Fine-grained Miocene sediments from Tianshui Basin, northeastern Tibetan Plateau, have received intense attention recently because these sediments were identified as loess. The presence of early Miocene loess pushes the timing of initiation of inland Asian desertification from 8 Ma back to 22 Ma. However, mudflat/distal fan and shallow lake sediments of Miocene have also been reported in Tianshui Basin. Consequently, the origin of these fine-grained Miocene sediments in this area remains controversial. Here we investigate then-alkane biomarker characteristics of Neogene sediments from a north-south transect of exposures within Tianshui Basin and compare these molecular distributions with those published Quaternary loess to help resolve the disputed origin. We found thatn-C23 and n-C25 alkanes, sourced from either aquatic macrophytes or palustrine plants, are ubiquitous in the Miocene sediments from Tianshui Basin but are largely absent in Quaternary loess. This striking difference between n-alkane distributions in the Tianshui samples and the Quaternary loess casts doubt on an eolian origin for the Tianshui samples and challenges the hypothesis of an early Miocene onset of Asian interior desertification.
 Initiation of desertification in the Asian interior is one of the most important events observed in the late Cenozoic paleoclimate record. One prevailing view is that initiation of Asian interior desertification did not occur until 8 Ma [An et al., 2001]. However, a recent study from the Tianshui Basin (QA-I section), northeast of the Tibetan Plateau (Figure 1), suggests that desertification may have begun by 22 Ma [Guo et al., 2002] mainly based on a) the similarity in whole rock geochemistry between the Tianshui samples and Quaternary loess, b) hundreds of paleosol layers observed in the QA-I section and c) the similarity of grain-size, quartz micromorphology, and rock magnetic parameters et al. [Guo et al., 2002, 2010; Liu et al., 2006; Oldfield and Bloemendal, 2011]. However, the similarity of whole rock geochemistry does not necessarily require an eolian transport mechanism because any well-mixed upper crustal materials would have a similar whole rock composition [Taylor and McLennan, 1985]. In addition, the presence of paleosol layers is not conclusive evidence for an eolian origin, because paleosols are commonly seen in alluvial fan, floodplain, or mudflat settings, where sediments are frequently exposed to air allowing pedogenesis [Kraus, 1999]. Rock magnetic properties and terrestrial snails may also be similar in paleosols whether they are developed in loess or in other fine-grained sediments, making it difficult to distinguish their origin on these criteria. In contrast, many subsequent sedimentological studies proposed that the set of fine-grained Miocene sediments were deposited under the conditions of mudflat/distal fan, floodplain and shallow lake environment in the Tianshui Basin [Alonso-Zarza et al., 2009, 2010; Flynn et al., 2011; Li et al., 2006].
 Because of the importance of initiation of inland Asian desertification in paleoclimate research, more diagnostic source-indicating proxies need to be applied to the fine-grained Miocene sediments from Tianshui Basin before one can confidently state that this set of sediments are loess and inland Asian desertification had occurred by 22 Ma. Lipid biomarkers are resistant to degradation and can be diagnostic of their biological source, yielding important environmental interpretations [Brocks and Summons, 2004; Castañeda and Schouten, 2011; Eglinton and Eglinton, 2008; Gao et al., 2011; Pearson et al., 2007]. However, no biomarker work has yet been applied to these sediments. Here we examine the n-alkane distribution patterns of fine-grained Neogene sediments in three sections (Yanwan, QA-I, and Yaodian sections) (Figure 1) from the Tianshui Basin (see auxiliary material for geological setting, materials and methods) and then compare them with that of published Quaternary loess from the Chinese Loess Plateau [Bai et al., 2009; Liu et al., 2008; Xie et al., 2003; Zeng et al., 2011; Zhang et al., 2008, 2006] in order to provide further insight into the disputed issue.
Figure 2 shows the n-alkane distributions of the Yanwan (n = 26), QA-I (n = 7), Yaodian (n = 6) sections. All sediments containn-alkanes ranging fromn-C15 to n-C37. Most of them display a bimodal pattern, whereas several samples show a tri-modal distribution.
 For the Yanwan section, aliphatic hydrocarbons of 22 samples exhibit a bimodal distribution: one peak at n-C17/n-C18with no odd-over-even carbon preference of the short-chainn-alkanes, and the other peak atn-C23 or n-C29/n-C31with no odd carbon preference of medium-chainn-alkanes (n-C21–25), but with apparent odd carbon preference from n-C26 to n-C35 (average CPI27–33 is 2.03) (Table S1), while 4 samples (YW-40.5 = d, YW-52.3 = e, YW-76.5 = g and YW-101.5 = i) show a tri-modal distribution with peaks atn-C17/n-C18, n-C23, and n-C29/n-C31, respectively.
 All seven samples from QA-I consistently display a bimodal distribution with peaks atn-C17/n-C18 and n-C23alkanes, which is similar to Yanwan. These samples have no carbon preference in the short-chain and medium-chain parts, but have distinct odd carbon predominance fromn-C26 to n-C35 (average CPI27–33 is 1.76) (Table S1).
 For the Yaodian section, two categories of bimodal distribution are detected: the reddish-brown mudstones (IV, V, VI inFigure 2) have peaks at n-C17 and n-C23, whereas the marls (I, III) have peaks at n-C17 and n-C31. The category with peaks at n-C17 and n-C23corresponds to fluvial and floodplain sediments, whereas the latter category corresponds to shallow lake sediments. All six samples manifest a strong odd carbon number predominance of long-chainn-alkanes (average CPI27–33 is 2.76) (Table S1).
 The lipid components of epicuticular waxes can be divided into several fractions having similar molecular affinities, such as n-alkanes,n-alkanoic acids, andn-alkanols [Eglinton and Hamilton, 1967]. Previous studies have demonstrated that the characteristics of n-alkane compositions can be used to differentiate their biological origins. For example, long-chainn-alkanes (n-C27, n-C29 and n-C31) with odd-over-even carbon predominance typically originate from the epicuticular waxes of terrestrial higher plants [Eglinton and Hamilton, 1967; Vogts et al., 2009]. In contrast, short-chainn-alkanes, with chain lengths of 14–20 carbons and no distinct odd-over-even predominance, are predominantly ascribed to the algal and bacterial contribution [Han and Calvin, 1969; Giger et al., 1980; Pearson et al., 2007]. Submerged/floating aquatic macrophytes sampled from lakes, such as Myriophyllum, Potamogeton, Charaand many lake sediments are mainly characterized by medium-chainn-alkanes (n-C23 and n-C25) [Aichner et al., 2010; Ficken et al., 2000; Gao et al., 2011; Herzschuh et al., 2005; Lin et al., 2009; Mügler et al., 2008; Pearson et al., 2007], although the growing evidence also indicates that some palustrine Sphagnumspecies also have higher abundance of medium-chainn-alkanes [Bingham et al., 2010; Nichols et al., 2006]. We consequently ascribe the short-chain, long-chain, and medium chainn-alkanes to origins from bacteria and algae, terrestrial higher plants, and typical macrophytes or palustrine plants, respectively. The detection ofMyriophyllum, Potamogeton, Typhaceae and Cyperaceae pollen and algal fossils from Tianshui Basin [Hui et al., 2011; Li et al., 2006] demonstrates that n-C23 and n-C25alkanes may be the biolipid remains of aquatic macrophytes or palustrine plants. Therefore, the above lines of evidence demonstrate that the medium-chainn-alkanes with peaks atn-C23 or n-C25originated from submerged/floating aquatic macrophytes or palustrine plants that were grown in lacustrine-fluvial conditions.
 On the basis of previous interpretation, we infer that the abundant n-C23 and n-C25 alkanes of most of the Tianshui samples (Figures 2 and 3), are probably derived from aquatic macrophytes or moss (e.g., Sphagnum) living in palustrine/peat condition [Bingham et al., 2010; Ficken et al., 2000]. In contrast, n-C23 and n-C25 alkanes are largely absent in Quaternary loess samples (Figure 3). In addition to abundant short-chainn-alkanes, almost all Quaternary loess and late Mio-Pliocene red-clay samples have low abundances in middle-chainn-alkanes, but they have abundantn-C27, n-C29 and n-C31 alkanes [Bai et al., 2009; Liu et al., 2008; Xie et al., 2003; Zeng et al., 2011; Zhang et al., 2008, 2006], which are derived from terrestrial higher plants [Eglinton and Hamilton, 1967; Vogts et al., 2009]. Abundant short- and long-chainn-alkanes and scarce medium-chainn-alkanes of Quaternary loess are consistent with the concept that typical Chinese loess is mainly transported by the winter monsoon and only accumulates on positive topography as proposed byLiu . However, abundant medium-chainn-alkanes for samples from the Tianshui Basin are hard to reconcile with the positive topography concept as claimed by several papers from Guo's group [Liu et al., 2006; Zhan et al., 2011]. In contrast, abundant medium-chainn-alkanes for samples from the Tianshui Basin are consistent with a distal fan and fluvial/lacustrine setting for these sediments as suggested byAlonso-Zarza et al. . Pristane/n-C17 and Phytane/n-C18 values, which have been widely utilized as indicators of depositional environment and organic matter maturation/biodegradation [Duan et al., 2008; Peters et al., 2005; Younes and Philp, 2005], for the three studied sections from the Tianshui Basin have similar values (Figure S1). Considering that sediments from the Yaodian section are obviously fluvio-lacustrine and floodplain deposits [Alonso-Zarza et al., 2009; Li et al., 2006], we argue that the similar Pr/n-C17 and Ph/n-C18values among the three studied sections suggest that sediments of the Yanwan and QA-I sections were deposited under a similar environment.
 The above inference is based on the assumption that these middle-chainn-alkanes are autochthonous instead of allochthonous. If the middle-chainn-alkanes in the studied sections come from eolian input from dried up lake sediments upwind, the above inference will be invalid. A comparison ofn-alkane distribution patterns between the Tianshui and red-clay samples of similar ages should help to resolve an autochthonous versus allochthonous origin of these middle-chainn-alkanes. If middle-chainn-alkanes in the observed sections come from dried up lake sediments, abundant middle-chainn-alkanes should also be observed in the red-clay samples east of the Liupan Shan which is ∼140 km away from the Tianshui samples. Yanwan samples YW-7.5 (No. a), YW-21.5 (No. b) and YW-27.5 (No. c) have similar ages with Chaona samples CN-8, CN-9 and CN-10 [Bai et al., 2009]. However, as we noted before, all of the Chaona samples have low abundance in middle-chainn-alkanes. Therefore, we conclude that the middle-chainn-alkanes are more likely autochthonous instead of allochthonous. In reality, climate drying during Plio-Quaternary should produce more dried up lakebed upwind to the Loess Plateau area. However, many loess plateau sites (the Luochuan and Xunyi sections [Zhang et al., 2006], Chaona section [Bai et al., 2009], Jiuzhoutai section [Xie et al., 2003] and Tawan section [Zeng et al., 2011]) have relative low abundances of middle chain n-alkanes, which suggests that eolian contributions of middle-chainn-alkanes should be limited and abundant middle-chainn-alkanes observed from Tianshui are most likely autochthonous.
 Biodegradation and thermal alteration may also change n-alkane distribution pattern [Peters et al., 2005] and invalidate our inference. Therefore, it is essential to evaluate the degree of biodegradation and thermal alteration before confirming our conclusions. In the Tianshui Basin, the apparent odd carbon predominance of long-chainn-alkanes shown by the CPI values (1.4–5.1) (Table S1) and high abundance of easily biodegradable short-chainn-alkanes imply that the Tianshui samples suffered slight biodegradation. Additionally, the existence of relatively small unresolved complex mixtures (UCMs) in some samples also supports the previous interpretation. Furthermore, Pr/n-C17 and Ph/n-C18 values for all samples are lower than 1 and 1.6 (Figure S1), respectively, indicating that only slight biodegradation occurred for the studied samples [Peters et al., 2005]. Therefore, we conclude that the extent of biodegradation of the Tianshui samples is equivalent to biodegradation level 0∼1 based on the scale of Wenger and Isaksen , which means that our samples have only been subjected to very slight biodegradation and that the medium-chainn-alkanes are unlikely to be the biodegradation byproducts of long-chainn-alkanes. It is also unlikely that thermal alteration has affected the originaln-alkane distributions because the three sections are less than 300 meters thick. Hence the differentn-alkane distribution patterns between the Tianshui samples and Quaternary loess are probably not caused by modifications of bio- and/or thermal-degradation, but are more likely to result from a different origin and depositional environment.
 Although the Tianshui samples have several physical and geochemical similarities to the Quaternary loess, their strikingly different biomarker characteristics are difficult to reconcile with an eolian origin for the Tianshui sediments. The high abundance of medium-chainn-alkanes originating from either aquatic macrophytes or palustrine plants is more consistent with a lacustrine and mudflat/floodplain setting [Alonso-Zarza et al., 2009], with sediments derived mainly from the Qinling Mountain and surrounding mountains. Consequently the biomarker evidence points to the need for caution in identifying loess deposits by field observations and whole rock geochemistry comparisons and casts strong doubt on an early Miocene initiation of desertification within the Asian interior.
 We thank Xiuxi Wang, Bin Liu, Shanpin Liu, Zhanfang Hou for sampling and helpful discussions, Junsheng Nie and Darryl Granger for language improvements, and Qianxiang Meng for laboratory assistance. Philip Meyers and the anonymous reviewer are appreciated for their valuable comments and suggestions. This work was co-supported by the National Natural Science Foundation of China (41021091, 41101012), National Basic Research Program of China (2010CB833401, 2011CB403003) and Ph.D. Foundation of Ministry of Education of China (20090211120027).
 The Editor thanks Philip Meyers and an anonymous reviewer for their assistance in evaluating this paper.