daily averaged SW BOA aerosol effect up to −21.1 Wm −2 was reported by McFarlane et al. (2009) during the atmospheric radiation measurement mobile facility deployment in Niamey and Niger.During the winter monsoon season, Satheesh et al. (2006) reported daily averaged surface SW radiative effect from −23.6 to 12 W m −2 over the Arabian Sea for 8 years from 1995 to 2002 with a maximum instantaneous value of −51 W m −2 .During the summer monsoon season, the daily averaged surface SW radiative effect was in the range of −24.2 to −13.1 W m −2 with a "maximum" instantaneous value of −45 W m −2 .At TOA, the SW radiative effect of dust was in the range of −16.5 to −9.1 W m −2 during the summer monsoon and −7.3 to −6.0 W m −2 during the winter monsoon.Haywood et al. (2003) revealed an instantaneous direct SW radiative effect at BOA from a large Saharan dust plume, advected off the coast of western Africa, to be approximately −130 W m −2 .The maximum instantaneous SW DRE at the surface in this study was −449 W m −2 on 19 March 2012.

The long-wave radiative effect
Figure 13c and d shows that the daily averaged LW DRE is mostly positive at TOA and BOA and is larger at the surface.Dust absorbs terrestrial radiation coming from the surface and atmosphere and generally re-emits it at lower temperatures than the underlying surface.The presence of dust reduces the outgoing LW radiation and increases the downward TOA LW flux.The maximum daily average LW DRE at the surface is 43 W m −2 with a domain average value of 6 W m −2 .At TOA, the maximum daily average value of LW DRE is 8.4 W m −2 with a domain average value of 0.25 W m −2 .The LW warming can offset the SW cooling effect of dust at TOA. Zhang and Christopher (2003) estimated the dust aerosol LW TOA DRE over the Sahara from instruments on-board a satellite and showed that there is a strong warming effect of 7 W m −2 over the cloud-free Sahara regions for September 2000 and, when averaged over six regions from the study area, the LW TOA DRE ranges from −1 to 15 W m −2 .Liao and Seinfeld (1998a) reported LW BOA DRF ranging from 0.9 to 1.4 W m −2 for the 0.1 g m −2 dust loading using a one-dimensional column radiation model.From low dust loading under dry tropics atmospheric conditions, Sokolik et al. (1998) calculated the LW DRF at BOA in the range of 7 to 14 W m −2 and the LW DRF at TOA in the range of 2 to 7 W m −2 .We found that under dust storm conditions, the LW DRE at the surface varies between 50 and 80 W m −2 and the LW DRE at TOA varies between 15 and 25 W m −2 .It is smaller but comparable with the SW DRE.

Net DRE
Figure 13e and f shows the net (SW + LW) DRE at BOA and TOA, respectively.At the surface, the net DRE is negative in most of the region with an extreme value of −94 W m −2 .The strong negative net DRE at BOA is seen over the region where the model and satellites show the highest dust atmospheric loading (Fig. 8).In Fig. 13 the domain average value of net DRE is −10 W m −2 and −4 W m −2 at BOA and TOA, respectively.The SW DRE exceeds the LW DRE at both BOA and TOA. Figure 13e and f shows that net DRE causes cooling at the surface, while at the TOA it causes warming over the land but cooling over the oceans.This pattern in the net DRE agrees with that found by Yue et al. (2010), P. Jish Prakash et al.: The impact of dust storms on the Arabian Peninsula and the Red Sea Ackerman and Chung (1992), Woodward (2001), and Liao et al. (2004).The sign of the net DRE at TOA could be either positive (warming) or negative (cooling) depending on several key variables, such as the surface albedo, particle size, vertical distribution of dust, optical depth, and the imaginary part of the dust refractive index (Tegen and Lacis, 1996;Liao and Seinfeld, 1998a, b).In this study, the imaginary part of the dust refractive index is set to 0.003 for the entire SW spectrum.Table 3 summarizes the domain average SW, LW, and Net DRE at TOA and BOA for 19 March 2012.

Atmospheric heating rates
The rate of change of temperature in an atmospheric layer (K day −1 ) due to radiative heating/cooling is called the radiative heating/cooling rate.The simulated instantaneous dustinduced perturbations (similar to definitions Eqs. 4 and 5) of SW heating rates over Jeddah, Abu Dhabi, and Kuwait at 12:00 UTC on 19 March are shown in Fig. 14.The available temperature profiles at 12:00 UTC from RS are also shown.
The heating rates vary with the altitude following the vertical distribution of the aerosols.The maximum heating rates over Kuwait (∼ 4.4 K day −1 ) and Abu Dhabi (∼ 4 K day −1 ) are at approximately 1 km altitude.The heating rates reach their maximum value at 3 km over Jeddah (∼ 4.9 K day −1 ).Over Mezaira (not shown), the dust-generated heating rates in the lower troposphere reach 9 K day −1 .The maximum heating over Kuwait and Abu Dhabi is at lower levels compared to that over Jeddah.This is because Kuwait and Abu Dhabi are closer to the dust source regions and hence there is more dust in the lower levels of the atmosphere.
The RS temperature profiles help to clarify the effect of aerosols on the vertical structure of the atmosphere under storm conditions.All RS temperature profiles show strong inversions above the layers with maximum heating rates.These inversions are formed by absorption of SW radiation at the top of a dense aerosol layer and further mixing above this layer.The temperature inversion is quite strong: e.g., over Jeddah at 2.5 km altitude, it reaches 4.4 K and significantly affects the atmospheric and dust dynamics by preserving a dense aerosol layer near the surface.Reported heating rates from the Arabian dust storm are consistent with the figures in the literature for other locations.Lemaître et al. (2010) reported average dust heating rates to be between 1.5 and 4 K day −1 depending on the altitude and location of a dust event observed over Benin and Niger, which is comparable with the maximum heating rate over Kuwait (∼ 4.5 K day −1 ) in our study.The instantaneous heating rates they observed were as high as 8 K day −1 in some limited areas, which is comparable with the strong heating of 9 K day −1 observed at Mezaira near Abu Dhabi.
The radiative heating/cooling by dust must be taken into account to predict the overall impact of aerosols on atmospheric circulation.Karyampudi and Carlson (1988) showed that radiative heating by Saharan dust helps to maintain a warm and deep Saharan air layer over the ocean, enhancing the strength of the mid-level easterly jets and reducing the convection within the equatorial zone.Chen et al. (1994) showed that the radiative heating rates of dust could affect the evolution of a dust storm, leading to stronger surface frontogenesis.They suggested that the heating rates could significantly affect mesoscale weather systems in arid and desert regions.

Meteorological responses to the dust DRE
Figure 15 presents our simulated spatial distributions of instantaneous changes in surface air temperature and latent and sensible heat fluxes caused by the dust over land at 12:00 UTC on 19 March 2012.The surface air temperature over the majority of the study domain cools in response to the net dust radiative effect.The temperature changes by up to −6.7 K.A number of previous studies demonstrated similar magnitudes of the short-term impacts of dust on surface temperature (Han et al., 2013;Wang et al., 2010).
The change in the surface latent heat flux is quite similar to the prediction by Miller et al. (2004b).The instantaneous latent heat flux decreases over the study region with a domain average value of −1 W m −2 in the presence of dust aerosols over land.Figure 15c shows the decrease in the instantaneous sensible heat flux over land due to dust aerosols with a domain average value of −23.6 W m −2 .The daytime cooling at the surface leads to reductions in the upward transport of heat from the surface to the atmosphere.The decreases in surface latent heat and sensible heat fluxes caused by dust aerosols have been reported previously by Yue et al. (2010), Miller et al. (2004b), and Shell and Somerville (2007).

Summary and conclusions
A severe dust storm from 18 to 22 March 2012 was simulated using WRF-Chem to account for interactive dust generation and radiative effects.The model is able to simulate the major spatial features of meteorological fields in the study domain reasonably well.The synoptic systems that generated these dust events were associated with a fast-moving cold front accompanied by a high-level trough.Strong wind (velocities exceeding 15 m s −1 ) occurred behind the cold front, which entrained large quantities of dust particles into the atmosphere.The meteorological conditions that led to strong winds capable of producing the severe dust event have been identified.There were several rich dust source areas in the region, activated by the passage of the front, where the major plumes of dust storm originated.The simulation suggests that the main dust sources during the dust storm include river valleys of the lower Tigris and Euphrates in Iraq, areas from Kuwait, Iran, UAE, and the major Arabian deserts, Rub' al Khali, An Nafud, and Ad Dahna.Dust sources were also identified along the western coast of the Arabian Peninsula.
Our calculations show that 94 Mt of dust was emitted into the atmosphere during the simulation period.Approximately 78 % (73 Mt) of the dust was deposited back to the study domain and 22 % (i.e., 21 Mt) were transported outside the model domain.Approximately 6.5 Mt of dust was deposited to the oceans within the model domain.The Red Sea received 1.2 Mt of dust during this event.Dust particles bring nutrients to marine ecosystems; by scaling the effect of one storm to the number of dust storms observed annually over the Red Sea, we estimate the annual dust deposition, associated with dust storms, to be 6 Mt.
The variation in terrain elevation and surface properties produces large variability in the PBL depth over the Arabian Peninsula.The deepest PBLs developed over the higher elevations of the western Arabian Peninsula and over desert areas.The model predicted a well-mixed, deep boundary layer up to a height of about 3.5 km in the afternoon.Over land, approximately 15 % of dust was entrained in the free troposphere, while over the seas it might be more than 50 %.This is extremely important for understanding the dynamics and long-range transport of dust plumes.
The modeled AOD compares well with AERONET at the three sites and with satellite data.The simulation shows that mineral dust heats the lower atmosphere (1-3 km) by SW absorption with a maximum rate up to 9 K day −1 over Mezaira at 12:00 UTC on 19 March.The lower atmospheric heating is also reflected by the temperature inversions observed in the RS temperature profiles Figure 1.Geopotential height (m) at 500 hPa at 00:00 UTC from WRF-Chem simulations on (a) 17 March, (b) 18 March, and (c) 19 March 2012 and from ERA-Interim reanalysis on (d) 17 March, (e) 18 March, and (f) 19 March 2012.

Figure 2 .
Figure 2. Sea-level pressure (hPa) with H and L markers indicating low and high pressure systems at 00:00 UTC from WRF-Chem simulations on (a) 17 March, (b) 18 March, and (c) 19 March 2012 and from ERA-I reanalysis on (d) 17 March, (e) 18 March, and (f) 19 March 2012.

Figure 7 .
Figure 7. Simulated daily mean domain integrated (a) dust emission (black) and deposition (red) (Mt day −1 ) and (b) dust deposition (Mt day −1 ) over all ocean areas within the domain (black), over the Arabian Sea (red), and over the Red Sea (blue) as a function of time.

Figure 11 .
Figure 11.Domain-averaged dust loading (g m −2 ) above (solid line) and below (dash line) daily maximum PBL heights for land (red) and ocean (blue) areas as a function of time for the entire period of simulations.

Figure 12 .
Figure 12.(a) Simulated (dashed lines) and RS profiles of potential temperature and water vapor mixing ratio (g kg −1 ) at the Jeddah's King Abdulaziz International Airport at 00:00 and 12:00 UTC on 19 March 2012; (b) vertical profiles of wind direction (degrees) (blue and green) and wind speed (m s −1 ) (red and black) from radiosonde soundings at the Jeddah's King Abdulaziz International Airport at 00:00 and 12:00 UTC on 19 March 2012.

Figure 13 .
Figure 13.Simulated daily mean clear sky direct radiative effect (DRE) of dust aerosols (W m −2 ) on 19 March 2012: (a) SW DRE at BOA, (b) SW DRE at TOA, (c) LW DRE at BOA, (d) LW DRE at TOA, (e) net DRE at BOA and (f) net DRE at TOA. Positive values correspond to the heating of the atmospheric column and underlying surface.

Figure 14 .
Figure 14.Simulated (dashed lines) solar heating rates profiles in the atmosphere over Jeddah (black), Abu Dhabi (red), and Kuwait (green) sites and radiosounding temperature profiles at the same locations (solid lines) at 12:00 UTC on 19 March 2012.

Figure 15 .
Figure 15.Simulated instantaneous differences in (a) air temperature (K), (b) latent heat flux ( W m −2 ), and (c) sensible heat flux ( W m −2 ) in the runs with and without dust aerosols at 12:00 UTC on 19 March 2012.
. The meteorological responses to the dust radiative effects indicate a cooling at the surface over the high dust concentration regions.The daily domain average values of the net dust DRE at the surface and TOA under clear sky conditions at the time of the dust storm on 19 March are −10 W m −2 and −4 W m −2 , respectively.The SW DRE is negative at the surface over the domain with a largest daily average value of −134 W m −2 and a domain-averaged value of −16 W m −2 .The daily average SW DRE at TOA is positive (∼ 10 W m −2 ) over high albedo deserts and negative over oceans due to the smaller surface albedo.The maximum daily averaged LW DRE at the surface is 43 W m −2 with a domain average value of 6 W m −2 .At TOA, the maximum daily average value of LW DRE is 8.40 W m −2 with a domain-averaged value of 0.25 W m −2 .At the surface, the SW DRE always leads to cooling and LW DRE to heating.The dust radiative effects cause maximum change in the surface air temperature up to −6.70 K and with a domainaveraged value of −0.26 K at 12:00 UTC on 19 March.The change in the surface air temperature in the model is a combined effect of the SW and LW DRE.The indirect effect of mineral dust is not implemented in the model for the present study.The changes in domain average values of latent and sensible heat flux are −1 and −23.60 W m −2 , respectively.

Table 3 .
Summary of domain average values of daily mean clear sky direct radiative effect (DRE) for shortwave (SW), long-wave (LW) and net (SW + LW) radiation at top of the atmosphere (TOA), and bottom of atmosphere (BOA) on 19 March 2012.