Organization of Dust Storms and Synoptic Scale Transport of Dust by Kelvin Waves

Abstract. Based on the large scale transport of dust driven by the winds parallel to the mountains in the Harmattan, Saudi Arabian, and Bodélé Depression dust storms cases, a detailed study of the generation of Kelvin Waves and its possible role in organizing these dust storms and large scale dust transport was accomplished. For this study, observational and numerical model analyses were done in an in depth manner. For this, MERRA reanalysis datasets, WRF simulated high resolution variables, MODIS/Aqua and Terra images, EUMETSAT images, NAAPS aerosol modelling plots, and MERRA-2 dust scattering AOD modelling plots, surface observations, and rawinsonde soundings were analyzed for each of these three case studies. We found there were meso-β scale (horizontal length scale of 20–200 km) adjustment processes resulting in Kelvin waves only in the Harmattan and the Bodélé Depression cases. The Kelvin wave preceded a cold pool accompanying the air behind the large scale cold front instrumental in the major dust storm. We find that this Kelvin wave organized the major dust storm in a narrow zone parallel to the mountains before it expanded upscale (meso-α to synoptic).



Introduction
The Sahara Desert is dominated most of the year by the northeasterly wind-driven dry and hot air originating from the anticyclonic system centered over the North Atlantic Ocean (Shao, 2000). It has been shown that the transport of dust from the Sahara takes place in discrete outbreaks of several days and outbreak location and direction changes with the fluctuations of the Intertropical Convergence Zone (ITCZ) seasonally (Swap et al., 1996;Moulin et al., 1998). The 20 annual maximum dust storm frequency and intensity in the Sahara usually takes place in late winter and spring (Swap et al., 1996).
In Africa there are two preferred dust source regionsthe first comprises Algeria, Mauritania, and Morocco from which dust is transported over the Atlantic and as far west as the Barbados Islands and the second is the Chad Basin which exports dust to countries around the 25 Gulf of Guinea (Balogun, 1974). The Bodélé Depression which lies in the Chad Basin is the most intense and perpetual dust source in the world followed by the Western Sahara and has become the biggest and most persistent dust emission source on the global scale because of the availability of large amount of deflatable material and strong wind systems to facilitate the long range transport of the material (Engelstaedter et al., 2006). The emission of dust from Bodélé 30 takes place throughout the year though the peak emission occurs in winter/spring (Washington and Todd, 2005;Washington et al., 2006). These facts indicate the importance of the study of the Harmattan and Bodélé dust storms. Studies of these phenomena have been increasing since these two systems deliver a significant amount of dust equatorward, causing severe negative impacts on visibility, agriculture, and human health (e.g., causes meningitis) (Burton et al., 2013; 35 Kalu, 1979;Pérez García-Pando et al., 2014). Alharbi et al. (2013) state that dust storms in Saudi Arabia are very frequent. Washington et al. (2003) indicate that eastern and central Saudi Arabia are also areas of dust storm activity. The primary local source of dust storm activity in Saudi Arabia is the Rub Al Khali. Additional more remote sources are the Saharan Desert for Western Saudi Arabia and Iraqi deserts for the 40 northern and eastern part of Saudi Arabia (Notaro et al., 2013). These frequent dust storms in Saudi Arabia and the arid and semiarid areas around the Arabian Sea are some of the most important global dust sources (Kutiel and Furman, 2003). Hamidi et al. (2013) mentioned that the Tigris-Euphrates alluvial plain was the major dust source in the Middle East. Alharbi et al.,3 (2013) mentioned that the severe dust storm of March 10, 2009 was one of the most intense dust storms experienced in Saudi Arabia over the last two decades period.
All three severe dust storms: 1) Harmattan dust storm of March 2, 2004 in north West Africa, 2) Saudi dust storm of March 9, 2009, and 3) Bodélé Depression dust storm of December 8, 2011, that occurred in the lee of the mountains were caused by downslope wind effects and jet 5 adjustment processes (Pokharel et al. 2017a and2017b). During these events the winds were also parallel to the lee of the respective mountains (Atlas Mountains in the Harmattan case, Sarawat Mountains in the Saudi case, and Tibesti Mountains in the Bodélé case). The location of strong dynamics for these case studies is shown by the Modern Era Retrospective-Analysis for Research and Applications (MERRA) analysis data sets. Over time the wide distribution of dust 10 ablated and transported away from the mountains in each of these cases indicated that a possible role of terrain-induced waves such as Kelvin waves may be important in transporting dust from these three severe dust storms events. For this study, it was hypothesized that first organization of dust storms and spread of dust into the larger scale circulations was facilitated by Kelvin waves (the generation of Kelvin wave is illustrated by a schematic diagram in Figure 1). In this 15 study, we have pursued an evaluation of the possible role of Kelvin waves in the organization of strong dust transport events. The next section includes a literature review of Kelvin waves followed by research methodology, results and discussion, and conclusion in our analyses of dust storm genesis and dust transport.  Wang (2002) stated that a Kelvin wave is a relatively long wavelength gravity wave affected by the earth's rotation that is trapped along a lateral boundary, e.g., mountain ranges or coastlines. The generation and type of the Kelvin wave relies on the relative importance of the restoring force of gravity accompanying stable stratification, the significance of Coriolis 25 acceleration, and the nature of the physical boundary including the possible role of the proximity of the equator. To create a Kelvin wave in this study's region of interest, e.g., the Atlas Mountains in northwest Africa, the following processes (schematic Figure 1) must occur 1) early on the shallow cold pool of air is blocked by the Atlas Mountains due to lack of sufficient kinetic energy to force air parcels over the mountain, causing a buildup of mass as a height 30 perturbation in the area immediately adjacent to it, 2) after a certain period of time < than an inertial period (<2π/f where f is the Coriolis parameter) this build-up of excess mass was released as a gravity wave or buoyancy wave signal or undulation in a free surface near the terrain (Thomson, 1879), 3) as time progresses towards 2π/f the Coriolis force acts on and turns the flow towards the right in a very slow effort to achieve balance, 4) as this process continued wind flow 35 became parallel to the mountains as a distinct wind perturbation in conjunction with the aforementioned free surface undulation, and 5) in time the wind accompanying this Kelvin wave accelerates parallel to the mountain (Tilley, 1990).

Materials and Methods
To study these three dust events in detail, we have used Meteosat-8 dust image captured 40 from the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) (http://www.eumetsat.int/website/home/Images/ImageLibrary/DAT_IL_04_03_06_F.html),a composite of the Moderate Resolution Imaging Spectroradiometer (MODIS) /Aqua and Terra (level 1b, collection 51, 1 km horizontal resolution, and RGB composite)  holdings/merra/merra_products_nonjs.shtml; Rinecker et al. 2011) were analyzed. These data were used to create horizontal cross sections at different pressure levels as well as vertical cross sections of u and v wind speed, and isentropic surfaces at a resolution of 0.50º X 0.67º.
In order to get finer temporal and spatial resolution of the atmospheric processes, which were involved in the dust storms, non-hydrostatic Weather Research and Forecasting (WRF) 5 model (Skamarock et al., 2008) were run for each dust storm, given in Table 1. Four nested domains with horizontal resolutions of 54, 18, 6, and 2 km were generated, of which resolutions higher than 18 km were run with moist convection turned off as a parameterization because convection was virtually nonexistent in these cases. The lateral boundary condition data of the parent domain with the lowest resolution were from the National Center for Environmental 10 Prediction (NCEP)/Global Forecasting System (GFS) (1° X 1°) products. Three domains were then nested into the parent domain having 18, 6, and 2 km. The WRF simulated domains for the three cases are presented in Figures 3 and 9. The following six physical parameterizations, such as (i) momentum and heat fluxes at the surface (Janjić, 1996(Janjić, , 2001 following Monin-Obukhov similarity theory, (ii) turbulent mixing following the Mellor-Yamada-Janjić 1. Momentum and heat fluxes at the surface using an Eta surface layer scheme [Janjić, 1996[Janjić, , 2001] that follows Monin-Obukhov similarity theory, 2. Turbulence processes following the Mellor-Yamada-Janjić 1.5 order (level 2.5) turbulence closure model [Mellor and Yamada, 1974;Janjić, 2001], 3. Convective processes following the Betts-Miller-Janjić cumulus scheme [Betts, 1986;Betts and Miller, 1986;Janjic, 1994]applied only on the 54 and 18 km grid, 4. Cloud microphysical processes following the Thompson double-moment scheme [Thompson et al., 2004[Thompson et al., , 2006], 5. Radiative processes following the Rapid Radiative Transfer Model for long wave radiation ] and Dudhia's scheme for short wave radiation [Dudhia, 1989], and 6. Land-surface processes following the Noah land surface model (Noah LSM) [Chen and Dudhia, 2001;Ek et al., 2003].  , 1974;Janjić, 2001), (iii) moist convection following the Betts-Miller-Janjić cumulus scheme (Betts, 1986;Betts and Miller, 1986;Janjic, 1994)-only for the simulations with 54 and 18 km resolution, (iv) cloud microphysical 5 processes following the Thompson double-moment scheme (Thompson et al., 2004(Thompson et al., , 2006, (v) radiative processes following the Rapid Radiative Transfer Model for long 10 wave radiation  and Dudhia's scheme for short wave radiation (Dudhia, 1989), and (vi) land-surface processes from the Noah land surface model (Noah 15 LSM) (Chen and Dudhia, 2001;Ek et al., 2003) were applied in WRF simulations in each of three cases.

Results and discussion
3.1 Harmattan dust storm 20 case study

Observational and model analyses
Since Bechar lies to the east of Atlas Mountain and is the nearest sounding station to the Atlas Mountains, it is logical to expect that the information regarding the vertical profiles of meteorological variables in the lee of the mountains could be well represented in the analysis of 25 before and after the Kelvin wave formation. The sounding of 0000 UTC March 2 shows that there was a thin inversion between 925-900 hPa ( Figure 4a). Above this a dry adiabatic lapse rate between 900-585 hPa was present; again above this deep dry adiabatic layer, there was a presence of the shallow inversion layer, which was caused by a warm stable thermal ridge. These different kinds of vertical temperature lapse rates indicate the presence of the discontinuous 30 stratification of the atmosphere over this area at that time and also shows the necessity of higher resolution data sets to reveal the detailed processes involved.
Surface archived observational data sets of Weather Underground at Tindouf (27°40′31″N 8°07′43″W) in Algeria showed that there was a major dust storm that occurred during the latter parts of March 2, 2004. This dust storm was associated with a surface north-northeasterly wind 35 with a speed of 10 m/s after 2200 UTC on March 2, respectively. Besides this, Bechar (31°37′N 2°13′W, northeast of Tindouf), Adrar (27°52′N 0°17′W, east of Tindouf), and Timimoun (29°15′46″N 0°14′20″W, east of Tindouf) experienced reduced visibility consistent with increased atmospheric optical depth and dust from 1300 to 2200 UTC, 1400 to 2300 UTC, and 1300-1600 UTC on March 2, respectively (Pokharel et al., 2017b;Pokharel, 2016)   10 the trough at the meso-α/synoptic scale (between 200 km and 2000 km). Given the location of trough amplification of the exit region of the jet streak which was becoming proximate to the stable layers in Figures 4a and 4b, i.e., proximate to mountain-induced thermal perturbation on the lee of the Atlas Mountains resulting in a mass perturbation (It is to be noted here that Figure  4a is observed sounding while 4b sounding is the product of WRF model. As both Figures are 5 consistent with each other in regards to vertical profiles of temperature and wind WRF model outputs can be said to be validated). This interaction of the perturbed mass field ahead of the jet and the exit region of the jet leads to a breakdown of geostrophic balance. At 1800 UTC and afterwards of March 2, the temperature and wind speed pattern at 925 hPa show the possible evolution and propagation of the Kelvin wave as it was propagating parallel to the Atlas 10 Mountains in the southwestward direction ahead of the cold pool accompanying the large scale cold front (Figure 6a). All these processes will be analyzed with high resolution and timecontinuous WRF numerical simulations.

3.1.2
WRF simulation analyses 15 Before entering in our specific analysis we would like to introduce a study of the jet adjustment processes, which are precursors of this large scale dust storm, discussed in recently published by Pokharel et al. (2017b). This jet adjustment processes were also one of the precursors for the generation of Kelvin waves. Pokharel et al. (2016Pokharel et al. ( , 2017b and Pokharel (2016) clearly show that there was an interaction of the exit region of the polar jet streak with a 20 local thermally perturbed air mass on the leeward side of the Atlas Mountains, and summarizes the detail processes after this interaction as follows: 1) the generation of a jetlet at 0900 UTC on March 2, 2004 in the leeward side of the Atlas over the 30-32°N 6-2°W region ; 2) an occurrence of the mass field adjustment processes modifying the wind field until it reaches to a new geostrophic balance; 3) in this adjustment process, a thermally direct transverse ageostrophic 25 circulation in the exit region of the jetlet developed downstream from the mountain leading to the upward motion and the formation of the cold pool under the right exit of the jetlet (where velocity divergence exists; and 4) this cold pool led to the rise of the low-level high pressure perturbation creating an ageostrophic/isallobaric wind as a return branch of the direct circulation of the exit region of the jetlet at the lower levels, i.e., 925 hPa. 30 Following this finding, we analyzed our WRF simulated data sets further to see if there was any additional signal generated 5 prior to or during the adjustment processes as discussed by Pokharel et al. (2017b) for the generation of the strong winds at the 10 lower levels. We see that there was an additional meso-β scale mass field adjustment process shown by the geopotential height  (Pokharel, 2016). The shading illustrates the orographic heights in meters. b. Enlargement of geopotential height in steps of 50 m (red contours), wind speed/direction (vectors), and axis of evolving ageostrophic wind (big arrow) at the 500 hPa level at 0000 UTC on March 2, 2004 from MERRA re-analysis with horizontal resolution of 54 km (Pokharel et al., 2017b). The shading illustrates the orographic heights in meter. A triangle of red (inside) and yellow (outside) shows the location of Bechar where the soundings were recorded.
(b) Figures 1a and 1b). This mass buildup by the jet adjustment process in the lower levels resulted in wind flow parallel to the mountains at 925 hPa by the ageostrophic isallobaric wind as discussed by Pokharel et al. (2017b). This indicated the possible evolution of the Kelvin wave (Thomson, 1879;Wang, 2002;Tilley, 1990) on the south to northeast edge of the lee of the Atlas (29-32°N 5°W-2°E) at 1100 UTC on March 2 (Figures 7c and 7d ) (consistent with schematic 5 Figures 1c, 1d, and 1e).  Figure 1f). Afterwards, when this cold northeasterly wind, which had already significant momentum, interacted with the warm air 40 column of stretched isentropes, which were present south/southwestward of the leeside of the Atlas Mountains (Figure 7g) as discussed above, there was a large production of the dust over the region on the south/southwest side of the Atlas (Figure 7g) (consistent with schematic Figure 1f). This Kelvin wave started to organize the Pokharel et al. (2017a). The Kelvin wave subsequently intensified the dust storm resulting in its growth upscale by first concentrating its energy in a narrow zone adjacent to the lee of the mountains in the afternoon on March 2 before it expanded over time as a suspension of dust in the atmosphere by the growing wind perturbation orthogonal to the Kelvin wave. It is also seen that the northeasterly wind accompanying the Kelvin wave expanded further with the arrival of 5 the Q-G cold pool from the region north/northeast of the Atlas Mountains parallelly (Figures 6b  and 6c. Figure 6b is consistent with Figure 6a, which is a product of observational data set as mentioned in the subsection 3.1.1). In other words, Figures 6a, 6b and 6c indicate the wind was responding to the Kelvin wave mass field as one wind component was going west-southwest close to the mountains as it turns to the right in response to the Coriolis force and another 10 component was turning to the left away from the mountains farther north and east as it goes southsouthwestwards (shown in 15 two red circled areas).
The presence of the low static stability on the south/southwestward sections of the Atlas This is consistent with the sequential dust storms at Tindouf, Bechar, and Adrar during this time period as discussed earlier in the sub-section 3.1.1. Over time the WRF simulations and MERRA observations support the fact that the Kelvin wave propagated ahead of the cold pool 45 accompanying the air behind the large scale cold front (Figures 6a, 6b, and 6c). This led the (c) 14 generation of the large volume of dust from the large areas on the leeward side (south and southwest of the Atlas, 25-31°N 10°W-2°E) of the Atlas at around 1800 UTC on March 2 and afterwards. This is also consistent with the dust storm at Tindouf at this time period. The presence of the area of warm air column on the south and southwestern edge of the Atlas was also supported by the presence of the strong temperature gradient at the edge of this cold pool 5 consistent with the studies by Parmenter (1976); Garreaud and Wallace (1998);Liebmann et al. (1999). Here, an evolution of the Kelvin wave process as discussed earlier is consistent with Tilley (1990), which states that Kelvin wave is the propagation of parallel wind with the boundary on its right in the Northern hemisphere. The generation of Kelvin wave occurs when there is a shallow barotropic layer of air column (fluid) impinging on a physical barrier with a 10 vertical height > fluid depth. In such a case, the flow of this air column is blocked by the barrier creating a buildup mass in the area immediately adjacent to the barrier. After that, this excess mass is released as gravity waves and once the fluid has been set in motion the Coriolis force begins to act upon it, turning the air flow towards its right. Over time this process continues and the air flow as a Kelvin wave becomes directed parallel to and is right-bounded by the barrier.  30 The positively tilted trough slightly deepened until the occurrence of the dust emission. Although the strength of the positively tilted trough was weak compared to the other cases, falling 35 heights still revealed the upper-air disturbance over the northeast region of Niger, including north and northwestern regions of Chad which included Aozou, and southeast of Libya. From 0900 40 UTC 8 December to 0900 UTC 9 December this trough propagated southeastward and this propagation of the trough along with the jet stream were both associated with quasi-45 (a) Figure 8 a. Bodélé Depression dust storm image captured by MODIS/Terra at 0950 UTC on December 9, 2011(source: https://ladsweb.nascom.nasa.gov). The red star indicates surface station at Ndjamena in Chad which captured the dust storms from 0600 UTC on December 8 to December 9, 2011 (source: wunderground.com) (Pokharel et al., 2017b). The x and y coordinates of this station are mentioned in observational and model subsection of 3.2.1 of Bodélé Depression dust storm case study.
(b) 20 geostrophic lifting ahead of the trough and sinking behind it. A jet at 500 hPa was over the Tibesti Mountains and two additional jets were north of Chad from 0000 UTC 8 December. The jet, which was over the Tibesti, was coupled to the trough until the occurrence of this dust storm. On the other hand, the Kelvin wave seemed to be evolving as it was flowing parallel to the Tibesti ahead of the large scale cold pool accompanying the large scale cold front at 0800 UTC 5 and afterwards shown by the temperature and wind speed pattern at the 850 hPa level at that time (not shown). So, to analyze all these aforesaid conditions, we again use high-resolution WRF simulations for the detailed dynamical development of a Kelvin Wave in this severe dust storm. 10

WRF simulation analyses
As mentioned above in the Harmattan case, Pokharel et al. (2017b) states in this case also that there was an interaction of the subtropical jet with the perturbed warm air mass on the leeward side (south/southwest/southeast) of the Tibesti Mountains that led to the different processes (i.e. establishment of a meso-β scale adjustment process) as mentioned in the above 15 Harmattan case. In addition to this meso-β scale adjustment process in the lower levels, there was an additional interesting meso-β scale feature shown in the model output. At and after 0700 UTC on December 8 of 2011, as discussed earlier in the Harmattan case we also 20 see that there was an additional meso-β scale mass field adjustment process shown by the geopotential height rise at 925 hPa ( not shown). 25 Similarly, the temperature pattern at 925 hPa shows that there was a cold pool over the 19-30°N 10-30°E region. Within this wide cold region, there was a comparatively colder region of air at the lower level than at upper 30 levels in terms of the structure of the isentropic surfaces in the east/northeast side of the Tibesti Mountains (21-23°N 19-21°E) ( Figure  11a) (consistent with schematic Figure 1a). This cold column indicates the presence of the 35 stability of the atmosphere at this level. This was the result of the presence of the blockage of the cold air column by the mountain range and generation of the initial mass impulse (Figures 11b and 11c) (consistent with schematic Figures  1a and 1b). From this buildup of mass by the jet adjustment process in the lower levels as discussed earlier, there was wind flow parallel to the Tibesti Mountains and equatorward at 925 40 hPa ( Figure 10). After flowing parallel to the Tibesti Mountains this equatorward-directed wind rotated anti-cyclonically or southwestward along the east slope of the Tibesti Mountains as a northeasterly wind which strengthened in the time. When the wind flow at this low level stable region became parallel to these mountains, it can be inferred that there was a generation of the Kelvin wave (Thomson, 1879;Tilley, 1990) (Figures 10, 11c, and 11d) (consistent with 45 schematic Figures 1c, 1d, and 1e). There was a close association between the subtropical jet 54 km 18 km 6 km 2 km Figure 9. WRF domain configuration for the Bodélé Depression dust storms dust storm case shown in Figure  11 (Pokharel, 2016). do1, do2, do3, and do4 represent domains of 54, 18, 6, and 2 km resolution, respectively. streak imbalance (Pokharel et al., 2017b) and the Kelvin wave formation due to the presence of the inversion layer in the lee of the Tibesti Mountains. The presence of the cold pool at the bottom of the lee slope led to the pressure rise resulting in the wind flow parallel to the Tibesti Mountains during the Kelvin wave formation at 925 hPa. This process was also supported by the generation of compensating jet generated by the low level pressure rise and its 5 ageostrophic/isallobaric wind flow as discussed in the earlier case study. This Kelvin wave, which was trapped along lateral vertical mountain boundaries, was the result of the mass perturbations propagating parallel to the mountains barrier as discussed by Tilley (1990) and was also representative of higher frequency form of adjustment process that might have played a vital role in generating this type of particular dust storm processes. It is important to note that when  where − . ∇ is TKE advection by the mean wind, * 2 ( / ) is generated shear, ( / ) 10 is buoyancy, and is the dissipation of TKE or eddy dissipation rate. This enhancement of the TKE developed a turbulent well-mixed circulation, which ablated the dust from the surface. At 0900 UTC and onwards, the Kelvin wave mentioned in the above paragraph propagated ahead of the cold pool accompanying the northeasterly wind behind the large scale cold front from the north/northeast side of the Tibesti Mountains towards the south/southwest region of 15 Chad, where the deep convective environment existed. The presence of the convective turbulent environment was consistent with the expansion of the isentropic surfaces ( Figure 11e) and the strong gradient of the temperature at the edge of the cold pool at 850 and 925 hPa as suggested by Parameter (1976); Garneau and Wallace (1998); and Liebmann et al. (1999). The interaction of this cold pool accompanying the northeasterly wind and the highly warm air column created 20 significant magnitudes of TKE ( Figure 12) accompanying the generation of the turbulent eddies, which ablated the large volume of the diatomite dust from the surface, which is also consistent with Pokharel and Kaplan (2017). The surge of the cold pool over this region followed the early enhancement of the warm air in the south/southwestward of the Tibesti is consistent with the description in Vizy and Cook (2009

Conclusions
Besides the downslope wind effects and the jet adjustment processes which were responsible to cause these three dust storms (Pokharel et al., 2017a(Pokharel et al., , 2017bPokharel, 2016) we found that the organization of the dust storms and the synoptic scale transport of dust from the Harmattan and Bodélé Depression dust storms were also seen to be caused by the 5 Kelvin waves. Though we also carried out a detail analyses of Saudi dust storm case with the help of observational and WRF model data sets as we did for Harmattan and Bodélé Depression dust storm cases, for the convenience of readers we have not included the detail analyses of Saudi case in this manuscript since we do not have unambiguous signals of Kelvin waves formation in this case. It is also to be understood that all validation of the Saudi case indicated 10 that the simulation was accurate but the dynamics not necessarily the same as the other two case studies.
The generation of Kelvin waves in Harmattan and Bodélé Depression dust storms were facilitated by the jet adjustment processes in the lee of the mountains and topographic blocking of the flow consistent with the lack of the sufficient kinetic energy of the air parcels to cross the 15 mountains. This lack of kinetic energy to cross the mountains led to the 1) buildup of mass in the area immediately adjacent to respective mountains (consistent with schematic Figures 1a and 1b) , 2) release of excess mass from this build-up mass after certain period of time (consistent with schematic Figure 1c) , 3) effects of Coriolis force on turning of the flow towards the right to achieve a balance (consistent with schematic Figure 1d), 4) wind flow accompanying Kelvin 20 waves parallel to the mountains (consistent with schematic Figure 1e), 5) interaction of this wind accompanying Kelvin waves with the highly warm air column created significant magnitudes of TKE resulting in dust from the surface(consistent with schematic Figure 1f). This result is consistent with our generation of Kelvin waves hypotheses shown in schematic Figure 1 and also as discussed in the literature reviews of this manuscript. This result contributes to a better 25 understanding of how this type of large scale dust transport can be organized from this region to the U.S., Amazon basin, and Europe (which are shown by numerous previous studies) and might be due in part to Kelvin waves. We also think there is still a need to study major historical dust events that occurred in this region to confirm that these locations are suitable and are responsible for the generation of the Kelvin waves and the transport the dust on a regular basis. 30

Author contributions
Dr. Ashok Kumar Pokharel collected relevant data, performed the WRF model simulations and prepared this manuscript. Dr. Michael Kaplan edited and provided feedback on it.