Integrated Analysis of Dust Transport and Budget in a Severe Asian Dust Event

Asian dust storms markedly affect the ecosystem, environment, ocean biogeochemical cycle, and regional climate. Numerous measurements and model simulations have been performed to investigate the sources and transport of Asian dust. However, until now, few studies have performed a comprehensive quantification of the dust budget, resulting in significant uncertainty about the characterization of dust transport, emission, and deposition. In this study, a severe dust event in East Asia that occurred from April 28 till May 3, 2011, was analyzed in terms of dust transport characteristics based on multi-satellite observations and the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model. In particular, the dust budget of the event was quantitatively estimated using a new atmospheric reanalysis dataset, namely the second Modern-Era Retrospective Analysis for Research and Applications (MERRA-2). The multi-satellite observations and models indicated that dust events such as this are uncommon. A two-layered dust structure was found in southeast China, the lower (< 1.5 km) and elevated (> 3 km) layers of which mainly originated from the Gobi and Taklimakan Deserts, respectively. The dust budget in East Asia, as estimated from MERRA-2, revealed a high dust mass loading (5.7–6.6 Tg) between 70°E and 140°E from April 29 till May 1, with the highest daily dust loading (approximately 6.6 Tg) reported on April 30. The total dust emission was 6.3 Tg over a 6-day period (April 28–May 3), and the maximum amount (nearly 5.9 Tg) of dust was deposited on the ground in the region. The dust flux amounts horizontally transported across the longitudinal boundary of 70°E and 140°E were 1.7 and 2.8 Tg, respectively.


INTRODUCTION
Asian deserts are crucial sources of mineral dust aerosols, which can affect the climate and environment by altering the radiative energy balance, cloud microphysical properties, and biogeochemical cycles (Wang et al., 2011;Narasimhan and Satheesh, 2013;Huang et al., 2014;Knippertz and Stuut, 2014).Some studies have reported that Asian dust aerosols can be transported to vast downwind areas, including the coastal seas of China, Korea, Japan, and even North America (Fang et al., 1999;Mori et al., 2002;Zhao et al., 2011;Cottle et al., 2013;Li et al., 2015).These transported aerosols are often mixed with high concentrations of regional pollution aerosols that largely reduce visibility and worsen air quality in downwind areas of East Asia (Guo et al., 2013).Moreover, the long-range transport and deposition of dust containing nutrients such as iron, nitrogen, and phosphorus into the oceans promote phytoplankton growth and result in feedback effects on the Earth's climate and ecosystem (Shi et al., 2012;Tan et al., 2014Tan et al., , 2016)).Therefore, understanding dust transport and budget in the Asian dust event that occurred from April 28 to May 3, 2011, in East Asia is imperative for evaluating their effects on air quality and ecosystem conservation (Niu et al., 2016).
Many studies have used satellites, ground-based measurements, and model data to investigate the long-range transport characteristics of the Asian dust event and their effects on air quality and the environment in East Asia (Liu et al., 2011;Chen et al., 2015;El-Askary et al., 2015;Li et al., 2015).However, the complexity of the transport and sources of dust aerosol remain great challenges in determining its effects on local air pollution and the global climate system (Prasad et al., 2010;Tan et al., 2016).Up to date, quantitative understanding of individual dust events is also still incomplete (Husar et al., 2001;Liu et al., 2003).In particular, the dust budget, which is crucial for quantitatively evaluating air pollution and biogeochemical effects (Knippertz and Stuut, 2014), has not been comprehensively investigated.Some researchers have estimated dust mass fluxes based on satellite observations (Kaufman et al., 2005;Yu et al., 2008).However, such estimates contain large uncertainties due to other uncertainties associated with the derived dust optical depth and mass extinction efficiency (MEE) (Kaufman et al., 2005;Yu et al., 2009Yu et al., , 2013)).Moreover, although model simulation is a useful tool for estimating dust mass (Tan et al., 2016), the dust budget is often underconstrained in dust models, resulting in large uncertainties in dust emission and deposition (Knippertz and Stuut, 2014).Conversely, these uncertainties hinder the comprehensive evaluation of the biogeochemical effects of dust (Knippertz and Stuut, 2014) and the effect of dust on air quality.Thus, accurately estimating the amount of dust emission, loading, and deposition is imperative for the quantitative determination of dust cycle effects.
A data assimilation tool provides a good opportunity to improve dust simulation by using available observations with high spatial and temporal resolutions (Dai et al., 2014;Buchard et al., 2016;Di et al., 2017) to potentially yield more accurate dust estimates than those obtained using either a model or observations alone.Di et al. (2013) and Dai et al. (2014) have reported that the assimilation of moderate resolution imaging spectroradiometer (MODIS) aerosol optical depth (AOD) data into a model can improve dust simulation relative to using observations alone.Yumimoto et al. (2016) asserted that the assimilation of aerosol data from Himawari-8, a next-generation geostationary meteorological satellite, is promising for future simulations of anthropogenic air pollution and natural dust outflow.Considering the advantages of data assimilation, we used the second Modern-Era Retrospective Analysis for Research and Applications (MERRA-2), a new atmospheric reanalysis dataset developed by NASA's Goddard Space Flight Center Global Modeling and Assimilation Office (GMAO).MERRA-2 contains the assimilation of bias-corrected AOD from the Advanced Very High Resolution Radiometer (AVHRR) and MODIS, AOD over bright surfaces obtained from the Multiangle Imaging SpectroRadiometer (MISR), and AOD from the AErosol RObotic NETwork (AERONET) (Randles et al., 2016).In our study, MERRA-2 was used to improve the quantitative understanding of the dust budget (e.g., dust emission and deposition) in a severe dust storm that occurred from April 28 to May 3, 2011, in East Asia.Han et al. (2015) have reported the impacts of this dust event on air quality in NanJing of China.However, the present study is the first to quantitatively analyze the dust budget of this event in East Asia.Moreover, the transport characteristics of this dust event were analyzed using integrated multisatellite observations such as measurements from the MODIS and Ozone Monitoring Instrument (OMI) and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO), as well as the hybrid singleparticle Lagrangian integrated trajectory (HYSPLIT) model and wind field data from National Centers for Environmental Prediction (NCEP) reanalysis, to comprehensively understand the dust transport process and delineate the effects of the Asian dust event on local air quality.
Overall, this study focused on the following objectives: (1) presenting the dust formation process and long-range dust transport characteristics based on the dust event that occurred from April 28 to May 3, 2011, in East Asia by using multisatellite observations and model data; (2) analyzing the dust source and vertical structure of this event; and (3) quantitatively estimating the dust budget from the emission, deposition, and transport in the event according to MERRA-2.The remainder of this paper is organized as follows.Section 2 describes the data and models used in this study.The dust formation process, transport characteristics, and budget analysis are presented in Section 3. The main conclusion is summarized in Section 4.

Satellite Observations
To identify the Asian dust event of April 28-May 3, 2011, a Chinese polar-orbiting meteorological satellite, Feng Yun-3A (FY-3A), which has sounding capabilities and natural color imagery with a high spatial resolution of 250 m (Dong et al., 2009), was used to obtain a direct view of the dust outbreak based on its true color composite images from channels 1 (640 nm), 9 (550 nm), and 7 (470 nm).
The MODIS equipped with the Terra and Aqua satellites, which observe the Earth over land (Kaufman et al., 1997) and ocean (Tanré et al., 1997), provides daily operational retrievals of global aerosol optical parameters.In this study, to understand the transport characteristics of the dust event, the AOD was used to determine aerosol optical properties along the transport pathway.The MODIS uses the Dark Target (DT) and Deep Blue (DB) algorithms for aerosol retrieval (Kaufman et al., 1997;Hsu et al., 2004).The DT algorithm is used over dark surfaces such as densely vegetated surfaces based on the linear relationship of surface reflectance between shortwave infrared (2.12 µm) and visible (0.47 and 0.65 µm) bands (Kaufman et al., 1997).The DB algorithm is applied to retrieve the aerosol parameters of bright surfaces by using blue (0.412 and 0.47 µm) channels (Hsu et al., 2004).Collection 6 (C6) of MODIS aerosol products contains merged AOD data combining DB and DT retrievals over land and ocean (Levy et al., 2013;Sayer et al., 2013;Tao et al., 2015).In this study, the Aqua MODIS C6 level 3 merged AOD at 1° × 1° resolution was used.
The OMI onboard the Aura satellite is a high-resolution spectrograph that measures the upwelling radiance at the top of the atmosphere in ultraviolet and visible (270-500 nm) spectra (Torres et al., 2007).The OMI aerosol products include the ultraviolet aerosol index (UVAI), aerosol absorption optical depth (AAOD), and aerosol extinction optical depth (AEOD).The AAOD in the near ultraviolet (UV) is more sensitive to boundary layer aerosols than the UVAI (Ahn et al., 2008).Moreover, the AAOD is less heavily affected by clouds than the AEOD because clouds enhance scattering but not absorption (Ahn et al., 2008).Therefore, level 3 AAOD data at 388 nm from the OMI were collected in this study to detect and analyze absorbing aerosols such as desert dust.
The CALIPSO satellite, launched in 2006, provides global vertical structures and optical properties of aerosol and cloud layers at 532 and 1064 nm, respectively (Winker et al., 2007;Omar et al., 2009).In this study, we used the level 1 attenuated backscatter and aerosol subtype classification products to obtain information on the vertical structure of the dust layers.

Model and Reanalysis Data
MERRA-2 replaced the original MERRA by using an upgraded version of the Goddard Earth Observing System Model, Version 5 data assimilation system produced by the GMAO (Randles et al., 2016).It places observations from NASA's Earth Observing System (EOS) satellites into climate contexts, thereby updating the MERRA system to include the most recent satellite data (Rienecker et al., 2011;Randles et al., 2016).The original MERRA assimilated only meteorological parameters, including wind and temperature (Randles et al., 2016).By contrast, MERRA-2 performs aerosol and meteorological assimilations simultaneously.Furthermore, MERRA-2 incorporates AOD measurements from various Polar Operational Environmental Satellites of the National Oceanic and Atmospheric Administration (NOAA), as well as EOS platforms and ground-based observations from NASA (Randles et al., 2016), which provide considerably more accurate estimates of dust transport, emission, and deposition than do single models and observations (Buchard et al., 2016).The annual variability range of dust emissions determined by MERRA-2 was 1900-2400 Tg yr −1 (Colarco et al., 2014).The difference between the high and low mean emissions was approximately 5% (Colarco et al., 2014).The global annual mean dust loading of MERRA-2 was 20.8-23.1 Tg.Moreover, the global annual mean dust emission and loadings were consistent with the multimodel AeroCom Phase I median emissions and loadings (Textor et al., 2006;Huneeus et al., 2011;Colarco et al., 2014).Thus, MERRA-2 was observed to be credible and was used to estimate the dust budget in this study.Until now, the available dust products from MERRA-2 include the dust column density, emission, and deposition with 0.625° × 0.5° resolution.
The dust loadings can be estimated using the following equation: where DL indicates dust loadings, D mean is the estimated mean dust column density based on MERRA-2 for a certain area in East Asia, and S is the area for estimated dust loading.The total dust emission and deposition (DAE tot ) can be estimated using the following equation (Han et al., 2016): where F mean is the estimated mean dust emission or deposition based on MERRA-2 for a certain area in East Asia, T is the time period for estimated dust emission or deposition, and S is the area for estimated dust loading.
The HYSPLIT model was used in this study to calculate backward trajectories at different heights for tracking the transport pathway during the dust event (Draxler et al., 1998).The model calculates the trajectories by using the Global Data Assimilation System meteorological dataset at a 1° × 1° resolution.Wind field data from the NCEP reanalysis project was also considered to analyze variations in meteorological conditions during dust transport.Table 1 provides detailed data.

Long-Range Transport Characteristics of Asian Dust
A severe dust storm event occurred from April 28 to May 3, 2011.The storm originated from the Taklimakan Desert in west China and the Gobi Desert in Mongolia and northwest China.The FY-3A true color composite images in Fig. 1 present a direct view of this event.A strong dust storm was clearly observed over both deserts on April 29.The storm started moving toward northeast China on April 30.
The MODIS-retrieved AOD can measure the global aerosol scattering or absorption effect and is often used to analyze variations in dust aerosol loading (Tan et al., 2014;Li et al., 2015).While OMI-retrieved AAOD in the near ultraviolet (388 nm) only reflects aerosol absorption.But it is also useful to track absorbing aerosol such as dust and smoke aerosols, which often interact with urban and industrial pollutants and other aerosol types during long-range transport (Ahn et al., 2008;El-Askary et al., 2015).A high AAOD (> 0.1) indicates dust or other absorbing aerosols.To analyze the dust transport characteristics of the Asian dust event, AOD products from MODIS-Aqua in addition to wind fields from NCEP reanalysis data and the AAOD from OMI at 388 nm were used in this study.Fig. 2(b) shows high AODs (> 1.6) in the Gobi and Taklimakan Deserts on April 29, thereby implying the existence of dust plumes in these regions.Significant absorbing aerosols (AAOD) as high as 0.14-0.22[Fig.3(b)] were also observed in these regions on the same day.This sharp increase in absorbance in the near ultraviolet range is likely a result of the mixture of dust with urban and industrial pollutants and smoke As shown in Fig. 2, wind fields at 850 hPa (≈ 1.5 km) revealed strong northwesterly winds in central Mongolia and northeast China.The winds were associated with the development of a Mongolian cyclone.Wang and Fang (2006) asserted that prevailing winds play a major role in the longrange transport of dust aerosols and favor dust formation and movement.Furthermore, high AODs [Figs.2(c)-2(f)] were observed and gradually distributed toward northeast and then southeast China on April 30 to May 3. Variations in high AAODs [Figs.3(c)-2(f)] showed strong agreement with those retrieved from the MODIS.These results reveal that the dust plumes from the Gobi Desert were gradually transported to downwind regions such as northeast, central, and southeast China, and even to the Korean peninsula from April 29 to May 3.During this transport process, the dust aerosols in the deserts mixed with anthropogenic pollutants and influenced the local air quality in the downwind areas.
As shown in Figs. 2 and 3, in contrast to the dust aerosols in the Gobi Desert, those in the Taklimakan Desert seemed to not be directly transported to southwest or southeast China, likely because of low wind speed and ambient terrains such as the Loess Plateau and Tibetan Plateau.However, passive satellite observations such as those from the MODIS and OMI can only provide the horizontal transport characteristics of dust aerosols during transport processes; they cannot determine the vertical distribution of the dust layer.Dust transport from the Taklimakan Desert is further investigated and identified in this study.

Dust Sources and Vertical Structure from HYSPLIT Model and CALIPSO Observations
In this paper, we report horizontal dust transport characteristics based on satellite observations of the Asian dust event.However, vertical dust transport could not be investigated in detail.To further investigate the dust source and layer altitude, the HYSPLIT model and CALIPSO active satellites were used to track the backward trajectory of dust masses and identify the vertical structure of the dust layers respectively.Fig. 4 shows the backward trajectories of dust masses in Beijing and Hangzhou along the dust transport pathway for 3 days at 6-hour intervals.Backward trajectories of dust masses in Beijing [Fig.4(a)] show that the dust at 500, 1500, and 3000 m mainly originated from the Gobi Desert, a finding in concordance with the MODIS and OMI observations.The vertical observations from the CALIPSO lidar (Fig. 5) show that the topmost dust layer was 8-10 km high in the analyzed dust event.The dust layer altitudes around Beijing were less than 5 km on April 29 [Figs. 5(c) and 5     from the Gobi Desert.The elevated dust plumes (> 3 km) in these areas mainly originated from the Taklimakan Desert and traveled across the Hexi Corridor and Loess Plateau through long-range transport to reach southeast China.These results indicate that the analyzed event was severe and uncommon in terms of two-layered dust structures found in southeast China.Moreover, the CALIPSO observations (Fig. 5) revealed that the dust plumes were mixed with other aerosols.

Dust Mass Budget Analysis
More crucial than the dust transport characteristics and vertical structures were the dust budget of the event analyzed in this study.Compared with available observations and models alone, MERRA-2 potentially provides more accurate dust estimations because of the assimilation of various aerosol observations (Buchard et al., 2016;Randles et al., 2016).Therefore, MERRA-2 offers many advantages for estimating and analyzing dust loadings, emission, and deposition in this severe dust event.In the dust budget analysis, the latitudinal range was set at 0°N-60°N.The dust mass loading was estimated using Eq. ( 1).Here, between 70°E and 140°E, the loading area was approximately 4.3 × 10 13 m 2 .
In this study, before estimating the dust mass loading, the spatiotemporal distribution of the dust column density in East Asia was determined using MERRA-2.As presented in Fig. 6, the spatial distribution of the dust column density showed some differences to that of the MODIS-retrieved AOD and OMI-retrieved AAOD, likely due to the differences in data sources.Briefly, the AOD from MODIS is retrieved from the instantaneous observation of satellites, whereas the dust column density from MERRA-2 is obtained through a 24-hour average model simulation.Nevertheless, Fig. 6 reveals an extremely high dust column density in the Taklimakan and Gobi Deserts on April 29 and that dust plumes gradually moved toward southeast China.This general trend of dust transport is consistent with the MODIS and OMI observations.Furthermore, the daily dust mass loadings were estimated between 70°E and 140°E during the analyzed event based on the dust column density of MERRA-2.As shown in Fig. 7, the dust mass loadings (5.7-6.6 Tg) were high from April 29 to May 1.The highest dust loading (6.6 Tg) was recorded on April 30 and was mainly due to the effect of a strong northwesterly wind and large-scale dust transport that occurred during the event, particularly on April 30.Moreover, the average dust loading between 70°E and 140°E was 5.2 Tg during this event.
The total dust emission and deposition during the event were estimated with Eq. ( 2) by using MERRA-2.Fig. 8 shows the amount of dust transport, emission, and deposition in this severe Asian dust event.The dust budget analysis revealed a total dust emission of 6.3 Tg between 70°E and 140°E during the 6-day storm and the deposited amount of dust on the ground was approximately 5.9 Tg in the region.The dust flux amounts horizontally transported across the longitudinal boundary of 70°E and 140°E were 1.7 and 2.8 Tg, respectively.Notably, few dust production values are available from observations; hence, we had no references for conducting direct comparisons with our dust budget analysis.Nevertheless, some studies have provided the dust values of individual dust events through modeling; for example, Eguchi et al. (2009) estimated that during the 11-day period of May 5-15, 2007, the total amount of dust mass was nearly 34.1 Tg and the deposition of dust in the region between 90°E and 140°E was 29.4 Tg.Although some discrepancies exist between our study and Eguchi et al. (2009) in terms of details of daily dust emission and deposition, the magnitude of total dust amount and deposition proportion showed relatively good consistency, thereby indirectly reflecting the reliability of our results.It is worth noting that some limitations such as dust optical properties (e.g., dust refractive index and particle size distribution) and dust MEE in the model simulation (Colarco et al., 2014) likely result in somewhat uncertainties in the MERRA-2 dust estimates.

CONCLUSIONS
The long-range transport and deposition of Asian dust   markedly affect the environment, climate, and ecosystem.To comprehensively understand the Asian dust effect, analyzing and evaluating the dust transport and budget is essential.Thus, we analyzed a severe and uncommon dust event that occurred in East Asia from April 28 till May 3, 2011, by using integrated multi-satellite observations, model data, and reanalysis data.The dust budget of this event was quantitatively estimated based on MERRA-2.
Satellite observations revealed that this Asian dust storm occurred almost simultaneously in the Taklimakan and Gobi Deserts and was associated with the development of a Mongolian cyclone (low-pressure system).FY-3A captured a direct view of the dust outbreak in this event.The high AODs (> 1.6) obtained from the MODIS and significant AAODs (0.14-0.2) obtained from the OMI in the Taklimakan and Gobi Deserts indicate the dominance of dust aerosols in both regions.Driven by strong northwesterly winds, the dust plumes were gradually transported to downwind regions such as northeast, central, and southeast China and even to the Korean peninsula during the event.
The CALIPSO lidar has an unparalleled capability to monitor the vertical structure of dust layers.The HYSPLIT model and CALIPSO data revealed a two-layered dust structure in southeast China.The dust plumes in the lower layer (< 1.5 km) mainly originated from the Gobi Desert in southeast China.The elevated dust plumes (> 3 km) in these regions mainly originated from the Taklimakan Desert.
The dust budget analysis revealed that the dust mass loadings (5.7-6.6 Tg) were high from April 29 till May 1.The daily dust loading was the highest (approximately 6.6 Tg) on April 30 mainly because of the effect of strong northwesterly winds and large-scale dust transport.Moreover, the dust budget analysis revealed that during the event, the average dust loading between 70°E and 140°E was 5.2 Tg.The dust fluxes horizontally transported across the longitudinal boundary of 70°E and 140°E were 1.7 and 2.8 Tg, respectively.During the event, the total dust emission was 6.3 Tg between 70°E and 140°E, and the deposited amount of dust on the ground was 5.9 Tg in the region.These results are relatively consistent with other modeled results.
In summary, the integrated multi-sensor and model approaches used in this study enabled a more comprehensive investigation into this severe dust event in East Asia.This study suggests that MERRA-2 is a useful tool to quantitatively estimate the dust budget of extreme dust events in the region.In future studies, we will further evaluate the dust budget through observations with a higher level of confidence in order to better understand the dust effect on air quality and environment.
(d)], likely because the dust particles in the lower layer were directly transported from the Gobi Desert to Beijing.Figs.5(g) and 5(f) show obvious dust layers in central and southeast China on May 2. Associated with the HYSPLIT model [Fig.4(b)], the results indicated that the dust plumes in the lower layer (< 1.5 km) originated

Fig. 2 .
Fig. 2. MODIS-Aqua AOD combining the DB and DT algorithms at 550 nm with wind fields at 850 hPa, as determined from NCEP reanalysis.(a)-(f) show the variations in the MODIS-retrieved AOD with wind fields during April 28-May 3, 2011, in East Asia.

Fig. 4 .
Fig. 4. Backward trajectories of dust masses for 3 days at 6-hour intervals in (a) Beijing and (b) Hangzhou at 500, 1500, and 3000 m during the dust event period.

Table 1 .
Data characteristics analyzed in this study.Satellite data include FY-3A, MODIS, OMI, and CALIPSO.Backward trajectory data are simulated from the HYSPLIT model.The reanalysis data contain MERRA-2-assimilated aerosol observations and NCEP wind field data.