Electron spin resonance (ESR) signal intensity and crystallinity index (CI) of fine- (<16 µm) and coarse-grained (>16 µm) quartz were measured in surface samples from the Taklimakan desert in western China, the Badain Juran, Tengger and Mu Us deserts in northern China, and the Gobi desert in southern Mongolia to evaluate whether these geophysical parameters can serve as reliable provenance tracers of Asian dust. The results indicate that spatial variability of both ESR signal intensity and CI is evident within the Taklimakan deserts and the Mongolian Gobi, but less significant in the three deserts of northern China. Coarse-grained quartz from the Mongolian Gobi and northern China deserts can be differentiated from the Taklimakan desert using the ESR signal intensity. Fine-grained quartz originating from three major Asian dust sources, i.e., the Gobi-sandy deserts in western China, northern China and southern Mongolia, can be distinguished effectively using the combination of ESR and CI signals. Our results suggest that ESR signal intensity and CI can discriminate the sources of fine-grained quartz better than coarse-grained quartz, providing an effective approach to trace the provenance of fine-grained dust deposition on the land and in the ocean.
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.
 Dust is an important component of the Earth climate system and affects the global climate by changing the solar radiation balance and oceanic biogeochemical cycles [e.g., Duce, 1995; Tegen et al., 1996; Jickells et al., 2005]. Eolian dust preserved in geological archives (e.g., continental loess, pelagic sediment, peat and ice core) can reflect past changes in source aridification and atmospheric circulation processes [e.g., Biscaye et al., 1997; Rea et al., 1998; Guo et al., 2002; Sun and An, 2005; Ferrat et al., 2012a]. In particular, the tempo-spatial variability of Asian dust provenance has been investigated intensively to constrain the drying history of Asian inland and to reconstruct fluctuations of past westerly and winter monsoon circulations [Pettke et al., 2000, 2002; Bory et al., 2002; Sun et al., 2008; Nagashima et al., 2007, 2011; Li et al., 2011; Ferrat et al., 2011, 2012b].
 Although the Gobi and the sandy deserts in inland Asia are the dominant sources of Asian dust [e.g., Liu, 1985], monitoring and modeling of modern dust emissions show that Asian dust mainly originates from three major regions, namely the Taklimakan desert in the northwestern China, the sandy deserts in northern China (e.g., Badain Juran and Tengger deserts) and the Gobi desert in southern Mongolia [Sun et al., 2001; Zhang et al., 2003a; Wang et al., 2007]. Alternatively, the Qaidam Basin [Bowler, 1987; Pullen et al., 2011], the Tibetan Plateau and its northern areas [Fang et al., 2004; Chen et al., 2007] or the huge alluvial fans distributed in the Gansu Corridor [Derbyshire et al., 1998] were also suggested as potential sources of Asian dust and Chinese loess. The deposition areas of Asian dust include the Chinese deserts, the Chinese loess plateau (CLP), the Tibetan plateau, the Japan Sea, and the North Pacific. It is estimated that after emission of Asian dust from the source regions, approximately 30% is deposited again in the original areas, 50% is transported to the Northern Pacific, and 20% is deposited in the CLP [Zhang et al., 1997].
 Numerous studies on Asian dust provenance focused on the elemental and radiogenic geochemistry of dust from source areas [Zhang et al., 1993; Takanori et al., 2005; Chen et al., 2007; Yang et al., 2009, 2008; Li et al., 2009; Ferrat et al., 2011]. The most controversial point is about the importance of the Taklimakan desert and Qaidam basin in regional dust transport [Liu et al., 1994; Zhang et al., 1997; Sun, 2002a]. Isotope and element tracers suggest that dust from the Taklimakan desert is an important source of dust input to the CLP [Liu et al., 1994; Zhang et al., 1997]. Uranisum-Pb ages of zircon crystals from Loess samples and potential source areas suggest that the loess of the CLP was partly derived from the Qaidam Basin and the northern Tibetan Plateau to the west [Pullen et al., 2011]. These inferences, however, are in contrast with evidences from electron spin resonance (ESR) signal and crystallinity index (CI) data of quartz [Sun et al., 2007, 2008], Nd-Sr isotope signatures [Chen et al., 2007; Li et al., 2009; Chen and Li, 2011], isotope and trace metal compositions of carbonate [Li et al., 2007, 2013], which suggest that the Gobi and sandy deserts in southern Mongolia and northern China are two predominant dust sources to the CLP.
 Late Cenozoic changes in dust provenance have been studied using element and isotopic tracers of Chinese loess, suggesting that dust sources are likely stable at tectonic and orbital timescales [Gallet et al., 1996; Jahn et al., 2001; Wang et al., 2007]. In contrast, element ratios [Zhang et al., 1997], Sr-Nd-Pb isotopic compositions [Sun, 2002a, 2005; Sun and Zhu, 2010], and ESR-CI results [Sun et al., 2008] suggested that the loess provenance likely shifted on tectonic to millennial timescales. Very recently, detrital-zircon age data of representative loess and paleosol samples from three distant sites from the CLP suggest that the dust provenance is heterogeneous and spatially variable [Xiao et al., 2012]. Clearly, a consensus regarding the tempo-spatial dust provenance variability has still to be reached.
 To address this important problem, effective provenance tracers that vary significantly between different dust sources are needed. Since quartz is the major and most stable mineral in dust and loess [Xiao et al., 1995; Sun et al., 2000], ESR signal intensity of the E1' centre (an indication of the age of the host rock, Ono et al. , Toyoda and Hattori ) and CI (an index reflecting the formation condition of the mineral such as temperature and speed of crystallization, Murata and Norman ) of quartz have been tested for the identification of the dust source identification in the CLP and the Japan Sea [Ono et al., 1998; Toyoda and Naruse, 2002; Nagashima et al., 2007, 2011; Sun et al., 2007, 2008]. While the preliminary results were encouraging, the sample set from the dust sources reported in earlier studies was very limited (only 3 to 5 samples were analyzed for one desert, Sun et al. ) and the coarse-grained quartz fractions were not assessed.
 In this study, we collected systematically surface sediments from the Taklimakan desert in the northwestern China, Gobi deserts in southern Mongolia and northern China, and three sandy deserts (Badain Juran, Tengger and Mu Us) in northern China (Figure 1). ESR signal intensity and CI of fine- and coarse-grained quartz of 80 surface samples from three Asian dust sources were measured to assess spatial variability of the ESR-CI results within each desert and between the deserts. The main objective was to develop the potential of the quartz ESR-CI signatures as source tracers to distinguish dust provenance in marine and terrestrial environments.
2. Setting and Sampling
 The Taklimakan desert with an area of 3.5 × 105km2 is the largest desert in China. The desert is located in the Tarim Basin of the Xinjiang Province in northwest China and surrounded by the Kunlun and Altun Mountains to the south, the Pamir Plateau to the west, and the Tianshan Mountain to the north. The Taklimakan desert contains almost all types of sand dunes, ranging from simple dunes with various shapes to numerous forms of compound or complex sand dune [Zhu et al., 1981]. This desert is regarded as one of the major Asian dust sources [Liu et al., 1994; Sun et al., 2001; Zhang et al., 1997, 2003]. The Gobi desert in southern Mongolia is located on the Mongolian Plateau, reaching the Khangai Mountain to the north and the Gobi Altai Mountain to the west. It is largely covered with stone and gravel with an even topography. Proluvial, alluvial and glacial debris from high mountains is intensively weathered and ground to provide enough detrital materials for the dust emission and transportation [Sun et al., 2001; Sun, 2002b; Natsagdorj et al., 2003].
 The Badain Juran desert is located on the western Alxa Plateau, surrounded by the Yabulai Mountain to the east, the Qilian Mountain to the south, and branches of the Gobi Altai Mountain to the north. The desert covers 47,000 km2 with the altitude of 1200–1700 m. Mobile dunes occupy 83% of the desert. The desert is strongly influenced by northwesterly winds, and fine-grained sediment is supplied by quaternary sediments to the south and huge sand hills and fluvial fans to the north [Yang et al., 2007]. The Tengger desert is situated on the southeastern Alxa Plateau (Figure 1), adjacent to the Yellow River to the southeast, and the Helan Mountain to the east. The desert covers an area of 42,700 km2, with the altitude of 1200–1400 m, mainly composed of mobile dune and many dry or drying lakes [Zhang et al., 2004]. The Mu Us desert is located in the depression of the Ordos Plateau with the altitude of 1100–1300 m and an area of 39,800 km2, which is underlain by red and gray sandstone from the Cretaceous in the middle and northwest, and by the Quaternary loess in the south and east [Zhu et al., 1980].
 From 2007 to 2009, we conducted several field excursions in these potential dust sources including the Taklimakan (TK), Tengger (TG), Badain Juran (BJ), Mu Us (MU), and Mongolian Gobi (MG) deserts (Figure 1). Surface samples were taken systematically from these sources by scratching off 1–2 cm thick clay mud crust at intervals of 50–100 km distance, mainly from dried riverbeds or hydrocephalus depressions within Gobi/sandy deserts (Figure 2). For this study, we chose 24 samples from the Taklimakan desert, 21 samples from the Mongolian Gobi, 14 samples from the Badain Juran desert, 12 samples from the Tengger desert, 9 samples from the Mu Us desert for ESR and CI measurements. Exact locations and descriptions of 80 surface samples are shown in Figure 1 and presented in Table 1.
Table 1. Location, the ESR Intensity, and CI of Fine- and Coarse-Grained Quartz in Surface Desert Samplesa
CI <16 (µm)
CI >16 (µm)
ESR <16 (µm)
ESR >16 (µm)
Also shown are the average values and relative standard deviation (%SD) of the ESR and CI for each source.
Average ± %SD
8.4 ± 1.9%
8.0 ± 2%
5.7 ± 21%
3.2 ± 29%
Average ± %SD
8.3 ± 2%
7.9 ± 2.4%
18.7 ± 10%
8.6 ± 28%
Average ± %SD
9.3 ± 0.9%
8.2 ± 0.7%
16.3 ± 10%
5.9 ± 21%
Badain Juran Desert
Average ± %SD
9.6 ± 0.6%
8.1 ± 1.3%
21.4 ± 9%
7.1 ± 21%
Mu Us Desert
Average ± %SD
9.6 ± 0.8%
8.2 ± 0.9%
14.0 ± 8%
6.5 ± 23%
3. Experimental Methods
 Wind tunnel experiments showed that dust emission and transport depend on grain size fractions [Pye et al., 1987]. Previous studies focusing on Asian dust suggest that most coarse-grained particles are transported by saltation or rolling mode, whilst the fine-grained fractions are transported by suspension [Zhang et al., 1999]. A range of modern observational data suggests strong winds particularly in the spring season as a primary driver of dust emission and transport in Asia [e.g., Kurosaki and Mikami, 2003; Roe, 2009; McGee et al., 2010]. Studies on the transport dynamics of loess sediment suggest that coarse-grained sediments are transported from the proximal sources by near-surface northwesterly wind, whereas fine-grained sediments are transported from distant sources by near-surface northwesterly wind and high-altitude westerly wind [Sun et al., 2004; Prins et al., 2007]. Thus, the grain size sorting effect on various provenance tracers has to be considered before evaluating the potential of source tracing approaches [Sun, 2005; Sun et al., 2008].
 Grain size composition of atmospheric dust indicates that the fraction below 16 μm constitutes about 70% of the total atmospheric loading [Zhang et al., 2003b], which is consequently transported over long range [Rea, 1994]. Sun et al.  demonstrated a grain-size effect on the ESR and CI signatures, with relatively high-ESR signal intensity and CI higher in the fine-grained fraction. For consistency with previous work, during this study we separated the bulk samples into two size fractions (>16 and <16 μm) and pretreated the samples to remove organic matter, carbonate, and Fe/Mn oxides before analytical measurements. Pretreatment procedures are described in Sun et al. .
 CI was measured using X'pert Pro MPD X-ray diffractometer (XRD) at the Institute of Earth Environment, Chinese Academy of Sciences. The procedure was as follows: (1) sample powder was mixed quantitatively with standard Si power (Wako, Co. Ltd, Japan) and ground in an agate ball mill; (2) the mixed power was placed in the glass holder and irradiated with a Cu Kα source with voltage at 40 kV and 40 mA. The XRD spectra was taken over a scan range from 66° to 69° (2θ), with a scan speed of 0.02° s−1 and a sampling angle of 0.006°. Measurements were done in duplicate on each side of the glass holder, and the relative standard deviation of the CI measurements was estimated to be 1.5% based on thirty reduplicate measurements of a laboratory standard sample (Wako Quartz, CI = 10). The CI was calculated using the equation:
where the factor F1 is estimated as 1.21 from the calibration curve of different mixing ratio samples between the wako quartz (CI = 10) and Si powder.
 The quartz content (QC) was determined using the internal standard addition and measured by the XRD with a scan range of 20°–30° (2θ), a scanning speed of 0.02° s−1 and a sampling angle of 0.004°. The QC was calculated using the following equation:
where WSi and WSample are the weight of silicon standard and the sample, respectively, IQtz/ISi is the peak ratio of the quartz diffraction at 20.9 Å to the silicon diffraction at 28.5Å, and F2 is a correction factor that varies with different crystallinity of the samples. To estimate F2, we chose three samples with different CI values: (1) the industrial standard quartz samples (Wako, Co. Ltd., CI = 10); (2) a natural carnelian from eastern Qilian Mts (CI = 8.7); (3) a surface sample from Tengger desert (CI = 7.3). The relationship between WSample/WSi ratio and IQtz/ISi ratio was used to determine the F2 to be 0.236 for CI = 10.0, 0.216 for CI = 8.7 and 0.199 for CI = 7.3. The F2 of a sample is then estimated from the CI value using the equation: F2 = 0.0137CI + 0.0976. Reproducibility of the QC was better than 3.0% based on 50 reduplicate measurements of a laboratory standard.
 ESR measurements were conducted using X-wave band EMX ESR spectrometer at the China Institute of Atomic Energy (Beijing). About 0.10–0.12 g of pretreated sample was radiated using a 60Co source with a total dose of 2.5 kGy. To convert the oxygen vacancies into E1' centre, the samples were heated at 300°C for 10 min in a Thermoluminescence heated stove (FJ-411). The measurements were conducted under the following conditions: room temperature, X-wave band, microwave power of 0.01 mW, scan range of 50 mT, centre field of 3502.35 mT, scan time of 20.48 mT and time constant of 655.36 ms. Measurements were repeated three times by varying the angle between 0°, 120°, and 240°, respectively. ESR signal intensity was calculated using the following equation:
where IE',WSample, QCSample and F3 are the peak value of E1' centre in the ESR spectrum (spins), the weight of the sample (g), the quartz content of the sample, and a constant to adjust the peak height of the E1′ center to the standard ESR, respectively.
 To estimate the constant F3, eight pure quartz samples with different ESR signal intensities and 10 desert surface samples with different quartz contents were measured in the China Institute of Atomic Energy. We find that the calibration constant F3 is variable along with different quartz contents, and can be estimated using the equation of F3 = −0.63 × QC+2.535. Based on 26 measurements of three laboratory standard samples (TG, MN and Japan-1), the error of ESR signal intensity is estimated to be ≤3%. The ESR signal intensity was normalized by quartz content and expressed as spin units, where one spin unit is equivalent to 1.3 × 1015 spins g−1 [Toyoda and Naruse, 2002]. In combination of the errors for QC and CI estimation, the propagated error for the ESR signal intensity is less than 10%.
 The ESR signal intensity of coarse-grained quartz (1.6∼11.8 spin units) is lower than that of fine-grained quartz (3.7∼24.1) (Figure 3a). ESR signal intensity of fine-grained quartz varies between 3.7 and 7.9 in the Taklimakan desert, 15.9 and 22 in the Mongolian Gobi, and 12.4 and 15.2 in the Mu Us desert, 14.2 and 18.6 in the Tengger desert, and 18.8 and 24.1 in the Badain Juran desert. The ESR signal intensity of coarse-grained quartz exhibits similar spatial variability, changing from 1.6 to 5.0 in the Taklimakan desert, 5.3 to 11.8 in the Mongolian Gobi, 4.4 to 8.1 in the Mu Us desert, 3.9 to 8.3 in the Tengger desert, and 4.9 to 9.6 in the Badain Juran desert. The average ESR signal intensity for fine- and coarse-grained quartz is significantly lower in the Taklimakan desert (5.7 and 3.2) than in other four source areas (14–21.4 for fine-grained quartz and 6.5–8.6 for coarse-grained quartz). Difference in the average ESR signal intensity between fine- and coarse-grained quartz is also lower in the Taklimakan desert (2.5) than in other four sources (7.5–14.3). Different pretreatment processes on pure quartz samples (i.e., grinding, irradiation, and heating) can result in strengthening of the ESR signals to varying degrees, particularly the heating/annealing can result in at least twice strengthening of the ESR signal intensity and has greater influence on fine fractions [Chen et al., 2011]. Since fine-grain quartz is likely originated from easily weathered elder rocks and recycled significantly compared to the coarse-grained quartz, the relatively high-ESR signal intensity is likely attribute to the recycling processes with more irradiation and heating effect on the fine-grained quartz.
 Spatial CI variability is evident for fine-grained quartz particles (8–9.7), but not obvious for coarse-grained quartz samples (7.6–8.3) (Figure 3b). The CI values of fine-grained quartz are lower in the Taklimakan (8.2–8.7) and the Mongolian Gobi (8–8.6) deserts than in the three deserts of northern China (9.4–9.7). The CI values of coarse-grained quartz, however, are mostly lower that those of fine-grained quartz, and similar between the different deserts with values ranging between 7.6 and 8.3. Difference in the average CI between fine- and coarse-grained quartz is significantly lower (∼0.4) in the Taklimakan desert and the Mongolian Gobi than in other three sources (1.1–1.4). In addition, spatial CI variability of both fine- and coarse-grained quartz is more significant within the Taklimakan desert and Mongolian Gobi relative to other three sources. The CI difference between fine- and coarse-grained quartz particles indicates that different size fractions are probably originated from different host rocks. For example, the fine-grained quartz may have higher contribution of strongly weathered rocks, in which the high-CI quartz is preferentially preserved during the weathering and grinding processes. It is likely that the fine-coarse differences in the ESR and CI data are attributed to different grinding, weathering and recycling processes during the formation of the fine and coarse quartz particles.
 Spatial variability of ESR signal intensity and CI of fine-grained quartz is significant within the Taklimakan desert and Mongolian Gobi, but insignificant within three deserts of northern China. For coarse-grained quartz, ESR signal intensity exhibits spatial variability within each desert, whereas CI shows spatial discrepancy only within the Taklimakan desert and the Mongolian Gobi. Combination of these two parameters suggests that the Taklimakan desert can be distinguished from other dust sources by its lower ESR signal and CI values. The Mongolian Gobi can be differentiated from the other sources by its higher ESR signal and lower CI values. Three deserts in northern China, however, are characterized by higher ESR signal and CI values. Notably, the Taklimakan desert can be further divided into two parts: higher ESR signal intensity in the northern Taklimakan (TK1) and lower ESR signal intensity in the southern Taklimakan (TK2). Similarly, the Mongolian Gobi can be separated into two subregions: lower CI values in the eastern Mongolian Gobi (MG1) and higher CI values in the western Mongolian Gobi (MG2) (Figure 3).
5.1. Homogeneity or Heterogeneity of Quartz Particles within Each Desert
 To test the homogeneity of fine- and coarse-grained quartz particles within each source, we compared the relative standard deviations (RSD) of ESR signal intensity and CI results with the relative analytical errors of these two parameters (1.5% for CI and 10% for ESR signal intensity). If the RSDs of both ESR signal intensity and CI results from one desert are larger than the relative analytical errors, the quartz particles within this desert can be tentatively considered as heterogeneous. RSDs of the ESR signal intensity and CI for both fine- and coarse-grained quartz are higher in the Taklimakan and the Mongolian Gobi deserts than in three deserts of northern China (Table 1). RSDs of the ESR signal intensity indicates that coarse-grained quartz particles are likely heterogeneous in these sources (RSDs > 20%), whereas the fine-grained quartz particles are well mixed in these sources except for the Taklimakan desert.
 Within the Taklimakan Desert, the ESR signal intensity and CI are estimated to 5.7 ± 21% and 8.4 ± 1.9% for fine-grained quartz, and 3.9 ± 29% and 8.0 ± 2% for coarse-grained quartz, respectively (Table 1). ESR signal intensity for both fine- and coarse-grained quartz reveal that quartz particles in the Taklimakan desert are likely heterogeneous, with higher ESR values (5.3–7.9) in the northern Taklimakan desert (TK1) and lower ESR values (3.7–6.5) in the southern Taklimakan desert (TK2) (Figure 4). Since ESR signal intensity is an indicator of the age of the host rock of the quartz [Ono et al., 1998; Toyoda and Naruse, 2002], the variability of the ESR signal intensity may result from the discrepancy of the different regional geological background. The Tianshan Mountains consist mainly of Paleozoic metamorphic and volcanoclastic rocks, with the thinner Mesozoic-Cenozoic sediments in the southern margin [Ma, 2002]. In contrast, the western Kunlun Mountains mainly are composed of the Cenozoic rocks, with vast development of alluvial fans covered with the Quaternary alluvial materials [Ma, 2002]. Large and thick Mesozoic-Cenozoic sediments take out in the Altun Mts to the southeast of the Taklimakan desert [Ma, 2002]. Therefore, the north-south gradient of the ESR signals indicates that the ages of host rock for quartz particles are relatively older in the southern Tianshan Mts compared to the western Kunlun and Altun Mts [Sun et al., 2007], as rocks surrounding the southern Taklimakan deserts likely suffered higher temperature during the formation to reset the ESR signal during the Cenozoic [Tada et al., 2010].
 The ESR and CI data exhibit significant variations for the coarse-grained quartz in the southern Mongolian Gobi, ranging between 5.3 and 11.8 and between 7.6 and 8.3, respectively. The ESR signal intensity in the eastern part (MG1, 5.3–8.4) is lower than the western part (MG2, 8.0–11.8), while the CI values in MG1 (7.6–7.9) are slightly lower than in the MG2 (7.8–8.3) (Figure 4b). This spatial variability of the physical parameters of the coarse-grained quartz is likely caused by the regional geological background. For example, the eastern part (MG1) is largely covered by the Gobi desert, where the coarse-grained quartz possibly originated from the underlain Mongolian Plateau that is primarily made up of Protero-Paleozoic tuva rock (amalgamated metamorphic and igneous complexes) and Permian to Jurassic intrusive rock [Ma, 2002]. The western part of the Mongolian Gobi (MG2) is surrounded by two high-altitude mountain chains (Gobi Altai Mountains in west and Khangai Mountains in northwest). Paleozoic metamorphic and igneous rocks are the main rock types of the Gobi Altai Mountains and Khangai [Ma, 2002]. These two high-altitude mountains may provide enough detrital materials for the dust transportation through glacial grinding processes [Sun et al., 2001; Sun, 2002b; Natsagdorj et al., 2003]. For the fine-grained quartz, the ESR and CI data varied from 15.9 to 22.0 and from 8.0 to 8.6, respectively. The small variability of the ESR and CI suggests that the fine-grained quartz might have been better mixed in the Mongolian Gobi compared to coarse-grained quartz.
 For the three deserts in northern China, the coarse- and fine-grained quartz particles are almost homogenous within each desert as inferred from the small RSDs of the ESR and CI results (Figure 4c and Table 1). Spatially, the ESR and CI variability is indistinguishable for coarse-grained quartz but evident for fine-grained quartz between these three deserts. For fine-grained quartz, the Tengger desert is separated from the Mu Us and Badain Jurain deserts by lower CI values, while Mu Us and Badain Jurain deserts can be further differentiated by the ESR signal intensity ranging from 12.4 to 24.1. The ESR-CI variability for fine-grained quartz in the three deserts is evident due to the different regional geological settings of their sources. The Mu Us Desert is overlaid on the Erdos Plateau, which consists of Cretaceous sandstone underlain by Mesoarchean crystalline basements [Liu and Yang, 2000]. The Badain Juran desert is located in the northern branches of the Qilian Mountains, while the Tengger desert is surrounded by the Helan Mountains to the east and the northeastern branches of the Qilian Mountains to the southwest. Qilian Mountains and Helan Mountains are mainly composted of Paleozoic metamorphic rocks [Ma, 2002]. The metamorphic rock is subjected to heat and pressure [Blatt and Tracy, 1996] causing recrystallization during metamorphism. The degree of crystallization of metamorphic rocks is higher than that of the parent rocks (e.g., sedimentary, igneous, and/or other metamorphic rock) [Lu and Sang, 2002], resulting in the higher ESR signals and CI of quartz grains. The different parent rocks surrounding these three deserts account for the spatial variability of the ESR and CI results.
5.2. Spatial Variability of the ESR-CI Results for Fine- and Coarse-Grained Quartz
 Our early work suggests that spatial variability of ESR and CI in the fine-grained quartz is significant among the different deserts [Sun et al., 2007], but if the fine-grained quartz particles are homogeneous within each desert remains unresolved due to the limited samples from these deserts. In this study, a more systematic characterization of the ESR signal intensity and CI of fine-grained quartz indicate a homogeneous nature of fine-grained quartz within three deserts in northern China (Figure 5a). These two studied parameters differ significantly between the Taklimakan desert and Mongolian Gobi although the ESR and CI values are variable with each desert. The spatial feature of the ESR and CI variability founded during this study is in line with previous work [Sun et al., 2007]. The Taklimakan desert is characterized by lower ESR and CI values, whereas three deserts in northern China are characterized by higher ESR and CI values; both contrast the Mongolian Gobi with higher ESR but lower CI values.
 We note that the ESR and CI values in northern China deserts and southern Mongolian Gobi in this study are higher then in early study. This is probably attributed to different spectrometers employed for the ESR measurements. Overall, the variability of the ESR signal intensity and CI is remarkable between these three major dust sources, confirming previous findings that fine-grained quartz from the three major Asian dust sources can be effectively discriminated using a combination of the ESR signal intensity and CI. Considering the geologic implications of these two parameters, the host rocks of the quartz in the Mongolian Gobi and in the three deserts of northern China are mostly older than those in the Taklimakan desert, while the host rock are mostly igneous in three deserts of northern China compared to the Taklimakan desert and the Mongolia Gobi (likely dominated by metamorphic rocks).
 With respect to the coarse-grained quartz, however, the spatial variation of the ESR and CI values between the three major dust sources is less clear (Figure 5b). The average values for the ESR signal intensity exhibit a small spatial variability, with ESR signal intensities increasing from the Taklimakan desert (3.2) to the three deserts of northern China (6.5) and the Mongolian Gobi (8.6). In contrast, the average values of the CI do not show spatial variations. The heterogeneous nature of coarse-grained quartz within each source area and insignificant spatial variability prevents us combining these two parameters for tracing the provenance of coarse-grained quartz. Nevertheless, the Taklimakan desert seems to be different from the other two major Asian dust sources with the relative low-ESR signal intensity.
 We examined systematically the ESR signal intensity and the CI of fine- and coarse-grained quartz from the Gobi and sandy deserts in western and northern China and southern Mongolia. The ESR and CI results suggest that quartz particles are not well mixed within the Taklimakan desert and Mongolian Gobi, but pretty homogeneous within three deserts in northern China. We find statistically significant differences of ESR and CI between the various deserts in the fine-grained but not in the coarse-grained quartz. For the fine-grained quartz, the Taklimakan desert in western China exhibits lower ESR signal intensity and CI, differing significantly from the Northern China sandy deserts (high ESR and CI) and southern Mongolian Gobi desert (high ESR and low CI). Such a spatial difference in the ESR signal intensity and CI is closely associated with the geological setting (i.e., age and type of host rocks surrounding each source). Our results confirm that a combination of the ESR signal intensity and CI for the fine-grained quartz can be employed efficiently to trace the provenance of Asian dust transported to nearby CLP and to distant marine and terrestrial ecosystems. The next steps are to test the ESR-CI fingerprints in modern dust storms to confirm whether these surface sediments are representative of dust transported long distances.
 We thank the constructive comments by two reviewers. This work was supported by grant from the Natural Science Foundation of China (No. 41072272) and Key Innovation Project (KZCX-EW-114) from the Chinese Academy of Sciences.