Lidar Observation and Numerical Simulation of a Kosa (Asian Dust) over Tsukuba, Japan during the Spring of 1986

Transcription

1 June 1988 K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 457 Lidar Observation and Numerical Simulation of a Kosa (Asian Dust) over Tsukuba, Japan during the Spring of 1986 By Kenji Kai*, Yoshitaka Okada*, Osamu Uchino*, Isao Tabata*, Hajime Nakamura**, Toshikatsu Takasugi** and Yoshinobu Nikaidou* *Meteorological Research Institute, Tsukuba, Ibaraki 305, Japan **Japan Meteorological Agency, Tokyo 100, Japan (Received 17 December 1987, in revised form 23 April 1988) Abstract A number of duststorms and/or sandstorms occurred in the deserts and loesslands of the Asian Continent in early March of After a few days the dust laden air was transported over the Yellow Sea to Japan by westerly winds. On March, a number of Japanese meteorological observatories reported a "Kosa (Asian dust)" phenomenon. The lidar observation of the Kosa was made at Tsukuba, Japan from 15 JST to 21 JST on 13 March The vertical structure and time change of the Kosa layer observed by the lidar are presented. At 15 JST, two Kosa layers existed at 4km and 2km, respectively. The upper layer had a thickness of about 1km and a scattering ratio of 3.2. The lower layer had a scattering ratio of 2.6 and appeared to be mixed with background aerosols. Subsequently, the Kosa layer at 4km increased in thickness and scattering intensity, with a thickness of 1.5km and a scattering ratio of 5.7. At 18 JST the Kosa layer at 4km separated into two sublayers at 4.5km and 3.5km. The total thickness of the upper and lower sublayers was 2.3km. The lidar derived optical thickness was (wavelength nm). From 18 to 20 JST, the Kosa layer gradually lowered 0.5km. At 20 JST the Kosa layer separated into three sublayers at 4.0km, 3.2km and 2.7km. Analysis of concurrent radiosonde data showed that the upper and lower sublayers were dry, while the middle sublayer was humid. A numerical simulation was carried out to investigate the long range transport of the Kosa particles. Simulated horizontal and vertical distributions of the tracers were in good agreement with the lidar observation at Tsukuba and the routine meteorological observations in Japan and China. In particular, the observed structure of the two Kosa layers was well simulated. The two Kosa layers were found to originate from different altitudes over the source regions. The numerical simulation reveals the Loess Plateau and its neighboring deserts as important sources for the Kosa. Another possibility includes the Takla Makan Desert. Travel time of the Kosa particles to reach Japan was two to three days from the Loess Plateau and its neighboring deserts, and five to six days from the Takla Makan Desert. 1. Introduction Dust originating from the Asian Continent and transported to Japan is known as a "Kosa (Asian dust, *)". In late winter and spring, the aerosols associated with a Kosa event play an important role in cloud physics, and optical and radiative transfer processes in the atmosphere. c 1988, Meteorological Society of Japan Isono et al. (1959) showed that the Kosa particles are active ice nuclei in the atmosphere. Hirose et al. (1983) evaluated the effect of the Kosa on the chemical composition of maritime aerosols over the North Pacific Ocean. Arao and Ishizaka (1986) estimated the volume and mass of the Kosa particles by using a size distribution model, and showed a significant peak in the range between 0.5 and 1.0*m radius. Okada et

2 458 Journal of the Meteorological Society of Japan Vol. 66, No. 3 al. (1987) studied the morphological features of the Kosa particles. Lidar is an active remote-sensing instrument that provides information about the fine spatial distribution of aerosols in the atmosphere. Iwasaka et al. (1983), Kobayashi et al. (1985) and Kobayashi et al. (1987) made case studies of Kosa layers by use of lidar. However, investigations of the vertical structure and time change of the Kosa layer appear to be in the preliminary stages, and further observational data will eventually be required. A number of duststorms and/or sandstorms occurred in the deserts and loesslands of the Asian Continent in early March of After a few days the dust laden air was transported over the Yellow Sea to the Japanese Islands by westerly winds. On March 1986, a number of Japanese meteorological observatories reported a "Kosa" phenomenon, which included reports of the fall of dust particles, a decrease in visibility, a yellow sky, etc. Lidar observation of the Kosa was made at Tsukuba, Japan for six hours on 13 March The vertical structure and time change of the Kosa layer observed by the lidar are presented in this paper. In addition, the long range transport of the Kosa particles is numerically simulated by a pollutant tracer model developed by Nakamura and Takasugi (1987). The Kosa also acts as a good tracer for studying tropospheric circulation. Recently, the long range transport of the Kosa has been a subject of concern (e.g., see Duce et al., 1980). Shaw (1980) reported an example of a long range transport of the Kosa to the Hawaiian Islands, and Uematsu et al. (1983) estimated that 6-12* 106 tons of particles from the Kosa are transported annually to the central North Pacific. Merrill et al. (1985) presented a model which showed the atmospheric transport of the Kosa to the Marshall Islands. In the previous studies (e.g., Iwasaka et al., 1983; Merrill et al., 1985), by assuming the conservation of potential temperature, the trajectory of the particles was computed from information of the two-dimensional wind field on an isentropic surface. In the present study, a large number of marked particles released from a source are directly traced by using the three-dimensional wind field supplied by the JMA global spectral model (NWPD/JMA, 1986). The numerical simulation is compared with the lidar observation at Tsukuba and routine meteorological observations taken in Japan and China. 2. Lidar observation The lidar observation was made at the Meteorological Research Institute at Tsukuba, Japan (36*04'N, 140*07'E, 24.5m MSL) from 15 JST to 21 JST on 13 March The observation provided 21 profiles of aerosol backscattering data. Profile data of temperature, humidity and pressure were obtained by radiosonde soundings at the Tateno Aerological Observatory near the institute. Figure 1 shows the lidar system used by the Meteorological Research Institute. The system is composed of a ruby laser transmitter, a receiving telescope, a signal processor and a controller. The wavelength of the pulsed laser beam is 694.3nm. The energy is approximately 2J per pulse, and the pulse repetition rate is 2ppm. A mechanical chopper is used to eliminate problems due to the fluorescence of the ruby rod. The backscattered photons are collected by the Cassegrain telescope with a diameter of 80cm, and detected by the photomultipliers. The signals of the photomultipliers are digitized by a transient waveform recorder with a sampling frequency of 10MHz, which corresponds to a maximum height resolution of 15m. The lidar signals are then fed into a microcomputer and stored on a floppy disk. The lidar signals were used to calculate the scattering ratio according to the standard method (Russell et al., 1976). The scattering ratio, R, is defined as follows: where *m(z) and *a(z) are, respectively, the molecular and aerosol backscattering coefficients at height z. The molecular coefficient *m(z) is determined from the radiosonde density profile closest in time to the lidar observation as:

3 June 1988 K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 459 Fig. 1. Schematic diagram of the lidar system of the Meteorological Research Institute (MRI). where d*(*)/d*=2.11*10-28 is the Rayleigh backscattering cross-section of an air molecule at the ruby wavelength (694.3nm), and Nm(z) is the molecular number density at height z. The optical thickness of the Kosa layer *(z1, z2) was derived from the lidar data, by integrating the aerosol extinction coefficient (z) from the base z1 to the top *ex z2 of the Kosa layer: Substituting Eqs. (1) and (2) into Eq. (3) and using the conversion value of A=*ex(z)/*a(z) for the ratio of the extinction coefficient to backscattering coefficient, Eq. (3) is rewritten: The optical thickness is calculated by Eq. (4), and as can be seen, the lidar derived optical thickness depends on the choice of A. Since the size distribution, refractive index and physical shape of the Kosa particles are not known, a value of A=50 is assumed to derive the optical thickness (Collis and Russell, 1976). Another source of error is due to omitting the extinction term in deriving the scattering ratio. The error is estimated to be about 15%. 3. Source region for the Kosa and meteorological conditions 3.1 Source region for the Kosa Figure 2 shows the geographical distribution of desert and loess over the Asian Continent and the observation site at Tsukuba (the Geographical Department of the Northwest Normal College, 1984; Zhao, 1986). The source region for the Kosa is the vast arid region between 33* and 47*N, and 75* and 115*E where the Takla Makan, Gurbantungut, Badain Jaran, Tengger, Gobi and Ordos Deserts and the Loess Plateau lie. The loess is distributed in the Hwang-Ho Basin. It is mainly composed of loose silt that is easily subjected to erosion (Zhao, 1986). The Gobi Desert on the borders of China and Mongolia is mainly a gravel desert as opposed to a sandy desert. One of the most notable characteristics of late winter and spring in this region is the large amount of dust and sand in the atmosphere

4 460 Journal of the Meteorological Society of Japan Vol. 66, No. 3 Fig. 2. Geographical distribution of deserts and loesslands in the Asian Continent and the observation site. (Arakawa, 1969). Duststorms and/or sandstorms occur most frequently in this season, when low rainfall and increased surface winds associated with cold fronts contribute to their occurrence (Ing, 1972). At times, the dust is transported to Japan during this season. The observation site, Tsukuba, is about 3000km east of the Hwang-Ho Basin, and about 5500km east of the Takla Makan Desert. 3.2 Meteorological conditions On March 1986, a number of Japanese meteorological observatories reported the Kosa phenomenon. Surface synoptic charts and meteorological reports are examined during the period of concern in order to establish the meteorological conditions for the Kosa phenomenon. Figure 3 shows a series of surface synoptic charts from 8 to 13 March. Low pressure appeared in the northern part of the China Plain on 8 March, while Japan was under the influence of an anti-cyclone. On 9 March the low pressure area moved to the Yellow Sea. From 10 to 11 March, the low along with an accompanying cold front moved eastward and passed over Japan, while an anti-cyclone moved to central China, following the cold front. Another low existed in the north-east China. On 12 March the anti-cyclone moved eastward over the Yellow Sea and reached the Kyushu. By the 13th of March the anti-cyclone covered western and central Japan. This anti-cyclone brought the Kosa to Japan, and on both the 12th and 13th of March, the Kosa phenomenon was reported at meteorological observatories, mainly in western Japan. Figure 4 shows the geographical distribution of the dates of occurrence of the duststorms, sandstorms and Kosa during 8 to 13 March. The data source is the meteorological messages collected by the Japan Meteorological Agency. The integer in the figure indicates the date of occurrence. The elevation of the hatched area is greater than 3000m. Sandstorms occurred in the Takla Makan Desert throughout the period. A cluster of duststorms occurred in the Loess Plateau and its neighboring deserts, i, e., the Ordos, Tengger, Badain Jaran Deserts on March. By taking into account the meteorological conditions and the geographical distribution of deserts and loesslands, it is suggested that the source region for the Kosa on March was the Loess Plateau and its neighboring deserts.

5 June 1988 K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 461 Fig. 3. Surface synoptic charts at 00 GMT during 8-13 March Fig. 4. Geographical distribution of the dates of the occurrence of the duststorms, sandstorms and Kosa during 8-13 March The evidence for this is as follows: (1) The Kosa phenomenon was reported in Japan on 12 to 13 March after a cluster of the duststorms occurred in the Loess Plateau and its neighboring deserts on March. (2) The speed of the eastward movement of the low pressure area and the anti-cyclone was about 10 degrees of longitude per day.

6 462 Journal of the Meteorological Society of Japan Vol. 66, No. 3 (3) The loess is a very fine textured soil, which can be easily lifted up into the atmosphere. Another candidate for the source region appears to be the Takla Makan Desert since sandstorms occurred there from 8 to 13 March. Iwasaka et al. (1983) pointed out the possibility of the Takla Makan Desert as a source region. They showed by use of lidar observation and air mass trajectory analysis that the two layers over Nagoya came from distinctly different sources: the lower layer from the Hwang-Ho Basin and the upper from the Takla Makan Desert. However, Murayama et al. (1984) and Arakawa (1969) presented evidence to the contrary. According to Murayama et al. (1984), the particles that originate in the Takla Makan Desert are too large to be dispersed over wide areas. In addition, sandstorms in the Takla Makan Desert can not affect such a wide area as those generated in the Gobi and Ordos Deserts because of the confining mountain ranges (Arakawa, 1969). In Section 5, a numerical simulation will be carried out to identify the source region for the present case. 4. Results and discussion of the lidar observation 4.1 Lidar signal and scattering ratio The lidar observation started at 15:03 JST and ended at 20:48 JST on 13 March The 21 vertical profiles of lidar signals are shown in Fig. 5(b), and the profiles of temperature and relative humidity before and after the lidar observation are shown in Figs. 5(a) and (c), respectively. In addition, the lidar signals are analyzed in terms of their scattering ratio defined by Eq. (1) and are shown in Fig. 6. In both Figs. 5 and 6, the boundary of the Kosa layer is indicated by dotted lines. The vertical structure and time change of the Kosa layer are described as follows. At 15:04 JST, two scattering layers existed at 4km and 2km. The upper layer had a thickness of about 1km and a scattering ratio of 3.2. There was no cloud over the site according to visual observations from the ground. This dust layer was identified as the main Kosa layer due to the distinct scattering layer around 4km although it had not been found in routine observations. The lower layer had a scattering ratio of 2.6 and appeared to be mixed with background aerosols. Subsequently, the Kosa layer at 4km increased in thickness and scattering intensity. At 16:46 JST it had a thickness of 1.5km and a scattering ratio of 5.7. At 18:03 JST the Kosa layer at 4km separated into two sublayers: the upper sublayer with a scattering ratio of 2.9 was centered at a height of 4.,5km, and the lower sublayer with a scattering ratio of 5.1 was centered at a height of 3.5km. The total thickness, including the upper and lower sublayers was 2.2km. From 18 to 20 JST, the entire Kosa layer gradually lowered 0.5 km. During this time, the lower layer at 1.5km became weaker, and its scattering ratio was less than 2.0. At 20:14 JST the lower sublayer of the Kosa layer further separated into two sublayers, Fig. 5. Profiles of lidar signals, temperature and relative humidity on 13 March 1986.

7 June 1988 K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 463 Fig. 6. Profiles of the scattering ratio on 13 March creating three sublayers at 2.7km, 3.2km and 4.0km. This vertical structure of three sublayers continued until the end of the observation. On the other hand, the lower layer at 1.5km experienced an increase in its scattering ratio. At 20:37 JST there was a large peak (scattering ratio=46) at 5.5km height, which was due to the passage of clouds over the site. A comparison of the lidar profile at 20:37 JST with the radiosonde profile at 20:30 JST shows that the upper sublayer at 4km and the lower sublayer at 2.7km were dry, while the middle sublayer at 3.2km was humid. Ing (1972) reported, by analysis of radiosonde data, that duststorms originating over central China were associated with an absence of moisture in the troposphere. However, the middle sublayer of the present study was humid. The temperature and relative humidity at 3km were approximately -9* and 80%, respectively. This raises the question as to why was the middle sublayer humid? The following is considered: (1) The Kosa particles originating over the Asian Continent might have mixed with a humid air mass while being transported to Japan. (2) The Kosa particles might have been modified to act as condensation nuclei or ice nuclei (Isono et al.,1959). It is beyond the scope of the present study to answer the above question, however it should be noted that a Kosa layer is not always dry. 4.2 Summary of the lidar observation Figure 7 summarizes the vertical structure and time change of the upper Kosa layer. The heights of the base and top of the Kosa layer are found from Fig. 6, and are plotted in this figure. The Kosa layer and the cloud are shaded. The optical thickness is derived by use of Eq. (4), integrating from the base to the top of the Kosa layer, and is Fig. 7. Time change of the Kosa layer and the optical thickness on 13 March 1986.

8 464 Journal of the Meteorological Society of Japan Vol. 66, No. 3 also plotted in the figure. The Kosa layer at the height of 4km had a thickness of about 1km at 15 JST, and then it gradually thickened. At 18 JST the Kosa layer separated into two sublayers and had a maximum value of optical thickness (about at nm). From 18 to 20 JST, the entire Kosa layer gradually lowered 0.5km. At 20 JST the Kosa layer further separated into three sublayers. The optical thickness was less than A temporal cloud passed over the site at 20:45 JST. The values of the lidar derived optical thickness are rather small. This indicates that the Kosa layer was optically thin owing to the fact that the site is far away from the source region. The Kosa observations were concentrated in western Japan as shown in Fig. 4, and were not observed in eastern Japan. This suggests that the Kosa may be undetectable from ground visual observations in eastern Japan. Another reason may be the accuracy of the conversion factor A in Eq.(4). The absolute value of the lidar derived optical thickness depends on this conversion factor A. It would be possible to specify a more accurate value of A if the size distribution, refractive index and physical shape of the Kosa particles were known. It is also important to determine a value of A experimentally. 4.3 Observational model of the Kosa Figure 8 shows a schematic illustration of the Kosa along 36*N from 100*E to 140*E on 13 March The duststorm that occurred in the deserts and loesslands of the Asian continent is shown in the figure. According to Murayama and Kimura (1984), the dust may be lifted up to 1-2 km, and at times as high as 4km. Then the dust was transported across the Yellow Sea and the Sea of Japan to Japan by the westerly winds. The time change of the Kosa structure shown in Fig. Fig. 8. An observational model of the Kosa on 13 March is converted into a spatial change by using the mean vertical profile of zonal component of the winds. The zonal component is calculated by averaging the wind profiles at Tateno (36*03'N, 140*08'E), Wajima (37*23'N, 136*54'E) and Yonago (35*26'N, 133*21'E) at 0 and 12 GMT on 13 March The zonal components found, for example, are 11.4m/s at 2km, 18.1m/s at 3km, 28.5m/s at 4km and 33.6m/s at 5km (westerly is positive). The resulting Kosa structure is schematically shown and shaded in the figure. The dotted line is interpolated. The spatial form of the Kosa appears as a tongue. The figure illustrates the stage at which the tip of the tongue-like Kosa moves over Tsukuba. The tip of the Kosa is at a height of 4 km and has a thickness of about 1km. At this stage the main part of Kosa is about 360km behind the tip and will be transported over the site in a few hours. 5. Results and discussion of the numerical simulation In order to investigate the long range transport of the Kosa particles, a numerical simulation is carried out using the pollutant tracer model developed by Nakamura and Takasugi (1987). The present paper is limited to the Kosa, for more details of the model and its application see Takasugi and Nakamura (1988). 5.1 Outline of the pollutant tracer model The pollutant tracer model (PTM) uses the three-dimensional wind field and diffusion coefficients supplied from the operational numerical weather prediction model of the JMA, which is a global spectral model (GSM) with 12 vertical levels and a 42 wavenumber triangular truncation. The performance and systematic errors of the GSM can be found in NWPD/JMA (1986) and Kanamitsu et al. (1983). The PTM computes the three-dimensional Lagrangian motion and vertical diffusion of tracers released from a source. The vertical diffusion is described in the PTM as a diffusion coefficient. The terminal velocity of the tracers is not incorporated in the PTM. A full description of the PTM can be found in Nakamura (1987). A 'cyclic run' of the PTM coupled with the

9 June 1988 K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 465 GSM is performed during the study period in order to avoid a decline in the accuracy of prediction. This means that the initial data of the GSM are cyclically renewed by observed data every 12 hours. Therefore, the wind field used in the PTM is close to that observed. According to the discussion in Section 3, the source regions of the tracers are chosen as two target areas: one is the Ordos Desert (40*N, 106*E) and the second is the Takla Makan Desert (40*N, 80*E). 5.2 Case 1: Column source on the Ordos Desert In this case, the source of the tracers is a rig. 9. Predicted three-dimensional distributions of the tracers for Case 1. Vertical line segments at the corners indicate the height of 800mb.

10 466 Journal of the Meteorological Society of Japan Vol. 66, No. 3 Fig. 10. Predicted horizontal distributions of the tracers for Case 1.

11 June 1988 K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 467 column of 4km in height, having a radius of 200 km, and is located on the Ordos Desert. This column released 101 tracers every hour, from 00 GMT 8 March to 23 GMT 14 March. The total number of the tracers was (168 hours). Figures 9-11 show the predicted threedimensional, horizontal and vertical distributions of the tracers during the study period from 8 to 13 March First, the horizontal distributions found in Fig. 10 will be discussed while referring to the three-dimensional illustrations in Fig. 9 and the meteorological conditions previously shown in Fig. 3. On 8-9 March some of the leading tracers followed along the cold front of the low near the Gulf of Chihli (Bo Hai), while most of tracers moved south-eastward to the China Plain (30-40*N, *E). On March the leading tracers along the cold front of the low passed over Japan. The remaining major portion of the tracers were separated into two groups. The majority of tracers north of about 40*N were trapped by the low pressure area in the north-east part of China, and remained there, with a few of them reaching Japan. On the other hand, the tracers south of 40*N came under the influence of the traveling anti-cyclone, and moved eastward to the Yellow Sea accompanied by descending motion as can be seen in Fig. 11. On 12 March the tracers transported by the traveling anti-cyclone reached western Japan, and spread to central and western Japan on 13 March. There is a distinct belt of high concentration of the tracers, which extends from the Howang-Ho Basin to Japan. It was these days that the JMA observatories reported the Kosa phenomenon. Tracing back to the source region, the tracers over Japan originated from the Ordos Desert two to three days before, i. e. on 10 or 11 March, when some of Chinese observatories Fig. 11. Predicted vertical distributions of the tracers for Case 1.

12 468 Journal of the Meteorological Society of Japan Vol. 66, No. 3 reported the phenomenon of duststorms, shown in Fig. 4. The simulated horizontal distributions of the tracers are in good agreement with the routine meteorological observations in Japan and China. Next, examination is made of the vertical distributions in Fig. 11. In this figure, only the tracers found in the latitudes between 30*N and 40*N are plotted in order to see the pattern near Japan more easily. As mentioned with the horizontal distributions, the tracers trapped by the traveling anti-cyclone descended from higher altitudes to the mb (1-1.5km) on March. This descending motion of the tracers can be clearly seen in the vertical distribution between 120*E and 125*E in Fig. 11(d). This motion is also illustrated in Fig. 9(c) and (d). The movement of the tracers at lower altitudes was slower, due to the lower wind speeds. On 12 March, the tracers released at later times at higher altitudes caught up with the tracers that were released earlier at lower altitudes. At this point, two layers existed between 125*E and 135*E in Fig. 11(e). This situation is clearly illustrated in Fig. 9(e) and (f). Tracing back to their origin, the tracers of the upper layer at mb (3-5.5km) were released from the source on 11 March, and those of the lower layer at mb (1-2km) were released on 10 March. On 13 March the structure of the two layers became clearer. The upper layer was found around 600mb (4km), and the lower layer was centered around 850mb (1.5km) at *E in Fig. 11(f). The vertical distribution at 140*E at 12 GMT (21 JST) on 13 March agrees well with the lidar profiles at Tsukuba (see in Fig. 6). In Case 1, both of the horizontal and vertical distributions are in good agreement with the lidar observation at Tsukuba and the routine meteorological observations in China and Japan. The numerical simulation reveals that the Loess Plateau and its neighboring deserts are important sources for the Kosa. The travel time of the tracers from the source to Japan is two to three days. 5.3 Case 2: Column source on the Takla Makan Desert In this case, a column with the same dimensions as Case 1 was located in the Takla Makan Desert. The column released 125 tracers every hour, from 00 GMT 6 March to 23 GMT 11 March. The total number of the tracers was (144 hours). Figure 12 shows the predicted horizontal and vertical distributions of the tracers released from the source on 12 GMT 13 March. The horizontal distribution is quite similar to that of Case 1. The tracers found over Japan were released from the source on 8 and 9 March. The vertical distribution also shows two layers, centered at mb (4-7km) and 850mb (1.5km). The gross features are similar to the previously simulated Case 1, but some discrepancies can be seen in the vertical distribution. The simulated upper layer found at 4-7km is higher and more dispersed than that found by the lidar observation. Fig. 12. Predicted horizontal and vertical distributions for Case 2.

13 June 1988 K. Kai, Y. Okada, O. Uchino, I. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 469 In Case 2, the horizontal distribution agrees with the routine meteorological observations in Japan and China, but the vertical distribution does not fully agree with the lidar observation. This discrepancy will be discussed later. The travel time of the tracers from the source to Japan is five to six days. 5.4 Case 3: Point source in the Ordos Desert In this case, the effect of the height of the source is examined. A point source was placed at a height of 1km in the Ordos Desert. The result of the numerical simulation is shown in Fig. 13. The gross features of the horizontal distribution are similar to those of Case 1 and Case 2, however the vertical distribution is different. Case 3 simulates only the lower layer at 850mb (1.5km), and it does not reproduce the upper layer such as that found in Case 1 and Case 2. This result shows that the lower layer at 1-2 km found over Japan was formed by the tracers released at lower altitudes (about 1km), and Fig. 13. Predicted horizontal and vertical distributions for Case 3. conversely, that the upper layers in Case 1 and Case 2 were formed by tracers released at altitudes higher than 1km. The two simulated layers in Case 1 and Case 2 originated from different altitudes over the source region. 5.5 Summary and discussion The numerical simulation (Case 1) with the 4- km column source located in the Ordos Desert agrees well with the lidar observation at Tsukuba and routine meteorological observations in Japan and China, both in the horizontal and vertical distributions. In particular, the observed structure of the two Kosa layers is well simulated. It was concluded that the Loess Plateau and the neighboring deserts were the major sources for the Kosa of the present study. Another possibility includes the Takla Makan Desert (Case 2). However, the case of the source region being located in the Takla Makan Desert is slightly different from the lidar observation. Case 3 shows that the two layers originated at different altitudes over the source regions. The travel time of the Kosa particles to reach Japan is two to three days from the Loess Plateau and its neighboring deserts, and five to six days from the Takla Makan Desert. The above gives a summary of the numerical simulation. Next, some discrepancies between the simulation and the observations will be considered. As previously mentioned, the terminal velocity of the tracers is neglected in the PTM. According to Kasten (1968), the terminal velocity varies rapidly with particle size, from 0.01cm/s for 1-*m particles to *1cm/s * for l0-*m particles. For example, particles with a radius of r=1*m fall *10m per day, particles of r=10*m fall *1km per day, and particles of r>10*m fall out for several days. Therefore, removal by gravitational sedimentation is negligible for particles less than *5*m, but increases progressively thereafter with increasing particle size. In Case 1, the removal of the loess particles by gravitational sedimentation is negligible, because the size distribution of the loess particles peaks at 1-2*m according to Murayama et al. (1981). Thus, the agreement between Case 1 and the observations is

14 470 Journal of the Meteorological Society of Japan Vol. 66, No. 3 reasonable. In Case 2, the simulated upper layer is higher and more dispersed when compared with the lidar observation. Taking account for the gravitational sedimentation of the tracer, this difference between Case 2 and the observation may be explained below. As Murayama et al. (1984) reported, the particles of the Takla Makan Desert are greater in size than those of the Loess Plateau and Ordos Desert. Also, the path from the Takla Makan Desert to Japan (about 5500km) and the travel time of 5-6 days are long enough so that the gravitational sedimentation becomes more effective than in the case of the Loess Plateau. If the terminal velocity could be incorporated in the PTM, the height (4-7km) of the simulated upper layer in Case 2 might be lower and more consistent with the lidar observation. Case 3 shows that the two simulated layers originated from different altitudes over the source region. This result is an important mechanism for the observation of the two Kosa layers by the lidar, and may be explained as follows. Dust was lifted to about 1km over the source region on 10 March Then, the dust at mb moved slowly eastward with the traveling anti-cyclone. This dust of lower altitude origin formed the lower Kosa layer over Japan. On the other hand, dust was lifted to altitudes higher than 1km over the source region on 11 March. The dust was transported eastward by the westerly winds at about 600mb. The speed of the dust of higher altitude origin was greater than that of the lower altitude origin. As a result, the dust of higher altitude origin caught up with that of lower altitude origin over Japan on 13 March, when the structure of the two Kosa layers was observed by the lidar over Tsukuba. Finally, the long range transport of the Kosa particles and the identification of the source region are well accounted for by the PTM coupled with the GSM of the JMA. These encouraging results affirm the usefulness of the PTM in the source identification problem. 6. Concluding remarks The lidar observation of the Kosa was made at the Meteorological Research Institute, Tsukuba, Ibaraki, Japan from 15 JST to 21 JST on 13 March The observation showed the vertical structure of the Kosa layer and its time change. The results of the lidar observation are summarized as follows: (1) At 15 JST two Kosa layers existed at 4km and 2km. The upper layer had a thickness of about 1km and a scattering ratio of 3.2. The lower layer had a scattering ratio of 2.6 and appeared to be mixed with background aerosols. Subsequently, the Kosa layer at 4km increased in thickness and scattering intensity, with a thickness of 1.5km and a scattering ratio of 5.7. At 18 JST the upper Kosa layer separated into two sublayers at 4.5km and 3.5km. The total thickness of both the upper and lower sublayers was 2.3km. The lidar derived optical thickness was (wavelength 694.3nm). From 18 to 20 JST, the Kosa layer gradually lowered 0.5km. At 20 JST the Kosa layer separated into three sublayers at 4.0km, 3.2km and 2.7km. (2) A comparison of the lidar profiles at 20:37 JST with the radiosonde data at 20:30 JST showed that the upper and lower sublayers were dry, while the middle sublayer was humid. (3) An observational model of the Kosa on 13 March 1986 was derived from the results. The model showed a cross-section of the Kosa along 36*N from 100*E to 140*E. In addition, the long range transport of Kosa particles is simulated by the pollutant tracer model (PTM). The results of the numerical simulation are summarized as follows: (4) The numerical simulation (Case 1) with a 4-km column source in the Ordos Desert agrees well with the lidar observation at Tsukuba and the routine meteorological observations in Japan and China, both in the horizontal and vertical distributions. In particular, the observed structure of the two Kosa layers is well simulated. It is concluded that the Loess Plateau and the neighboring deserts were the major sources for the Kosa under consideration. (5) Another possibility includes the Takia Makan Desert (Case 2). However, the case of the Takla Makan Desert is slightly different from the lidar observation. In Case 2, the simulated upper layer is found at a higher altitude and is more dispersed compared with the lidar observation.

15 June 1988 K. Kai, Y. Okada, O. Uchino,1. Tabata, H. Nakamura, T. Takasugi and Y. Nikaidou 471 (6) Case 3 shows that the two layers at 4km and aerosol particles in spring season. J. Meteor. Soc. 2km originated from different altitudes over the Japan, 61, Ing, G.T.K., 1972: A dust storm over central China, source regions. The lower layer at 2km over April, Weather, 27, Japan was formed by the tracers released at Isono, K., M. Komabayasi and A. Ono, 1959: The lower altitudes (about 1km), and conversely, the upper layers of Case 1 and Case 2 were formed nature and origin of ice nuclei in the atmosphere. J. Meteor. Soc. Japan, 37, by the tracers released at altitudes higher than 1 Iwasaka, Y., H. Minoura and K. Nagaya, 1983: The transport and spatial scale of Asian dust-storm km. clouds: a case study of the dust-storm event of April (7) The travel time of the Kosa particles to reach Japan is two to three days from the Loess 1979, Tellus, 35B, Kanamitsu, M., K. Tada, T. Kudo, N. Sato and S. Isa, Plateau and its neighboring deserts, and five to 1983: Description of the JMA operational spectral six days from the Takla Makan Desert. model. J. Meteor. Soc. Japan, 61, Kasten, F., 1968: Falling speed of aerosol particles. J. Acknowledgments Appl. Meteor., 7, Kobayashi, A., S. Hayashida, K. Okada and Y. Iwasaka, The authors would like to express their thanks 1985: Measurements of the polarization properties to Dr. N. Murayama and Mr. T. Motoki of the Meteorological Satellite Center for their helpful of Kosa (Asian duststorm) particles by a laser radar in spring J. Meteor. Soc. Japan, 63, suggestion, and Prof. M. M. Yoshino of the, Y. Iwasaka, M. Yamato and A. -,- Ono, 1987: Consideration of depolarization ratio University of Tsukuba for his kind suggestion measurements by lidar - in relation to chemical with respect to the literature on the deserts in composition of aerosol particles. J. Meteor. Soc. China. The authors also wish to thank Mrs. A. Koshiba and K. Nakayama of the Applied Japan, 65, Merrill, J.T., R. Bleck and L. Avila, 1985: Modeling atmospheric transport to the Marshall Islands. J. Meteorological Division, JMA for providing them Geophys. Res., 90, with the SYNOP data, and the staff members of Murayama, N., 1980: The duststorm over the sea as the Tateno Aerological Observatory for providing revealed by the GMS satellite. Umi to sora, 55, them with radiosonde MT data. This study was (in Japanese). T. Kobayashi, -, T. Yano, K. Morita and H. partly supported by a scientific research fund Yasoshina, 1981: On the size distribution and dust (No ) from the Ministry of Education, loading of the Kosa (in Japanese). Paper presented in Science and Culture of Japan. the Autumn Meeting of the Meteorological Society References Arakawa, H. (ed.), 1969: Climates of Northern and Eastern Asia. Elsevier Publishing Company, 248p. of Japan. - and F. Kimura, 1984: Long range transport of the Kosa (in Japanese). Paper presented in Spring Meeting of the Meteorological Society of Japan. Arao, K. and Y. Ishizaka, 1986: Volume and mass of yellow sand dust in the air over Japan as estimated from atmospheric turbidity. J. Meteor. Soc. Japan, 64, Collis, R.T.H. and P.B. Russell, 1976: Lidar measurement of particles and gases by elastic backscattering and differential absorption. Laser monitoring of the atmosphere, Ed. ED. Hinkley, Springer-Verlag Berlin Heidelberg, Duce, R.A., C.K. Unni, C.K. Ray, J.M. Prospero and J.T. Merrill, 1980: Long-range atmospheric transport of soil dust from Asia to the tropical North Pacific: Temporal variability. Science, 209, Geographical Department of the Northwest Normal College, 1984: Atlas of physical geography of China (in Chinese). Map Publishing Company, 200p. Hirose, K., Y. Dokiya and Y. Sugimura, 1983: Effect of the continental dust over the north Pacific Ocean: Time variation of chemical components in maritime

16 472 Journal of the Meteorological Society of Japan Vol. 66, No. 3 Japan. J. Meteor. Soc. Japan, 65, Russell, P.B., W. Viezee, RD. Hake, and R.T.H. Collis, 1976: Lidar observations of the stratospheric aerosol: California, October 1972 to March Quart. J. Roy. Meteor. Soc., 102, Shaw, G.E., 1980: Transport of Asian desert aerosol to the Hawaiian Islands. J. Appl. Meteor.19, Takasugi, T. and H. Nakamura, 1988: A tracer diffusion model coupled with the JMA global spectral model (to be published). JMA/NWPD Technical Report. Uematsu, M., R. Duce, J. Prospero, L. Chen, J.T. Merrill and R.L. McDonald, 1983: Transport of mineral aerosol from Asia over the north Pacific Ocean. J. Geophys. Res., 88, Zhao, S., 1986: Physical geography of China. Science Press, and John Wiley and Sons, 209p.