The Japan-Sino joint project, Aeolian Dust Experiment on Climate impact (ADEC), was initiated in April 2000 in order to understand the aeolian dust impact on climate via radiative forcing. As a part of the ADEC project, we have conducted field research in a sand dune and a gobi (i.e., a desert in which the soil surface consists of sand and pebbles with flat surfaces) in the south of the Taklimakan desert, China. The purpose of this study is to understand the wind erosion process and its relation to the meteorological and soil physical parameters. For this purpose, we measured the vertical profiles of wind speed, air temperature, and humidity as well as the other meteorological elements using an automatic weather station. A new sand particle counter (SPC) was newly developed to measure the saltation process. The SPC detects a signal change when a saltation particle passes through the slit between the laser beam transmitter and receiver. From this signal change, we can measure saltation particles from 30 to 667 μm diameter with 32 bin classes and particle numbers of each bin class every second. We have operated this SPC in the field, and it proved to be useful for the saltation process study when data corrections and calibration were properly made. During the observation period (1–21 April 2002), a total of eight dust events occurred; we analyzed two events: 5 April and 14 April cases. The results can be summarized as follows: (1) Total saltation fluxes in the 5 April case from 1223 to 1430 UT were 37.93 kg m−2 at 30 cm height and 43.71 kg m−2 at 20 cm height for the gobi site and 2.61 kg m−2 at 30 cm height for the dune site. (2) In the 14 April case, from 0327 to 0830 UT, the total saltation flux was 8.95 kg m−2 at 30 cm height for the gobi site. (3) Saltation flux at the gobi site in the 5 April case was more than 10 times larger than that of the sand dune, though the distance between the sites is 4 km. This is because the number of the parent soil particles around 80 μm at the gobi site was more than 10 times greater than that of the dune site. (4) Height dependency of saltation particle size and number was found in the gobi site; that is, the particle size distributions at the gobi sites in the 5 April case indicated that the number size distribution of the coarse particles, 117 to 554 μm, at 20 cm height was greater than that at 30 cm height. This size-height dependency was reasonable from a physical point of view. However, present theory cannot explain this well.
 A huge amount of mineral dust is emitted into the atmosphere from arid and semiarid regions of continents. According to Intergovernmental Panel on Climate Change (IPCC) , the global estimate of such dust using the dust global climate model (GCM) ranges from 1000 to 5000 Mt yr−1, although reliability of these estimations is not so high. Aeolian dust in the atmosphere is thought to be a major contributor to optical thickness, and its radiative forcing plays an important role in the atmosphere and the ground surface via direct and indirect effects. However, there still remains a large uncertainty in the understanding and model representations of all processes of aeolian dust, such as wind erosion, uptake in the free atmosphere, deposition, optical properties of particles, and its radiative forcing in the global and/or regional scale.
 With this background, the Japan-Sino joint project, Aeolian Dust Experiment on Climate impact (ADEC) was initiated in April 2000 in order to understand the aeolian dust impact on climate via radiative forcing [Mikami et al., 2002; http://www.aeoliandust.com]. The main objective of this project is to understand the whole process of aeolian dust including its outbreak, long-range transport, and radiative effects. In situ observations and network observations were planned in China and Japan in the area from 80°E to 140°E, which includes source regions and deposition areas. As part of the ADEC project, we conducted field research at a sand dune and a gobi (sand and pebble) desert in the south of the Taklimakan desert, China, during the ADEC IOP 1 (April 2002) and IOP 2 (March 2003). The purpose of our study is to understand the wind erosion process and its relation to the meteorological and soil physical parameters.
 Although the saltation process is a key mechanism in dust emission [Shao et al., 1993], the present saltation theory adopted the assumption of homogeneity and the uniform particle size and is assumed as identical trajectories. For this, saltation particle size distribution with height is expressed uniformly in size. The vertical profile of saltation mass flux is not adequately described either. This hypothesis is not reasonable from the physical point of view because actual saltation is a heterogeneous process. The theoretical consideration of the heterogeneity of the saltation process was first made by Anderson and Hallet  and Anderson . Recently, Shao  proposed the outline of a new heterogeneous saltation theory that will be more suitable for expressing saltation in nature. However, field data, which include information of the whole saltation process and can be used for verifying the theory, are lacking. Many saltation measurements have been made in the field using portable wind tunnel. However, the turbulent structure within a tunnel is not a natural condition because of the size limitation of the width and height of the tunnel. In such a condition, the saltation structure does not well represent the natural saltation that is heterogeneous in nature. Only Leys and McTainsh  tried to observe the saltation process at different heights with high time resolution and size information in natural field conditions.
 In this study, we sought to monitor the heterogeneous saltation process and meteorological elements controlling the saltation process. We observed these processes under natural field conditions of the gobi and sand dune deserts. The observations were made in the suburbs of Qira, a small oasis located in the south of the Taklimakan desert, China. In this region, a sand dune desert and a gobi desert are distributed, and meteorological observations to monitor the characteristics of the desert environment have been conducted [Nagashima et al., 1991; Mikami et al., 1995]. In addition, according to the statistical analysis of meteorological observatory data in the Taklimakan desert [Kurosaki and Mikami, 2002], dust events occur most frequently in Hotan, which is located in the south of the Taklimakan desert, 100 km east of Qira, among the meteorological observatories in the desert. Therefore this location presents good conditions for monitoring dust outbreaks.
2. Newly Developed Sand Particle Counter
 Saltation process field measurements have been conducted using various types of sand traps. Most recently, passive samplers and fast responding samplers have been widely used. The Fryrear  passive sampler is famous for its capability for long-term use in the field environment. The Leach sampler [White, 1982] is mainly used in wind tunnel experiments. Both samplers are aligned in the wind direction and accumulate the saltation sand particles within the trap. However, the passive sampler is not suitable for monitoring the time variation of the saltation process.
 SENSIT [Stockton and Gillette, 1990] is widely used as a fast responding sampler. It senses the impact kinetic energy of saltation particles and has sufficient time resolution for measurements. It can thus be used for the saltation process study. This sensor, however, detects the kinetic energy of sand particles, so that it is impossible to separate the kinetic energy information according to particle size and velocity.
 Streamwise saltation flux Q is basically a function of particle size and height. It is generated by the wind shear energy and changes with the time variation of wind field near the ground surface. In order to monitor the saltation process, the time change of the number-size distribution of saltation particles, including their vertical profile, must be monitored together with related parameters, such as friction wind velocity, soil wetness, and parent soil information. The present sand traps do not fully satisfy these requirements.
 For these reasons, we developed a new sand particle counter (SPC) for use in field measurements of the saltation process [Yamada et al., 2002]. This instrument must satisfy the following conditions: (1) it must be rugged for use in the field and must not require AC power, (2) its measurement range of particle sizes must be wide enough to cover saltation sand particles, e.g., 40–600 μm, with satisfactory size class number, and (3) the sampling frequency must be high enough, e.g., 1 s, to monitor the physical process of saltation.
 The new SPC is designed to sequentially measure the saltating sand particle size in the field. We use a semiconductor laser in the sensor unit of the SPC. The principle of this sensor is originally based on the snow particle counter, which was first developed by Schmidt . The original Schmidt snow particle counter employs a beam from a tungsten lamp. Kimura and Sato  replaced this with a laser beam to achieve a fast response (Niigata-denki, SPC-S7). We applied this sensor with modification to measure saltating sand particles in a hyper arid environment. Figure 1 shows the principle of the SPC sensor. When a sand particle passes through the laser beam, the output signal of the detector drops in proportion to the particles cross-sectional area. If we digitize the output signal with a very fast response time, we can acquire information of the sand particle size and number from a sequence of signals, assuming that each particle has a spherical shape.
 The SPC sensor detects the information of particle size from signal change in proportion to its cross-sectional area. However, when particle is fine and smaller than the smallest detectable size of the instrument, accurate measurement becomes difficult owing to the diffractive effect of the sensor beam. In addition, white noise from the electric circuit of the signal-processing unit will mask the signal from extremely fine particle, say smaller than 30 μm. For these reasons, digitized signal was filtered by a threshold level so that signal smaller than the smallest particles accounted by the SPC can be eliminated. That is to say, the effect of the signal change from particles smaller than the smallest size particle accounted for by the present sensor is eliminated.
Table 1 presents the specifications of the SPC. There are 32 size bins with sizes ranging from 30 to 667 μm in diameter. The particle number of each bin is counted every second and stored in the compact flash memory in the control box. For use in a desert environment, every unit and box of the system is protected against dust, and all electric power is supplied from solar cells installed on the ground. Figure 2a shows the SPC sensor unit set on the gobi desert, and Figure 2b shows the sand dune in Qira in the southern Taklimakan desert, China.
Table 1. Specifications of the Sand Particle Counter
luminescent diode laser with collimeter
PIN photo diode
2 mm (height) × 25 mm (length)
30–667 μm (32 channels)
Operating temperature range
Width × length × height
175 × 383 × 201 mm
Input signal range
0 to ∼2.5 V
Electric power consumption
maximum 1 A (DC12V)
Width × length × height
400 × 600 × 187 mm
 The raw SPC data will have some systematic errors caused by its ambient temperature dependency, pollution of detector lens due to dust, and so on. Before analyzing the data, we must therefore calibrate and correct the raw data. The procedures are listed below.
2.1. Edge Effect Correction
 The width (height) of the slit between laser oscillator and detector is 25 mm (2 mm). The sensor detects the cross-sectional area of a particle. However, when a sand particle is crossing the edge of the slit area, the sensor will underestimate its size. For this reason, we correct the particle size using the following equation [Sato, 1987], assuming that the spatial distribution of sand particles in the neighborhood of the sensor slot is homogeneous:
where L is the diameter of laser beam and Dobs is the observed particle diameter.
2.2. Temperature Correction
 The data processing and correction units are packed in the control box installed on the ground. The electrical coefficient of the circuit and its resultant laser beam intensity are thus affected by the environmental temperature within the control box. To avoid this, we made an environmental test on the temperature dependency of the data processing and correction units within a chamber in the laboratory. On the basis of this, we developed a temperature correction equations for each sensor unit. The environmental temperature within the control box was monitored every minute during the observation. The raw data were then corrected by the temperature correction equation for each unit.
2.3. Correction of Dust Pollution
 The raw signal will be attenuated when dust particles adhere to the surface of the detector. Although we cleaned the surface of the detector during the IOP, we calibrated the output signal and corrected the data after each dust event.
 After every observation period, we performed a data calibration test using a calibration tool. This tool has rotating strings equivalent to particle sizes of 138, 226, 252, 319, 400, and 451 μm with a fixed rotational frequency of 3000 rpm. When a string crosses the sensor slit, we can check the signal output and correct it if necessary. In fact, the output of the data processing unit was quite stable during the field observation period.
 The field campaign was a part of ADEC IOP1 [Mikami et al., 2002]. Figure 3 presents the site location map of ADEC IOP1. The saltation observation was made on the outskirts of Qira, a small oasis in the southern edge of the Taklimakan desert, China. We have set two observation systems on the gobi desert (36°54′07″N, 80°47′07″E; “gobi site” in Figure 4) and on the sand dune (36°56′11″N, 80°45′47″E; “dune site” in Figure 4). The observation sites are located on a slope of the Kunlun Mountains with an inclination of about 1/100. The ground surface condition of the gobi site is gobi, consisting of sand and pebbles in a topographically flat condition. The dune site is located in a sand dune area, and its undulation is quite large. We surveyed the surface undulation of 30 m by 30 m areas around the observation points using a laser surveying instrument (Leica Geosystems TPS-400). The standard deviations of the surface undulation within the area were 9.2 cm for the gobi site and 121.4 cm for the dune site. The surface soil textures of the sites are 0% clay, 1% silt, and 99% sand for the dune site and 1% clay, 31% silt, and 68% sand for the gobi site.
 In order to monitor the whole saltation process, we collected the meteorological conditions, visibility at two levels, and surface soil physical conditions, in addition to the measurements of the saltating sand particles. Mikami et al.  made a long-term meteorological observation around this area and found a typical diurnal wind variation between the Kunlun Mountains and the desert area. However, even when the katabatic wind blows during nighttime, the maximum wind speed at 10 m height is around 5 m s−1. The diurnal wind circulation is thus unlikely to induce a dust outbreak. They also found that strong westerly winds, exceeding 7 m s−1, blow frequently from May to June and that dust storms are often accompanied by such strong westerly winds. Our sensor instruments, therefore, faced west in order to monitor dust outbreaks and their related meteorological elements such as wind speeds, wind directions, air temperature, and visibility. The fetches of both sites are long enough for the observation. The upwind distances of the dune sites are several kilometers from west to north, 500 m from north to east, and 200 m from east to south, and 200 m from gobi south to west. The upwind distances of the sites are several kilometers from west to north, 500 m from north to east, and 200 m from east to south, and several kilometers from south to west.
 Specifications of the observation system are listed in Table 2. During ADEC IOP1, all the data, except for SPC data and sand catcher samplings, were collected for every minute and stored in the data logger (Campbell Scientific Co. CR-10X). The time interval of SPC was 1 s, and the sample of the sand catcher was taken just after each dust event. As Kurosaki and Mikami  pointed out, snow cover and surface soil water are important controlling factors on wind erosion, especially during the spring season in northern China. However, during IOP1, the average volumetric soil water content at the dune site and the gobi site were 0.03% (14 samples, 0–1 cm in depth) and 0.27% (24 samples, 0–1 cm in depth), respectively. Still, we do not have a concrete understanding about the soil moisture effect on the threshold value of the saltation, as referred to in some experiment results [McKennna Newman and Nickling, 1989; Fécan et al., 1999]. The effect of the soil moisture on dust outbreaks during the IOP1 can be negligible in such a hyperarid condition.
Table 2. Specifications of the Meteorological Observation
Start Time, UT
End Time, UT
WS (3.6, 5.7, 8.7 m), WD (8.6 m), Tair (3.6, 5.5, 8.5 m), RH (3.6, 5.5, 8.5 m), SWR (up and down ), LWR (up and down), SoilT (5, 20, 40 cm), G (11 points), Swater (11 points), rain sensor, and visibility meter (2.4, 9.1 m).
WS (0.45, 1.38, 3.8 m), WD (3.8 m), Tair (0.45, 1.24, 3.48 m), RH (0.45, 1.24, 3.48 m), SWR (up and down), LWR (up and down), SoilT (2, 5, 20 cm), G (11 points), Swater (11 points), rain sensor, visibility meter (1.5, 2.9 m).
 ADEC IOP1 was conducted from 8 to 21 April 2002. During this period, all the observation sites simultaneously monitored a dust storm from its outbreak to its long-range transport to Japan. Figure 5 illustrates the time variation of wind speed at 3.8 m height and visibility at 2.9 m height on the gobi surface from 1 to 21 April. We used two criteria to select dust outbreak events from the time series of wind speed and visibility measurements: (1) the wind velocity at 3.8 m height exceeds 7 m s−1 and (2) the visibility at 2.9 m height is less than 500 m. The former criterion is from the observational result of wind-blown sand in this region [Nagashima et al., 1991]. According to their result, the sand on the sand dune in this region moved when the surface wind exceeded 7 m s−1. The latter criterion is taken from that of a heavy dust storm or sandstorm of the present weather code (WW = 33, 34, and 35) of the World Meteorological Organization (WMO). From 1 to 21 April, a total of eight dust events were classified (arrows 1 to 8 in Figure 5). According to the statistical analysis of the Qira meteorological observatory data, the average seasonal total numbers of days with a sandstorm and dust storm in Qira were 0.9 in the winter, 8.9 in the spring, 7.7 in the summer, and 1.2 in the autumn [Yoshino, 1997]. Thus dust outbreaks from 1 to 21 April were much more frequent than the average. This is also consistent with the fact that the dust outbreak frequency in this season was higher than that in normal years in the eastern Asian continent [Kurosaki and Mikami, 2003].
 From these events, we selected two cases, the 5 April case and the 14 April case, for analysis. On 5 April, a weak dust storm occurred and lasted for several hours (from 1223 to 1430 UT; arrow 3 in Figure 5). In this case, we set three SPCs, one in the sand dune and the other two in the gobi site. However, the intensive observation period had not started yet, so only 30-min average data were available for the meteorological elements. In the 14 April case, a severe dust storm occurred from 0327 to 0830 UT, and low visibility (below 500 m) lasted for several days (arrow 7 in Figure 5). We measured the meteorological elements on the gobi desert with 1-min intervals, although the SPC observation was conducted only at 30 cm height in the gobi site. The other six cases lacked either SPC data or meteorological data. For this reason, we omitted these cases from the analysis.
4.1. The 5 April Dust Event
Figure 6 illustrates the time variation of the total number of saltation particles taken every minute for 84 and 169 μm bin classes at 30 cm height at the gobi site. We can also get the particle numbers for the other bins in the same manner. Figure 6 reveals that the dust storm lasted for about an hour and then the saltation flux gradually decreased to zero. The time sequence of the 30-min average wind speed at 3.8 m at the gobi site exhibits similar changes, such as 7.85 m s−1 from 1201 to 1230 UT, 11.52 m s−1 from 1231 to 1300 UT, and 9.27 m s−1 from 1301 to 1330 UT. The 30-min average of visibility at the gobi site had a minimum value, 324 m, at 1300. Figure 7 shows the time variations of the saltation particle numbers on the gobi site at 30 cm height in 1-s intervals (thin line in Figure 7) and 21-s moving averages (thick line). The fluctuation of saltation flux, which is caused by the turbulent motion of the wind field close to the ground surface, is well represented. These results indicate that the SPC represents the detailed saltation process well with high time and size resolutions.
 During this event, we placed three SPCs in the field. One was at the dune site whose observation height was 30 cm above the ground surface. The other two SPCs were placed on the gobi site. One, gobi-H, was set at 30 cm height, and the other, gobi-L, was set at 20 cm height. Figure 8 illustrates the number size distributions of saltation particles at the dune site and the gobi sites. The particle numbers of the gobi sites are more than 10 times greater than that of the dune site for all classes of particle sizes ranging from 30 to 667 μm.
 We can evaluate the saltation flux, g m−2 s−1, of sand particles that passed through the sensor grid, 2 mm × 25 mm, from the summations of particles in every size class. We set the soil density as 2.71 g cm−3 on the basis of soil sampling measurements made at the sites. Figure 9 shows the resultant saltation fluxes at 30 cm height at the dune site, and at two heights, 30 and 20 cm, at the gobi site. The fluxes at the gobi sites are very similar, although the lower sensor shows slightly smaller values. In contrast, the flux at the dune site exhibits extremely smaller values compared with the gobi sites.
4.2. The 14 April Dust Event
Figure 10 indicates the time variation of total saltation flux at the gobi site, 30 cm height. A dust outbreak started from 0327 UT and lasted for about 5 hours. Unlike the 5 April case, the saltation flux gradually increased and reached a maximum value 40 min after the dust occurrence. During this event, the total amount of saltation flux measured at the gobi-H site was 8.95 kg m−2.
 Since we measured the meteorological elements every minute during the event, we can compare the relation between wind speeds and saltation flux. According to the result, we can clearly confirm that a threshold value of the wind velocity controls the occurrence of dust outbreak (Figure 11). In this case, the threshold wind velocity at 3.8 m height was 7.5 m s−1.
5.1. Superpose Effect of the SPC Sensor Under High Number Density of Saltation Particle
 The SPC sensor counts the particles within a measuring space (25 mm × 2 mm × 0.5 mm) between source and detector units. If particles within this volume are extremely dense, particle silhouettes viewing from detector units, with cross-sectional area of 2 mm × 0.5 mm, superposed each other. This will result in an under estimation of the particle number and over estimation of the particle size. To check this possibility in SPC, we made an examination under severe dust storm condition on 1236 UT 5 April 2002. The total saltation flux at this moment was 68.63 g cm−2 s−1 and the wind speed at 45 cm height was 8.0 m s−1. If we assume the density of dust as 2.71 g cm−3, the resultant volume of dust within a measuring space is 0.314 × 10−5 cm3 cm−3 × 2.5 × 10−2 cm3 = 0.78 × 10−7 cm3.
 If we assume that all the saltation particles are spherical with 10 μm diameter, number of the dust particle within a measuring space is 149. Then total cross-sectional area of all the particles is 1.17 × 10−4 cm2 assuming every particle do not superposed. As the cross-sectional area of the detector is 1.0 × 10−2 cm2, the percentage of the saltation particle silhouettes area is only 1.2% of the cross-sectional area of the detector. If we assume the saltation particle size as 100 μm diameter, this result decreases to 0.12%. Thus, in the actual condition, the influence of the superpose effect can be negligible.
5.2. Streamwise Saltation Flux
Table 3 presents the total streamwise saltation fluxes during the dust events in the 5 April case, 1223 to 1430 UT, and the 14 April case, 0327 to 0830 UT. Although the 5 April case does not last as long as the 14 April case, the total streamwise saltation flux of the 5 April case is larger than that of the 14 April case. This means that the dust storm on 5 April was stronger than that on 14 April. This difference in timescale and strength is caused by the difference of the structure of the atmospheric disturbances that gave rise to the dust storm. As mentioned in section 2, the sites are located in the southern edge of the Tarim basin, which is surrounded by steep mountains on three sides, south, west, and north. For this complicated topographic condition, the mechanism of dust storm in this region will have a peculiar characteristic. In order to understand this, analytical and numerical studies have been made [Mikami, 1997; Aoki, 2003]. Aoki  made a numerical simulation of dust outbreak in the Tarim basin and found that some dust storms are caused by the advection of the cold air mass from the northeast accompanied by the passing of a low-pressure trough around the Tarim basin. On the other hand, Mikami  made a case study of dust storms around this region and showed that a mesoscale gust front passes through this region from the west and a strong dust storm occurs. He also pointed out that the topography of the slope of the Kunlun Mountains would play an important role in enhancing strong wind. The horizontal scale of the disturbance is smaller than the former one, and the time duration is shorter. The 5 April case may be caused by a mesoscale disturbance, although we do not have detailed synoptic meteorological information to confirm this.
Table 3. Total Saltation Flux at the Dune Site and the Gobi Sitesa
5 April (1223–1430 UT)
14 April (0327–0830 UT)
Total Flux, kg m−2
Maximum Flux, g m−2 s−1
Total Flux, kg m−2
Maximum Flux, g m−2 s−1
Maximum fluxes indicate the maximum instantaneous saltation flux during each event.
5.3. Saltation Fluxes at the Gobi and Dune Sites
 As shown in Figure 12, the saltation fluxes at the sand dune and gobi sites differ quite significantly. The total fluxes during this dust event from 1223 to 1430 UT were 2.61 kg m−2 for the dune, 37.93 kg m−2 for the gobi-H site, and 43.71 kg m−2 for the gobi-L site (Table 3). The gobi and dune sites are separated by about 4 km, so the synoptic meteorological conditions seem to be the same. Thus it is difficult to account for this difference as the synoptic meteorological conditions. For both sites, the parent soil particle size distributions (PSD) were obtained by laboratory experiments using the samples taken at 0 to 1 cm depth at each site. We used the laser diffractometry method with microtrac FRA (Leed and Northrup Company Ltd.). The analytical range of the soil particle size is 0.12 to 704 μm with 50 channels. The soil samples were fully dispersed in a solvent of hexametaphoshate. Figure 13 illustrates the size distributions for the sand dune and the gobi. The size distribution of the gobi is represented by the average of three samples taken at around the AWS. The maximum distribution of the dune is 200 μm, which is considerably larger than the maximum of the gobi (80 μm). According to the Bagnold and the Greeley-Inversen saltation scheme, the minimum threshold friction velocity for individual particle is around 80 μm [Greely and Inversen, 1985; Shao, 2000]. This means that saltation occurs most effectively at sizes around 80 μm. Accordingly, this difference of PSD is the main reason for the large difference of the saltation fluxes between the sand dune and gobi. Most of the ground surface in the Taklimakan desert is sand dune, and it is mainly located in the center of the Tarim basin. The gobi is located on the outskirts of the Taklimakan desert. However, if the above relation is common in the Taklimakan desert, dust outbreaks seem to be dominant not only in the central part but also in the outskirts of the basin.
5.4. Size Distributions of the Saltation Particles
Figure 14a illustrates the normalized number density of saltation particles collected during the 5 April dust event at the gobi-H site and the dune site. The normalized number density n(D) is defined as
where Dmin is the observable minimum of diameter detectable by SPC (30 μm) and Dmax is the maximum of diameter detectable by SPC (667 μm). The observation heights are the same (30 cm). Although the total saltation fluxes of both sites are quite different, the normalized number densities of saltation particles indicate that both distributions are very similar in its spectral peak size and its spectral slope, from 70 to 600 μm.
 In contrast, the normalized number densities of saltation particles at the gobi-H and gobi-L sites possess slightly different spectrum shapes, as shown in Figure 14b. The maximum particle sizes are around 70 μm, which corresponds well with the particle size of the minimum threshold friction velocity predicted by the Bagnold and the Greeley-Iversen's schemes [Bagnold, 1941; Greely and Iversen, 1985; Shao, 2000]. However, the normalized particle number density at the gobi-L site indicates that the coarse particles from 117 to 554 μm are slightly larger than the number density at the gobi-H site. Although this height dependency of the saltation particle number density cannot be explained by the present theory of saltation, this variation of particle size distribution with height is reasonable from a physical point of view. A larger saltation particle will drop faster than a smaller particle owing to its greater terminal velocity. In order to fully explain the characteristics of the saltation process, a new saltation theory must be developed.
 In order to understand the mechanism of dust outbreak from arid region, we made intensive observations in the south of the Taklimakan desert, China. The purpose of the observations is to monitor the saltation processes both on the gobi and sand dune. For this, we have developed a new sand particle counter, SPC. The sensor part of the SPC is based on the snow particle counter that uses a luminescent diode laser in the transmitter. The SPC enables us to measure saltation particles from 30 to 667 μm in diameter with 32 bin classes and measure the particle number of each bin class every second. We have tested this SPC in the field, and it proved to be useful for the saltation study when the temperature dependency, edge effect, and dust pollution were properly corrected.
 Eight dust events occurred during the intensive observation from 1 to 21 April 2002. Of these eight events, we analyzed the 5 April case and 14 April case. The results are summarized as follows.
 The total saltation fluxes at 30 cm height during the 5 April case (1223 to 1430 UT) for the gobi was 37.93 kg m−2 and for the dune was 2.61 kg m−2. The saltation flux at the dune site was only 7% of that at the gobi site, although the distance between the sites is only 4 km. The reason of this was to be the difference in the parent soil size distributions: that is, the number of particles around 80 μm was considerably greater for the gobi surface than for the dune.
 The threshold wind velocity at 3.8 m height in the gobi site was 7.5 m s−1 from the observation of the 14 April case (0327 to 0830 UT). In this case, the total saltation flux was 8.95 kg m−2 at 30 cm height.
 The saltation particle size distributions at the gobi sites in the 5 April case indicated that the number size distribution of the coarse particles ranging from 117 to 554 μm at 20 cm height was greater than that at 30 cm height. This is reasonable from the physical point of view, because the higher the height, the fewer the number of coarse saltation particles owing to the gravitational effect. However, the present saltation theory cannot explain this particle size dependency on height.
 The authors would like to thank H. Nagashima of Tokyo Marine University, Y. Shao of Hong Kong City University, K. Okada, and S. Kurita of Meteorological Research Institute of Japan for their helpful discussions and comments. This observation is made as a part of the Aeolian Dust Experiment on Climate Impact (ADEC) project sponsored by the Ministry of Education, Culture, Sports, Science and Technology of the Japanese Government.