Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Vertical dust structures and meteorological conditions over the Taklimakan Desert during the summertime periods (July–September) in 2006 and 2007 were investigated using satellite data including the CALIPSO lidar (CALIOP) measurements and the Weather Research and Forecasting (WRF) model simulation. Summertime convective velocity simulated by WRF was greater than 3 m/s, and the convective time scale was about 1000 s (ca. 15 min). We examined 42 CALIPSO paths (24 daytime and 18 nighttime paths) of data acquired under convective dusty conditions. The dust layer thickness derived from the CALIOP measurements reaches 3000–4000 m, which is approximately equal to the Tarim Basin depth. This thickness shows a good correlation with the WRF simulated depth of the convective boundary layer (BL). The dust remains suspended during both daytime and nighttime; the CALIOP average dust extinction coefficients in the BL are respectively 0.151 ± 0.102 km−1 and 0.128 ± 0.079 km−1 for daytime and nighttime. Finally, we estimated the dust amount transported from the BL to the free atmosphere. Typically, 10–20 Gg/day of dust (assuming area of 600 km × 300 km as a main part of the Taklimakan Desert) is transported from the Taklimakan Desert by vertical mixing. The daily horizontal dust flux above the BL was estimated to be 40–50 Gg/day over the Taklimakan Desert.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Mineral dust originating from Chinese deserts, especially from the Taklimakan Desert, has been found in the ice in Greenland [Bory et al., 2003] and the French Alps [Grousset et al., 2003]. It is considered that the dust from Taklimakan Desert has a long lifetime and is distributed globally. The dust can play an important role in the formation of cirrus clouds and therefore can affect the global radiation budget [Mikami et al., 2006]. It also supplies nutrients to the open ocean [Duce and Tindale, 1991].
 The Taklimakan Desert, one of the largest deserts in the world, is located in the Tarim Basin (average elevation 1000–1500 m ASL) in Northwest China. It is surrounded by high mountains of 4000–5000 m ASL, including the Pamir Plateau, Tibetan Plateau, Tian Shan Mountains, and Kunlun Mountains. The Taklimakan Desert has an east-west dimension of approximately 1400 km and a north-south dimension of 550 km. In this large area, however, there are only a few observation sites located in oasis cities at the desert's edges, where only the limited WMO SYNOP weather stations provide routine surface observations. Therefore, details of both the horizontal and vertical dust structures remain largely unknown.
 Aeolian Dust Experiment on Climate impact (ADEC), a research project conducted during April 2000–March 2004 [Mikami et al., 2006], set up a network for observations of dust emissions, transport, and deposition over a large geographic area from the Taklimakan to Japan. A polarization Lidar was operated at Aksu at the northern edge of the Tarim Basin during the ADEC campaign by Kai et al. . They observed springtime dust layers lofted up to 4–5 km, which is the first lidar measurement over Taklimakan Desert. Unfortunately, the Intensive Observation Period was performed only during the spring dust seasons, and the dust vertical distribution during summer was not available from the campaign. However, thick dust layers were observed frequently during the commercial flights (B757) from Urmuqi to Hotan, Tarim Basin. The basin was completely filled with white/grey floating dust especially during summer which was lofted up to the height of the Tian Shan mountains (M. Mikami, private communication, 2005).
 Continuous global measurements of aerosol and cloud vertical distributions with very high spatial resolution have become available since the successful launch of the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) onboard the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite in April 2006 [Winker et al., 2007]. The CALIPSO satellite overflies the Tarim Basin region at least once each day. Therefore, CALIOP provides an excellent chance to study the details of the dust structure over the Tarim Basin. Liu et al.  reported that the Taklimakan dust concentration increases during the early spring to early autumn based on the CALIOP first year observations. Similarly, Huang et al.  reported that the Taklimakan dust can be transported vertically up to the level of the Tibetan Plateau based on the CALIOP summertime measurements in 2006.
 Very recently, Uno et al.  showed the capability of an integrated analysis of the CALIOP measurement and dust transport models to explore the 3D structure of Asian dust transportation. These previous analyses focused on the dust transport from the Gobi Desert. In the present study, we have investigated the Taklimakan dust phenomena during summertime based on the first two year CALIOP measurements (2006, 2007). We extracted the typical dust profile under convective dusty conditions. Furthermore, we applied the Weather Research and Forecast (WRF) meteorological model [Skamarock et al., 2008] and clarified the typical dust structures over the Taklimakan Desert in summer as well as the meteorological conditions under which the dust layers were observed by CALIOP.
2. Observation Data and Meteorological Model
 The CALIOP Level 1B (Ver. 2.01) data are mainly used in this study. Those data provide profiles of total attenuated backscatter coefficient at 532 nm and 1064 nm and volume depolarization ratio (δ) at 532 nm. The dust extinction coefficients are derived using the Fernald's inversion method [Fernald, 1984] by setting the lidar ratio S1 = 35 sr, as described by Shimizu et al. . The inversion is started from an altitude of 14 km down to the ground surface. The Level 1B data are averaged to a horizontal resolution of 5 km before the inversion. The Cloud-Aerosol Discrimination (CAD) result from Level 2 data is used to screen out cloud signals. Detailed descriptions are available in the work by Uno et al. .
 The Weather Research and Forecast (WRF) meteorological model (ver. 2.2) is used to reproduce meteorological conditions and extract the information related to structures of the convective boundary layer. Because the Taklimakan region has a very complicated topography, we employed a nested two-grid system in our numerical simulations. The numerical model horizontal domain is centered at 37°N, 86°E for Grid 1 and 39.39°N, 83.12°E for Grid 2 (see C1 and C2 in Figure 2) on a Lambert projection system, comprising 100 × 90 (45 km resolution) and 211 × 126 (9 km resolution) grids. The vertical domain extends from the surface to 20 km and has 31 grid layers. The NCEP FNL Global Tropospheric Analyses result (1° × 1°, http://dss.ucar.edu/datasets/ds083.2/data) is used for model initialization. We analyzed two summer periods (July–September) of 2006 and 2007. We note that, however, the WRF model does not treat the dust concentration, and cannot consider the dust–radiation feedback. As a consequence, it can possibly cause over-cooling (overheating) under dusty calm night (daytime) conditions. We also used the OMI Aerosol Index (AI), MODIS Deep Blue AOT retrieval [Hsu et al., 2006], and WMO SYNOP visibility reports in our analyses for comparisons.
3. Results and Discussion
3.1. Meteorological Screening for Convective Conditions
 Dust emission amounts from the Taklimakan Desert are large from the early spring to early autumn [Liu et al., 2008]. Dust emission is controlled largely by strong winds produced by developing low-pressure and cold fronts, which are common meteorological phenomena in spring. Convective meteorological conditions on the other hand play a more important role in the dust emission in summer because the surface wind speed (WS) in this season is usually not as large as in spring.
Figure S1 shows the 3-month-averaged aerosol optical thickness (AOT) based on the MODIS Deep Blue AOT data averaged for 2006 and 2007 (see auxiliary material). Table 1 shows the 3-month-averaged AOT over the Taklimakan and Gobi Deserts. The location of the averaged area is specified in Table 1. We confirmed that the summertime AOT level over the Taklimakan is the second largest relative to other seasons and equal to 60% of the springtime AOT level, but higher than that over the Gobi Desert for the same time period. The WS is high in spring and its prevailing direction is ESE at 500 hPa; in summer, the WS lessens and its direction is more eastward.
Table 1. MODIS/AQUA Deep Blue Algorithm 550 nm AOT Averaged for 2006 and 2007
Number in parentheses is ratio to the Spring mean.
Averaged area is 75.6–88.1°E × 36.5–41.5°N.
Averaged area is 100–114°E × 36–44°N.
 We examined the WRF simulated surface sensible heat flux (SHF), and convective mixing layer height (Zi) in a box in the Tarim Basin as defined in Figure 2a. SHF is averaged over the defined box for a time period of 3–9 UTC; Zi is determined from the potential temperature profile and then averaged over a time period of 7–9 UTC. We also calculated the convective velocity scale w* = [HBLg/ΘSHF]1/3 (where HBL = Zi − zg; zg is the average elevation for the selected box region, g is gravitational acceleration and Θ is potential temperature).
Figure 1 portrays the daily variation of meteorological parameters for July, August, and September of year 2006 (upper two panels) and 2007 (lower two panels). Figure 1 also presents the Deep Blue AOT (1330LT), OMI AI (1345LT), and SYNOP visibility report base extinction (Vis_Ext). Vis_Ext is calculated by 4/Vis_min (km−1), where the minimum visibility between Hotan and Ruoqiang sites is taken as the value of Vis_min (see Figure 2 for the locations of these sites).
 The daily averaged SYNOP WS ranges 1.0–3.6 m/s and is less than 2.5 m/s (>85% of observation time). The SHF ranges 250–320 W/m2 until the end of August and then decreases gradually to 200 W/m2 by the end of September. It is surprising that Zi ranges 2600–5300 m ASL, which is almost the same as the elevation of the surrounding Tibetan Plateau. The convective scale w* has a large value of 2.0–3.2 m/s. The simulated summertime meteorological conditions can be characterized by a small surface wind speed with a very large w* (>3 m/s). The convective time scale HBL/w* becomes around 1000 s (ca. 15 min), indicating that the convective mixing is very strong and that the dust in BL must be vertically uniform.
 We also confirmed that the Deep Blue AOT and OMI AI are reasonably correlated to the visibility degradation. Vis_Ext also shows a good correlation with the Deep Blue AOT (e.g., R > 0.65 for 2007). Some large spikes in Vis_Ext correspond to WS > 2.5 m/s (e.g., days 208, 246 of 2006, and 237 of year 2007), indicating a strong-wind base dust emission, which is excluded from this study.
 Based on these basic meteorological and AOT daily information, we determine the convective dusty days among the CALIPSO orbits passing over the Tarim Box region during July–September of 2006 and 2007, using criteria of WS < 2.5 m/s, w* > 2.8 m/s, and in a relatively cloud free condition (judged based on the CALIOP CAD information). We selected 32 CALIPSO paths, including 19 daytime paths around 7 UTC, and 13 nighttime paths around 20 UTC, that meet these criteria. We also examined some cases in September where w* < 2.8 m/s (SHF decreases in September) when the CALIOP observations were judged to have undergone a typical convective condition (we extracted 5 daytime and 5 nighttime cases). In total, 42 CALIPSO paths were selected. Because of the high cloud occurrence frequency in August 2006, the selected days in August are fewer than those in the same time period of 2007. The CALIOP AOTs integrated over 0–1.5 HBL for these selected days are also presented in Figure 1. Variation of CALIOP AOTs is consistent with other dust observation results.
3.2. Snapshot of CALIOP Observations
Figure 2 portrays a snapshot of selected typical daytime and nighttime CALIOP measurements (dust extinction cross-sections), along with the WRF simulated wind fields and potential temperatures. For the typical daytime dust case observed on 10 August 2007 (the upper panels), Zi = 4370 m ASL, w* = 2.83 m/s and the average dust extinction coefficient within the BL is 0.135 km−1. The WS is low, and the dust profile is almost uniform. The dust is lofted in the air up to the elevation level of Tibetan Plateau. HBL becomes nearly 3300 m and the dust AOT in the BL reaches 0.45.
 The lower panels in Figure 2 show the nighttime dust case observed on 30 July 2006. The potential temperature for this nighttime case is also compared with that for the daytime case (broken line in the lower panel in c). The prevailing wind field is SE; The WRF simulation indicates a strong stable boundary layer (SBL). Above the SBL, the potential temperature gradient is rather mild and retains the residual layer. The top height of the nighttime dust layers is greater than the top height of convective boundary layers generated during daytime, and a very high dust concentration remains during nighttime (the average dust extinction coefficient within the BL is 0.21 km−1), which is carried over from daytime. These results show that thick dust layers exist over the Tarim Basin, even during nighttime.
3.3. Scaled Dust and Meteorological Parameters Over the Taklimakan Desert
 To characterize the Taklimakan dust, we compute non-dimensional profiles for the selected 42 CALIOP measurements: the dust extinction coefficient (Ext_dust) scaled by the averaged dust extinction coefficient in the BL (defined as Ext_dust_BL), and the differences of the potential temperature and WS from the BL averaged values. For these profiles, the height is scaled by the BL thickness HBL.
Figure 3 presents these scaled profiles for daytime (upper row) and nighttime (lower row). The spatial averaging method is the same as that used in Figure 2. The nighttime profiles are also compared with daytime profiles for the dust extinction coefficient and potential temperature. Table 2 presents the averaged dust extinctions and the WS.
Table 2. Characteristics of Summertime Taklimakan Dust Profiles
Number in parentheses is the standard deviation. Upper row shows the zg < z < HBL + zg; lower row shows the HBL + zg < z < 2 HBL + zg.
Number in parentheses is the standard deviation on HBL (m).
Daytime (6–8 UTC)
 From the scaled daytime extinction profile, the dust concentration is shown to have a peak at about z/HBL = 0.3 and decreases gradually to 0.7 at z/HBL = 1. The dust level linearly decreases above z/HBL = 1. One of our interests is to estimate the amount of dust above HBL. We examine the dust loading between 1 < z/HBL < 1.5, defined as
D1_1.5 is 0.28 in daytime, indicating that about 30% of dust exists above the BL. The averaged Ext_dust below HBL for the selected 24 daytime observations is as large as 0.151 ± 0.102 km−1.
 The nighttime scaled extinction profile shows that the lower part of the dust layer exhibits a stable stratification from the surface up to z/HBL = 0.3. An isothermal profile is apparent below HBL as a daytime residual layer. The scaled dust extinction coefficient shows a small peak within the surface stable layer at z/HBL = 0.1. Above the surface layer, the scaled dust extinction profile is distributed almost uniformly within the BL. The nighttime averaged Ext_dust_BL is 0.128 ± 0.079 km−1 (ca. 85% of the daytime value). Similar to the daytime dust distribution, the dust concentration decreases above the BL; the dust loading D defined by equation (1) is 0.27, almost the same as that for daytime.
 Next we estimate the vertical dust flux from the BL to the free atmosphere (FA) based on the similarity theory of the convective mixing layer. The vertical dust flux can be approximated as
where we assume that α ranges 0.2–0.4 based on the heat flux similarity within the convective boundary layer. Taking a typically observed gradient of dust extinction and potential temperature between 1 < z/HBL < 1.5, d[EXT_dust] = 0.04 km−1 and dΘ = 5 K, the mass/extinction scale factor of 1.7 mg/m3 km [Sugimoto et al., 2003], and the surface heat flux 300 W/m2, we get the vertical dust flux ∣Zi = α × 12.0 Kg/km2/h. If this value is invariant all day, and assuming a horizontal size of the main Taklimakan Desert area of 600 km × 300 km, then the total vertical dust flux can be estimated as
 Therefore, under the typical convective condition, on an order of 10–20 Gg/day of dust is transported from the Taklimakan Desert to FA by vertical mixing processes. This estimate includes neither vertical advection nor mountain/valley local wind transport contributions. For this reason, the estimated value is expected to be smaller than it should be.
 Another important quantity associated with the dust mass budget is the outflow by horizontal transport. We assume that the dust below HBL is trapped within the Tarim Basin and that the main outflow to FA occurs above HBL. Thus, the horizontal dust mass transport flux (HMF) can be estimated as
 Furthermore, if we assume a vertical cross-section of 300 km × 3000 m, then HMF becomes
 That is, the estimated daily horizontal dust flux is on an order of 40–50 Gg/day above the Taklimakan Desert.
 Atmospheric dust loading can also be evaluated as the product of (averaged dust extinction coefficient) × (mass/extinction scale factor) × (volume of Taklimakan airborne dust; assuming 600 km × 300 km × 3000 m height). Applying these numbers, we get
 It is interesting to note that the dust loading above HBL is almost the same as HMF, which means that the pumping up process of the Taklimakan dust above HBL is the bottle neck process. Once it is transported up to the FA from the basin, the dust is transported rapidly away from the Taklimakan region. Our estimate of 40–50 Gg/day for the horizontal dust transport in the FA is the first reported estimate for the Taklimakan region and is important in quantifying the global dust budget and radiation balance as a background level of dust.
4. Concluding Remarks
 We clarified the summertime Taklimakan dust structure based on the NASA/CALIPSO satellite borne lidar data and the Weather Research and Forecasting (WRF) model for summertime seasons (July–September) in 2006 and 2007. The WRF results showed that the daytime convective mixing is very strong within the Tarim Basin; the convective scale w* is larger than 3 m/s, and the convective time scale HBL/w* reaches around 1000 s (ca. 15 min). We selected 42 CALIOP paths including daytime and nighttime measurements under convective dusty conditions. The dust layer thickness observed by CALIOP showed a good agreement with the simulated convective boundary layer thickness (HBL) of 3000–4000 m, which is approximately equal to the full depth of the Tarim Basin. The CALIOP measurements showed that dense dust layers remained in the air not only during daytime but also during nighttime. The averaged dust extinction coefficients below HBL derived from the CALIOP are respectively 0.151 ± 0.102 km−1 and 0.128 ± 0.079 km−1 for daytime and nighttime. Finally, we estimated the vertical dust flux transported from the BL to the FA. The results showed that a dust amount of 10–20 Gg/day is transported by vertical mixing processes from the Taklimakan Desert under the summertime convective conditions. The daily horizontal flux above the BL was estimated to be on an order of 40–50 Gg/day over the Taklimakan Desert. These results suggest that the Tarim Basin supplies a large amount of dust to the FA ordinarily in summertime and the dust acts as a background flowing northwesterly from the Taklimakan Desert area.
 This work was partly supported by the Global Environment Research Fund, Ministry of Environment, Japan (C-061), and a Grants-in-Aid for Scientific Research under grant 20244078 from the Ministry of Education, Culture, Sports, Science and Technology, Japan. The CALIPSO data were obtained from the NASA Langley Research Center Atmospheric Sciences Data Center.