Seasonal dependence of the long-range transport and vertical distribution of free tropospheric aerosols over east Asia: On the basis of aircraft and lidar measurements and isentropic trajectory analysis
 Seasonal changes in the vertical structure of free tropospheric aerosols over east Asia, on the basis of aircraft-borne and lidar measurements, and on the pathway of the long-range transport of Asian dust particles inferred from isentropic trajectory analysis are discussed. Aircraft-borne measurements held in situ in the free troposphere over central Japan in 2000–2001 revealed a small in scale yet steady transport of dust in the lower-middle free troposphere (2–6 km altitude) during spring including days with no evident dust outbreak. Such dust, found as background, was observed even in summer in the regions higher than 4 km under the influence of remaining westerly winds but not in the lower regions. From a series of lidar observations over Nagoya (35°N, 137°E), Japan, noticeable changes in aerosol characteristics were obtained in the free troposphere from spring to summer. Taklimakan desert is suggested as possible important source of the background dust.
 Over the past decades, many reports have been made on the aeolian transport of mineral dust to the vast regions of the Northern Pacific, and it is now well recognized that it originates in the continental arid regions upwind [Duce et al., 1980; Gao et al., 1992; Leinen et al., 1986; Merrill et al., 1989; Rex and Goldberg, 1958; Uematsu et al., 1983]. There is a growing concern over the impact which the dust may have on the global climate system and biogeochemical cycle because of the vast scale involved in the transport. While they travel in the atmosphere, not only do dust particles interact directly with incoming radiation to the atmosphere [Sokolik and Toon, 1996; Tegen et al., 1996], they also provide an effective reaction surface (hence a sink) for acidic gaseous species such as SOx and NOy [Denterner et al., 1996], further increasing the dust's potential to act as a giant Cloud Condensation Nuclei: CCN [Levin et al., 1996]. These features may cause changes in the cloud cover and precipitation patterns [Albrecht, 1989; Pincus and Baker, 1994; Twomey, 1974], thereby affecting hydrological and geochemical cycles. In addition, it has been suggested by a number of investigators that the aeolian input of soil dust in remote ocean sites distant from any estuary is the key source of trace metals, especially Fe, which critically limits the primary production of phytoplankton [Falkowski, 1997; Gao et al., 2001; Martin et al., 1989]. This implies a significant role played by the dust in controlling not only the marine ecosystem, but also the level of atmospheric CO2.
 Dedicated investigations in the past on the long-range transport of dust from the Asian continent revealed that the maximum transport takes place in spring when dust outbreaks most frequently occur. However, most measurements have been made from the ground and have tended to focus only on severe individual events or on the spring period alone. Iwasaka and his colleagues frequently detected lidar signals suggesting the presence of weak dust layers (i.e., background KOSA) in the free troposphere over Japan during periods with no evident dust outbreak or even in seasons other than spring [Iwasaka et al., 1983, 1988; Iwasaka and Kwon, 1997; Kwon et al., 1997; Sakai et al., 2000]. As can be inferred from this description of previous measurements, there is a lack of in situ free tropospheric measurements, as well as cross-seasonal measurements to fully characterize the spatial and temporal variabilities of free tropospheric aerosols over the region.
 The distinct seasons of eastern Asia create strong monsoons that cause variation not only in the passing of the westerly jet, but also in the source strength of continental aerosols. Hence the long-range transport of aerosols by the westerly wind in this particular region should exhibit large spatial and temporal variations. Knowledge on the aerosol distribution in the free troposphere is essential since this is where the long-range transport actively takes place. In the following sections of this report, we will discuss the seasonal changes in the vertical structure of the free tropospheric aerosols over east Asia, on the basis of aircraft-borne and lidar measurements, and the transportation pathway and origin of the Asian dust particles inferred from the isentropic trajectory analysis.
2. Measurements and Analysis
2.1. Aircraft-Borne Measurements
 Six series of aircraft-borne measurements were conducted over Japan during 2000–2001. Measurements took place on 10 March (16:24–19:30 Japan Standard Time, JST), 11 March (8:56–12:19 JST), 29 April (10:57–13:51 JST), and 19 July (9:09–12:26 JST) of 2000, and 21 March (9:11–13:23 JST) and 20 July (17:25–20:29 JST) of 2001. For the sake of safety, days of unstable weather conditions, such as strong frontal activities, were avoided. The flight tracks are shown in Figure 1. The four series of measurements made in the spring were conducted mostly over Wakasa bay (36°00′N, 135°30′E) on the Japan Sea side of Japan, and the two series of measurements made in the summer were taken over the sea off the eastern coast of Kii peninsula (33°30′N, 136°30′E) on the Pacific side. In every flight, a Cessna 404 aircraft took off from Nagoya (35°N, 137°E) and followed several straight tracks at multiple levels over the destination covering the range from the top of the boundary mixing layer to the mid free troposphere (up to 5–6 km) to obtain a vertical structure of the aerosols (Figure 1b).
 Ambient air was introduced into the aircraft cabin through an inlet as shown in Figure 2. A manifold placed directly below the inlet distributed sampled air to various instruments. The pressure inside the aircraft cabin was maintained to be the same as the ambient air. The inlet was aligned parallel to the aircraft axis, and airflow entered the thin-walled tip at a rate of 640 L/min; subsequently, the conic diffusing section decelerated the axial velocity from 60 m/s to 17 m/s. Stainless steel was selected as the inlet material to minimize the formation of electrical charges on the smoothed wall surface, and no anti-bounce coating was applied to the inlet walls. Particle loss is inevitable in any measurement involving inlet sampling. Generally, larger particles are more prone to loss. For example, the 90° bent tube of the inlet may have caused large particles to impact upon the inner walls. To examine the extent of loss caused by the bend, we modeled the transmission characteristics using calculations from Pui et al. . The curvature ratio, which is the ratio of the bend radius to the radius of the tube cross-section, in our case was 12. This is well within the range (5 < Rbend/Rtube < 30) where the curvature effect has been proven to be insignificant. Calculations were made by assuming a flow velocity of 17 m/s and particle densities of 1.0, 2.0, and 3.0 g/cm3. The density of dust is assumed to be around 2.6 g/cm3 [Ishizaka, 1972]. Table 1 shows the diameter at which particles of various densities are transmitted with 90%, 50%, and 10% efficiency. Theoretically, 50% of particles with a 5.3 μm diameter and 3 g/cm3 density will be deposited in the bend. Given the high Reynolds numbers (Re) of 40,000 inside the diffuser and 32,000 inside the tube, turbulence is also likely to have occurred inside our inlet, thus causing particle deposition. Calculations from Muyshondt et al.  were used to estimate the turbulent loss caused by the inlet length. The transmission efficiencies of the diffuser cone (5.2 cm) and straight tube sections (50 cm, including the connection between inlet and manifold) were calculated separately and were then multiplied to deduce the overall transmission efficiency of the entire straight sections. For simplification, the diffuser section was assumed to be cylindrical with the diameter being the mean of the tip and the tube diameters. Transmission efficiencies calculated for particles of various sizes and densities are shown in Table 2. According to these results, 5 μm particles with 3 g/cm3 density would pass through the straight sections of the inlet with 65% efficiency. The above estimations indicate that particle deposition would have been slightly more pronounced in the bend than in the straight tube section. Particle residence time for the entire inlet length was estimated to be about 0.05 s. Hence even a 10 μm, 3 g/cm3 density particle with sedimentation velocity of 0.9 cm/s would fall only 0.04 cm inside the inlet, which makes gravitational sedimentation insignificant. Difficulties in maintaining isoaxial sampling at the inlet tip, or obtaining a representative aerosol size distribution from external probes, remain as potential sources of uncertainty in our observational results. However, depositions in bends and high Re flow are often considered the most important loss mechanisms [Blomquist et al., 2001; Okazaki et al., 1987]. Overall, the fairly slow aircraft (60 m/s) used in our study allowed the inlet to be short in length, thereby helping to avoid any significant loss of large particles inside the inlet.
Table 1. Estimated Transmission Efficiencies for the Bent Section of the Inlet (at 20°C, 1013 hPa)
Particle Density g/cm3
Particle Diameter Transmitted With Various Efficiencies, μm
Table 2. Estimated Transmission Efficiencies for the Straight Sections of the Inlet (at 20°C, 1013 hPa)
Particle Diameter, μm
Transmission Efficiency With Various Densities, %
 Aerosol particles were sampled directly with an onboard two-stage low-volume impactor (LVI). Particles were directly collected onto sampling substrates placed inside the impactor. Carbon-coated nitrocellulose (collodion) films supported by Ni or Cu grids were used as the substrates. The jet diameters of the first and second stages of the impactor were 1.3 mm and 0.4 mm, respectively. Approximately 1000 cm3 of sampled air was introduced into the impactor every minute. By assuming a particle density of 1.0 g/cm3 under standard atmospheric conditions (1013 hPa, 20°C), the diameters at which particles would be collected with 50% efficiency were expected to be 1.6 μm and 0.2 μm at the respective stages. The morphology and chemical constituents of individual particles were analyzed with a scanning electron microscope (SEM) (HITACHI, S-3000N) equipped with an energy dispersive X-ray (EDX) analyzer (HORIBA, EMAX-500). Particle size was estimated from the direct appearance of individual particles in the electron micrograph. The maximum length of a particle was considered the diameter for an irregularly shaped particle [Trochkine et al., 2002]. In parallel with the particle collection, the number-size distribution was measured by two optical particle counters (OPC). The KC18 counter (RION) divided 0.1- to 0.5-μm particles into five channels and the TD-200S counter (SIGMATEC) divided 0.3- to 10-μm particles into 15 channels.
2.2. Lidar Measurements
 The lidar observation site in the eastern part of Nagoya city, Japan (35°N, 137°E), allowed continuous monitoring of vertical profiles of the free tropospheric aerosols through the period of March to August 1994. The lidar used here consisted of a pulsed Nd:YAG laser with wavelengths of 1064, 532, and 355 nm, a receiving telescope with a diameter of 1000 mm, and a multichannel photon-counter (dual five channels). (The main characteristics of the lidar system are summarized in Table 3.)
Table 3. Main Specifications of the Raman Lidar System at Nagoya
>350 mJ/pulse (1064 nm)
>100 mJ/pulse (532 nm)
>150 mJ/pulse (355 nm)
Laser beam divergence
Field of view
Photon counting method: Multichannel counter
Range resolution 50 and 100 m
PMT-1 (404 nm)
H2O Raman Nd:YAG THG
PMT-2 (375 nm)
O2 Raman of Nd:YAG THG
PMT-3 (532 nm)
P component of Nd:YAG SHG
PMT-4 (532 nm)
S component of Nd:YAG SHG
PMT-5 (1064 nm)
Basic pulse of Nd:YAG
Raman scattering (355 nm)
H2O, O2, N2
Mie/Rayleigh scattering (532 nm)
Scattering ratio, polarization ratio
Mie/Rayleigh scattering (1064 nm)
 The scattering ratio of atmospheric particulate matter, R (Z), is defined as
where Bm(Z) and Bp(Z) are the atmospheric molecular and particulate backscattering coefficients, respectively, at altitude Z. Equation (1) is rewritten as
From (1'), we can regard the value of R(Z)-1 as a parameter indicating the mixing ratio of particulate matter at an altitude of Z measured with an optical technique. We assumed the height of the matching point (the aerosol-free atmosphere) to be above 25 km when deducing the backscattering coefficient of particles.
 The total depolarization ratio Dt is defined by
where B is the backscattering coefficient, subscripts ∥ and ⟂ refer to measurement made with the plane of the polarization of backscattered light parallel and orthogonal to that of the transmitted laser pulse, respectively, and subscripts p and m refer to components due to atmospheric aerosol particles and air molecules, respectively.
 For single scattering by spherical particles, the polarization of the incident wave is retained in the backscattering. The particle depolarization component in the lidar return would be 0 if only that scattering mechanism was present. When the depolarization ratio Dp is high, nonspherical particles are expected to be present in the atmosphere. It has been widely accepted that dust particles have strong nonsphericity and show a depolarization ratio ≫0 [Iwasaka et al., 1988; Kwon et al., 1997; Sakai et al., 2000]. Here we assumed that the particle backscattering effect is a major component in backscattering light from the lower atmosphere (the boundary mixing layer and the lower troposphere), and ignored the contribution of atmospheric molecule backscattering in our estimation of the depolarization ratio of particles, as follows;
3. Results and Discussion
3.1. Particle Number-Size Distribution Measured by Aircraft-Borne OPCs
Figures 3a and 3b show the particle number-size distributions observed over central Japan by the onboard OPCs, and distributions from spring and summer are compared. In the middle of the free troposphere at altitudes greater than 4 km, the distributions from both seasons were quite similar and all distributions showed a bimodal structure with concentration peaks at 1 to 2 μm in diameter. However, in the lower part of the free troposphere (below 4 km), while the spring distributions showed no significant change with altitude, the summer distributions showed a marked drop throughout the size range and the peak found in the coarse range was somewhat obscured. Vertical variation of the coarse (D > 1 μm) particle numbers (Figure 3c) provided a better picture of this seasonal trend where there were similar numbers of particles at higher altitudes, and a summer decrease in lower altitudes. On average, the summer drop accounted for 86% of the number of coarse particles measured in spring. Whereas in the case of the mid free troposphere, the drop was only 9%. Emphasis is on the fact that the series of aircraft measurements were conducted under relatively stable atmospheric conditions, and hence these distributions may represent the “background” state of the atmosphere. Values of dN/dlogD values in the event of a dust outbreak, on the basis of measurements by Xu and Kai  are compared in Figures 3a and 3b. The highest points of the vertical bars indicate values measured in Nagoya during the prominent 8 April 2000 episode. The referred values, however, are based on ground measurement. Therefore values in the Figure 3 were converted for standard temperature and pressure conditions (STP: 20°C, 1013 hPa). Apparently, the contribution of a dust outbreak to the particle concentration is more pronounced in larger-particle-size ranges. The number of 1.2-μm particles could rise by 1–2 orders of magnitude compared to the background values.
3.2. Representative Particle Types From the Coarse-Particle Range
 Representative types of the coarse particles collected in the free troposphere are shown in Figure 4. One of the most abundant types found was mineral dust (Figure 4a). As shown in the electron micrograph, they were often irregularly shaped and rough on the surface. An X-ray spectrum showed that they are abundant in Mg, Al, Si, K, Ca, and Fe, characteristic of the crustal component. Sea-salt particles were also commonly found as coarse particles in the free troposphere (Figure 4b). They were mainly composed of NaCl and often resemble its crystal-like shape. Other types included giant ammonium sulfates, and soot agglomerates and fly-ash particles that appear to be of anthropogenic origin.
3.3. Seasonal Changes in the Vertical Distribution of Coarse Particles
 The fraction of coarse particles attributable to each particle type can be determined through an individual particle analysis (section 3.2) where we count how many particles of each type were found among the total number of particles analyzed. Combining this information with the relative change in the total number of coarse particles measured by OPC (section 3.1), we could find out which particle types were responsible for the drop observed in summer. In Figure 5, the relative seasonal changes in the total number of coarse particles are indicated separately with respect to altitude. Fractions attributable to the major particle types, such as dust, sea salt, and others, are superimposed. In Figure 5, 100% denotes the average number of coarse particles measured in spring within the respective altitude ranges.
 In the altitude range above 4 km, both the particle number and the proportion of each particle type did not show significant changes between the seasons; also, dust particles remained dominant (about 80%) in both seasons. In the altitude range below 4 km, there was a marked decrease in the total number of coarse particles in summer, as mentioned above, and the absence of dust was the most likely cause of this decrease.
Figure 5 implies that in much of the free troposphere during spring, mineral dust is always major in the coarse particle range even when there are no visual signs of major dust outbreaks. Even though the concentration will be too small to be detected as a dust outbreak, the persistence and high speed of the westerly jet will create a significant flux of dust from the continent to the vast regions farther east. Matsuki et al.  pointed out the possibility that the steady state in spring alone could generate a horizontal mass flux of dust comparable to that transported in the event of a dust storm.
 The most striking information contained in Figure 5 is that the particle concentration in the higher part of the free troposphere remained comparable to that in spring and the dust remained dominant even in summer. This indicates that dust transport is not a phenomenon confined to the spring, but continues on a modest scale that we cannot visually detect through, for example, satellite images or ground observations.
3.4. Relative Humidity Distribution
Figure 6 shows the relative humidity profiles obtained during the aircraft measurements. In spring, a dry air mass with relative humidity of less than 20% was dominant in the free troposphere, usually at an altitude greater than 2 km. The one exception occurred on 11 March, 2000 when humidity increased at an altitude of over 4 km because of overlaying cloud. Summer profiles were quite different in structure. A layer with high humidity of 50 to 70% extended from the boundary mixing layer to as high as 4.5 km, and humidity dropped markedly above that to a level comparable to that in the spring free troposphere. A dry air mass observed in the spring free troposphere and a similar one observed in the mid-higher free troposphere in summer coincided with the dust-dominant layer mentioned above; we observed similar coincidence in the presence of humid air masses and a relatively dust-free air mass. The most likely cause of this spatial variation was the difference in the origin of the observed air parcels.
3.5. Backward Trajectory Analysis
 We used isentropic backward trajectory analysis to estimate the origin of the air parcels. An air parcel was traced backward from the point when and where particle collection was actually done. The meteorological field used in the analysis was global objective analysis data provided by the Japan Meteorological Agency (GAPLX). The 6 hourly data included geopotential height, horizontal wind, temperature, and humidity (available below 300 hPa) at 18 pressure levels with horizontal resolution of 1.25°. An air parcel was traced backward on the isentropic surface by using the fourth-order Runge-Kutta scheme [Sakai, 2001]. Trajectories were calculated for a duration of 5 days with a time step of 1 hr. The isentropic assumption cannot stand in saturated air or convective conditions as are often encountered inside the planetary boundary layer. For this reason, we stopped the calculation in cases where the air parcel became saturated with respect to water or ice, or intersected a vertically unstable layer or the Earth's surface. Figure 7 shows the estimated trajectories of air parcels arriving at the sampling sites. As is evident in Figure 7, for the sampling points higher than 4 km, air parcels were estimated to always originate within the Asian continent regardless of season. For the sampling points lower than 4 km, though, there was a clear seasonal difference in air parcel origin, as we expected from the differences in the observed aerosol and humidity characteristics. In spring, air parcels also approached from the continent carrying dust as was the case for the higher free troposphere. However, in summer, air parcels became more maritime in origin and approached from the sea south-west of Japan, thus bringing humid and clean air.
 Given the limited number of our measurements (four series in spring and two in summer), the question arises as to whether we have observed representative states of the two seasons, or rather, some sporadic cases. Daily trajectories of the air parcels approaching Nagoya in April and July during 2000 and 2001 are shown in Figure 8. Two test areas, dust (25–50°N, 80–105°E) and marine (10–35°N, 120–145°E) source areas, were set to better distinguish the arid continental and marine origin of the air parcels. The fraction of trajectories passing through each area is summarized in Table 4. Because Nagoya is close to the marine source area, trajectories passing over both test areas were regarded as continental. At every point of air parcel arrival, the relative humidity over Nagoya was obtained from the meteorological field used in the calculation; these are presented in the table as monthly mean values. The mean transit time of air parcels from the continent is given in the table as an average value taken from those that arrive after crossing the 90°E meridian.
Table 4. Fraction of Trajectories Passing Through Hypothetical Dust (25°–50°N, 80°–105°E) and Marine (10°–35°N, 120°–145°E) Source Areas Before Arriving 3 and 7 km Over Nagoya (35°N, 137°E)a
Monthly mean relative humidity at the point of arrival. Mean transit time required for an air mass to reach Nagoya from the 90°E meridian.
Altitude: 7 km
Transit time, days
Altitude: 3 km
Transit time, days
 Air parcels approaching Nagoya at higher altitudes, 7 km in this case, both in April (spring) and July (summer) exhibited similar features. Pathways were mostly from over the Asian continent, centering over the arid regions of China and Mongolia. Table 4 shows that the majority (77–80%) of air parcels came after passing over the dust source region in spring. In summer, about half (39–61%) were still from the dust source regions. Air parcels were relatively dry in both seasons with a similar average humidity of about 40%. However, the mean transit time from a point at 90°E to Nagoya was only 2.7–3 days in spring, while it was 4.6–5 days in summer because of the weaker westerly wind.
 The pathways of air parcels approaching at 3 km over Nagoya showed a distinct seasonal change, as was also the cases shown in Figure 7. In spring, air parcels took a path that was probably affected by the high pressure system stationed over Siberia; the parcels then descended and headed south at about 110°E. Despite the strong continental tendency, only a small fraction of the air parcels passed through the dust source area (10–30%), and most pathways followed a northerly detour. In summer, on the other hand, the paths were mostly along the edge of the extending subtropical high stationed over the Pacific, and the majority (65–84%) of the trajectories originated in the marine source area. Also, the average relative humidity was higher at about 60%.
 Since the trajectories of observed air parcels (Figure 7) well reflected the main path for each season (Figure 8), it seems likely that our results represent the typical aerosol characteristics of each season.
3.6. Seasonal Changes in Aerosol Vertical Profiles Measured by Lidar
Figure 9 shows the seasonal change in the vertical profiles of the scattering ratio, R(z) at a laser wavelength of 1064 nm and the aerosol depolarization ratio, Dp(z) at 532 nm, derived from a series of lidar measurements made at Nagoya (35°N, 137°E) from March to August 1994. The white horizontal lines in the top panel of Figure 9 indicate tropopause heights obtained from upper air sounding at Hamamatsu station (47681; 34°45′N, 137°42′E). Tropopause heights ranged from 10–13 km during spring months, but increased with time and eventually exceeded the heights covered in Figure 9. Such layers having a high scattering ratio (>5) and depolarization ratio (>3) were frequently observed throughout the free troposphere during the spring months (March–May). As the season changed, these layers gradually shifted to higher altitudes. Layers with a high scattering ratio inside the planetary boundary layer (<2.5 km) in summer months suggest the dominance of spherical particles because of the inconsistency with the depolarization ratio.
Sakai et al.  have found a systematic relationship between relative humidity and the origin of air masses within the free troposphere. An air mass passing over the Asian continent tends to be dry with relative humidity of less than 20%, whereas one coming from the Pacific Ocean holds more water (a relative humidity exceeding 70%). Also, aerosol layers with a high depolarization ratio appeared most frequently in the air masses coming from the continent. This is in good agreement with what we have observed in our study, and that the high depolarization ratio occurring in the dry continental air mass is due to dust has been confirmed by our individual particle analysis.
 Taking all into account, the layer with high scattering and depolarization ratios shown in Figure 9 was due mainly to dust coming from the continent. The upward shift of such a layer toward summer is a fine illustration of the monsoonal transition gradually taking place in this region. Thus we have confirmed that the replacement of the continental air mass of winter by the subtropical air mass of summer does not occur uniformly with respect to altitude, but proceeds from the lower troposphere depending on the elevation of the intruding subtropical high.
3.7. Potential Source of the Background Dust
 Current results have demonstrated, at the very least, that background dust is associated with continental air masses. Unlike the dust generated in the event of dust outbreaks, such dust does not always involve strong frontal activities over inland arid regions. Hence we expect background dust to have different generation mechanisms and transport pathways than the dust driven by spring storms. The Taklimakan (38°N, 82°E) and Gobi (40°N, 105°E) deserts are recognized as the two major source regions of dust in east Asia. In this section, we describe a number of findings which have led us to hypothesize that the Taklimakan desert is an important source of background dust.
3.7.1. Persistence of Dust in the Tarim Basin
Kurosaki and Mikami  compiled records of three hourly data on current weather (the SYNOP report) collected by 12 local meteorological observatories inside the Tarim basin, where the Taklimakan desert is located. Their statistics revealed that the occurrence of floating dust is one order of magnitude higher in the southern half of the basin than in the northern half. The frequency in Hotan (37°N, 80°E) located in the southern rim of the basin is especially outstanding (35–60%), reaching a maximum in spring, which is the driest of the seasons, and remaining high until August to September despite a decrease in July due to precipitation.
3.7.2. Topography of the Tarim Basin
 The Tarim basin is surrounded by high mountains exceeding 5000 m except at the eastern edge. However, easterly or north-easterly winds dominate the basin floor preventing dust from blowing eastward [Sun et al., 2001]. The only way for the dust to get out of the basin is to rise above 5000 m. As is evident in Figure 8 and Table 4, air parcels more frequently travel in the rather high altitude range (7 km) than in the lower range (3 km), above the main two source regions before approaching Japan, even in summer. Once the dust is lofted above 5000 m, there is a better chance that it will be transported over long distances because of the stronger westerly jet dominating the altitude range. One possible mechanism which could drive dust from the basin floor over the mountains is the daytime anabatic winds caused by the local circulation system between the Kunlun mountains (a mountain range along the southern rim of the Tarim basin) and the basin floor [Abe et al., 2002; Mikami et al., 1995]. The bare dry surface and high elevation (>1000 m ASL) of the basin floor would efficiently warm the air directly above, and is therefore likely to cause a very strong thermal plume, possibly carrying dust to high elevations along the northern slopes of the Kunlun mountains.
 In the case of the Gobi desert, on the other hand, even dust outbreaks may not be able to lift dust as high as the mid-troposphere. Sun et al  compiled 40 years of dust storm reports in China and concluded that 90% of the dust raised from the Gobi would remain in the lower part of the troposphere and would be the main source of aeolian deposits in the proximal regions. Their conclusion was further supported by the depositional record which showed the highest limit of loess deposition on the slopes of the downwind Chinese Loess Plateau to be less than 2850 m. This is in good agreement with the observational findings by Iwasaka et al. . In this sense, although the contribution of dust from this source is still not entirely deniable, the Gobi is less likely to be a significant source of the background dust observed in the mid free troposphere.
3.7.3. Overall Likelihood of Causing Long-Range Transport
 A significant fraction of the dust observed over Japan in the steady state of spring and in summer probably has its origin in the Tarim basin. This is further supported by Bory et al. , who reported that the mineral composition and the isotopic ratios of Sr and Nd from the dust entrained in the snow deposits of Greenland indicated the dust's origin was mainly the Taklimakan desert. It is reasonable to think the dust can be transported at higher altitudes where the westerly flow is more intense when explaining such long-range transport extending literally halfway around the globe. However, the local circulation system and accompanying vertical flux of dust in the mountainous regions are not well understood. Since it is also possible that the background dust originated directly from high mountain ranges where it could have been generated by weathering processes [Sun, 2002], investigation of the behavior of the boundary mixing layer between the Tarim basin and Kunlun mountain ranges, as well as chemical identification of the dust source, remain as future tasks necessary to fully prove the Taklimakan desert's role in generating the background dust.
 An ensemble of aircraft-borne measurements, lidar measurements, and isentropic trajectory analysis, has revealed the characteristic seasonal variations in the vertical structure of aerosols over east Asia.
 Through in situ measurements from an aircraft, we found coarse particles in much of the springtime free troposphere over Japan to always consist of mineral dust even at times when there were no visual signs of major dust outbreaks. This led us to conclude that there is a steady transport of dust in the lower-middle free troposphere (2–6 km altitude) during spring, and its persistence and the high speed of the westerly jet would together create a significant flux of dust from the continent to vast regions farther east.
 In the mid free troposphere (>4 km), we discovered that the particle concentration in summer remained comparable to that in the spring, and the dust remained dominant, contrary to the general understanding that the summertime free troposphere over the region should be dust free because of the prevailing subtropical high. This indicates that dust transport is not a phenomenon confined to the spring season, but continues on a modest scale in the layer under the persisting westerlies.
 Layers with high scattering and depolarization ratios in the dry free troposphere were frequently observed by lidar, and we confirmed that these layers were mainly composed of dust coming from the continent. The upward shift of such layers toward summer illustrated the monsoonal transition characteristic of this region. We suggest that the replacement of the continental air mass of winter by the subtropical air mass of summer does not occur uniformly with regard to altitudes, but proceeds from the lower troposphere depending on the elevation of the intruding subtropical high.
 Tarim Basin is a stable dust source at the southern rim of the Taklimakan desert, and taking into account a possible dust uplifting mechanism created by the unique topography and the strong persisting westerlies hanging over the high mountains, we speculate that it is a major source of the dust observed over Japan under steady conditions of spring and summer. However, investigation of the local circulation system inside the basin and chemical identification remain as future tasks necessary to determine if this is indeed a major source location.
 This investigation was supported by the Japan Ministry of Education, Culture, Sports, Science and Technology (Grant-in-Aid for Specially Promoted Research, 10144104).