The rotation pattern and fault activity in the southeast margin of the Tibetan Plateau (SEMTP) provide meaningful constraints on the geodynamic evolution of the plateau. However, the lack of Cenozoic paleomagnetic studies and accurate age constraints on Neogene sediments prevents a better understanding of the late Cenozoic tectonic activity in this area. To clarify the tectonic rotation pattern and deformation history of the SEMTP, we report new magnetostratigraphic data from a late Neogene sedimentary sequence in the Dali Basin, northwestern Yunnan Province, China. Rock magnetic analyses indicate that both magnetite and hematite are the main carriers of the characteristic remanent magnetizations (ChRMs). Magnetostratigraphic results show that the sedimentary profile spans from Chron C4n.1r to Chron C2n. The age of the sedimentary sequence in the Dali Basin can thus be paleomagnetically constrained to an interval from late Miocene to early Pleistocene. The basal age of the sediments is ~7.6 Ma, which indicates that the unroofing of Diancang Shan and activation of the Dali fault system were initiated at this time. The appearance of conglomerates and syntectonic sediments suggests the reactivation of the Dali fault system at ~2.5 Ma. Moreover, the overall mean ChRM direction suggests that the Dali Basin experienced significant (4.4 ± 2.5°), but minor post–late Miocene rotation. This indicates that most of the clockwise rotation demonstrated by previous paleomagnetic studies in the SEMTP occurred prior to late Miocene and may be concentrated between Eocene and Miocene, which is contemporaneous with the sinistral slip of the Ailao Shan-Red River fault.
 The Tibetan Plateau, the largest and highest plateau on Earth, is the product of ongoing collision between the India and Eurasian plates [Molnar and Tapponnier, 1975]. The timing and style of the development of high elevation in central Asia have great implications for understanding the processes of continental collision [Houseman and England, 1993; Tapponnier et al., 2001; Royden et al., 2008]. Paleoclimatic studies have suggested that the onset of the Asian monsoon system and aridification of central Asia may be related to the uplift of the Tibetan Plateau or the retreat of the Paratethys Sea [Kutzbach et al., 1989; Molnar et al., 1993; An et al., 2001; Guo et al., 2002; Bosboom et al., 2011]. Thus understanding the evolution history of the Tibetan Plateau has important implications to the evolution of the plateau and the interaction between the lithosphere and atmosphere.
 The southeast margin of the Tibetan Plateau (SEMTP) (Figure 1) is one of the key areas to understanding the geodynamic mechanisms of the Tibetan Plateau evolution, such as continuous deformation [e.g., Houseman and England, 1993], rigid block deformation [Tapponnier et al., 1982; Tapponnier et al., 2001], and lower crustal flow [Bird, 1991; Clark and Royden, 2000]. Paleomagnetic constraints on the precise timing of uplift processes, tectonic rotation patterns, and deformation history of this area can help distinguish between these deformation processes in the SEMTP region as well as the Tibetan Plateau. However, most of the previous paleomagnetic studies were concentrated on Mesozoic, specifically Cretaceous, or early Cenozoic rocks [e.g., Achache et al., 1983; Yang and Besse, 1993; Yang et al., 2001; Sato et al., 2007; Kondo et al., 2012] (Figure 1). The lack of late Cenozoic paleomagnetic studies and precise age constraints on Neogene sediments prevents us from better understanding the effects of the India-Eurasian collision on the SEMTP [Zhu et al., 2008; Li et al., 2012]. Additional and more detailed paleomagnetic studies of Cenozoic rocks from this region, therefore, are essential.
 In this study, we conduct a high-resolution magnetostratigraphic investigation of the late Miocene to Quaternary sediments of the Dali Basin, located at the northeast of the Diancang Shan (DCS) in the SEMTP (Figure 2). Our work is focused on the Sanying Formation because it may record a series of tectonic events and geomorphic processes, such as the uplift of SEMTP [Clark et al., 2005; Schoenbohm et al., 2006a], the reorganization of the regional drainage network [Clark et al., 2004; Kong et al., 2009, 2012], the reversal of the Red River fault [Leloup et al., 1995; Replumaz et al., 2001], and the reactivation of the left-lateral Dali fault system [Wang et al., 1998]. The Sanying Formation is widespread at the front of hills and in the fault-controlled basins in western Yunnan [Ge and Li, 1999]. It was thus extensively used to constrain the age of the above-mentioned events as well as paleoclimate reconstruction [Xu et al., 2008; Xia et al., 2009; Sun et al., 2010]. Wang et al.  also found that the Sanying Formation is deposited uncomformably on rocks of Paleozoic to early Cenozoic age, and hence, they concluded that the age of the formation represents the upper age of the pre-Pliocene erosion surface throughout much of western and central Yunnan [Clark et al., 2006]. However, none of these studies provide unambiguous age constraints for the Sanying Formation.
 In this study, we first describe detailed rock magnetic analyses to determine the magnetic carriers and then present a magnetic polarity stratigraphy to accurately constrain the age of the sedimentary sequence. Finally, based on the sediment accumulation rates derived from magnetochronology, combined with lithological changes, we will present an interpretation of the infilling process and the tectonic rotation of the Dali Basin and discuss its implications to the evolution of the SEMTP.
2 Geological Setting and Sampling
2.1 Geological Framework of the SEMTP
 The SEMTP is a fan-shaped area broadened outward to the southeast and narrowed to the northwest [Yunnan Bureau of Geology and Mineral Resources (YBGMR), 1990]. It is bounded by the Gaoligong-Sagaing fault to the west and the Xianshuihe-Xiaojiang fault to the east (Figure 1). In contrast to the steep topographic gradients in the southern and eastern margins of the Tibetan Plateau, the SEMTP is characterized by lower relief with a gradual change in topographic elevation and crustal thickness from the plateau to the outer margin [Clark and Royden, 2000; Wang et al., 2003; Royden et al., 2008]. Also, in contrast to the southern and eastern periphery of the Tibetan Plateau, where the crustal thickening is mainly accommodated by folds and thrust faults [e.g., Decelles et al., 2002], there is little evidence of upper crustal shortening in the SEMTP in the late Cenozoic [Wang and Burchfiel, 1997]. Instead, the GPS measurements, widespread normal faults as well as N-S directed basins and lakes in this area, indicate that most of the SEMTP are currently undergoing E-W extension rather than contraction [Leloup et al., 1995; Wang et al., 1998; Gan et al., 2007].
 The most striking geological feature in the SEMTP is the occurrence of large-scale strike-slip faults [Tapponnier et al., 1990; Leloup et al., 1995], such as the Sagaing fault, the Gaoligong shear zone, the Chongshan shear zone, the Ailao Shan-Red River (ASRR) shear zone and the Xianshuihe-Xiaojiang fault from west to northeast (Figure 1). The SEMTP is divided into three terranes by these strike-slip faults: the Chuandian, Shantai, and Indochina terranes (also see Figure 1). The Chuandian terrane is further divided into Yunling Collage, Dali Highland, and Chuxiong Basin by the Dali fault system [Wang et al., 1998] (Figure 2b). The area of this study, the Dali Basin, is located in the Dali Highland, south of the first bend of the Yangtze River.
 The dextral strike-slip Sagaing fault (Figure 1) is initiated at ~11 Ma, with a 500 km displacement resulting from the opening of the Andaman Sea [Khan and Chakraborty, 2005]. The ASRR shear zone is a major geological discontinuity belt of gneissic massifs, including Xuelong Shan, DCS, Ailao Shan, and Day Nui Con Voi from northwestern Yunnan to Vietnam, extending more than 1000 km long but no more than 20 km wide [Leloup et al., 1995]. It was a ductile left-lateral shear zone formed from Oligocene to Miocene and a brittle right-lateral fault from late Miocene to present. The time and displacement along the early left-lateral ASRR shear zone are strongly debated over the last decades. One school argues that the amount of sinistral offset along the ASRR shear zone is 700 ± 200 km and that the slip was activated from Oligocene to mid-Miocene (35–17 Ma) [Leloup et al., 2001; Gilley, 2003]. They also propose that the activity of the ASRR shear zone is kinematically linked to the opening of the South China Sea [Briais et al., 1993]. The second school of thought, however, argues that the displacement of the ASRR shear zone is 250 km [Hall et al., 2008; Searle et al., 2010] and that the age of slip ranged from 32 to 22 Ma [Searle et al., 2010], or a shorter time of 27–22 Ma [Wang et al., 2000]. Although the recent restoration of Cenozoic deformation in Asia by van Hinsbergen et al.  argues that a smaller displacement is more reasonable and consistent with available geological constraints, more detailed studies are still needed to better characterize this crucial fault. The dextral transtensional slip along the Red River fault began only ~5 Ma and has since accumulated ~40 km of displacement [Leloup et al., 1995; Replumaz et al., 2001; Schoenbohm et al., 2006b]; how this slip is partitioned into the Dali fault system remains unclear.
 The ductile sinistral Chongshan and dextral Gaoligong shear zone were formed in Oligocene and are thought to be conjugate shear zones [Wang et al., 2006; Akciz et al., 2008]. They deformed contemporaneously with the ASRR shear zone, and all of their motions accommodated the southeastward extrusion of Southeast Asia. The left-lateral Xianshuihe-Xiaojiang fault, which is considered to be a boundary fault to accommodate the different rotations around the eastern Himalayan syntaxis, initiated at ~13 Ma on the Ganzi-Yushu segment in the north and at ~5 Ma on the Xiaojiang fault in the south, and the total offset is 78–100 km [Wang et al., 1998; Wang et al., 2009].
 The Dali left-lateral fault system, a series of second-order faults in the SEMTP, including the NE-trending Jianchuan, Lijiang, and Heqing faults; the NW-trending Tongdian, Daju, and Zhongdian faults; and the north-trending Chenghai fault (Figure 2), are thought to be active in the last 2–4 Myr based on the sediments deposited in the fault-related basins, but the detailed history of these faults remains unclear [Wang et al., 1998].
2.2 Neogene Sediments of the Dali Basin
 The Dali Basin is formed by the east-dipping Erhai normal fault system, which is expressed by a series of triangle fault facets dipping eastward [e.g., Wang et al., 1998; Fan et al., 2006]. Most of the Dali Basin is covered by Quaternary sediments, with few outcrops of Paleogene conglomerates and Neogene silts (Figure 3) [YBGMR, 1990]. The Neogene Sanying Formation is assigned to the Pliocene based on the occurrence of plant fossils, such as Quercus pannosa (H.-M.), Ulmus cf. miopumila (Hu et Chaney), and Ulmus hedini (Chaney) [YBGMR, 1990].
 The Sanying Formation is well preserved in the northeast part of the Dali Basin (Figures 3 and 4). It overlays Silurian limestone and is disconformably onlapped by Quaternary unconsolidated conglomerates (Figures 3b and 4a). The section in this study is a combination of two sections exposed in road cuts and separated by ~1 km (Figure 3). At the lower part of the upper section, the sediments are predominantly red medium-coarse sands. Two dark gray peaty beds (~0.5 m each) are exposed at the bottom (Figure 4b). The sediments at the upper part of the lower section are also red medium-coarse sands but are replaced by sands interbedded with coal-bearing peaty clay up section (Figures 4c–4e). We correlate the two peaty beds of the upper section to the first appearance of peaty clay in the lower section (Figure 3b). To confirm this correlation, we also collected samples with an overlap of ~30 m of the two sections. The same lithostratigraphy comprising two marker layers and the same polarity recorded by the overlapped sections (see Supplementary Table 1) are consistent with this correlation. The tape measured composite section has a thickness of ~1000 m, including approximately 35 m of Quaternary conglomerates (Figures 3b and 4a). The section tilts homoclinally to the east, with little change in dip. At the top of the section (above 806.8 m), however, the tilt of the beds shallow from ~40° to 8°. The decrease in bedding dip toward the top of the section, a notable decrease in bedding thickness toward the crest of the fold, and a series progressive onlapping strata (Figures 3b and 4f) are all hallmarks of growth strata [e.g., Riba, 1976], or a syntectonic sediments caused by normal fault.
 The sequence can be divided into four facies associations (FA) based on up-section lithological changes. FA1 mainly consists of coal-bearing clays interbedded with fluvial medium- to coarse-grained, poorly laminated sands (interpreted as swamp and fluvial deposits; Figures 4c–4e); FA2 is chiefly fine- to medium-grained, multiple-colored sands and silts, with individual beds being a few to tens of centimeters thick (interpreted as fluvial or lacustrine deposits; Figures 4b and 4g); FA3 is dominated by gray and light brown thin-bedded clays and silts with interbeds of massive matrix-supported conglomerates and sands (interpreted as lacustrine deposits; Figures 4h and 4i); and FA4 is comprised of alternating matrix-supported, pebble-cobble conglomerate facies and laminated coarse-grained sand facies (interpreted as fluvial and alluvial deposits; Figures 4f, 4j, and 4k).
 Samples were collected using either a gasoline-powered drill (n = 4060) or as oriented hand samples (n = 45) based on the out-crop quality of the sediments. The sampling interval is typically 0.2–0.5 m but is as much as ~10 m in the conglomeratic intervals. Two specimens were drilled at each sampling level. All samples were oriented with a magnetic compass. All samples were cut into either specimen cylinders 2.5 cm in diameter and 2 cm in height or cubes with 2 cm edges.
3.1 Rock Magnetic Measurements
 A total of 449 specimens with a nearly equal sampling interval (~2 m) were selected to measure anisotropy of magnetic susceptibility (AMS) using a KLY-4s Kappabridge before thermal demagnetization was conducted. Other rock magnetic measurements, including thermal demagnetization of three-component isothermal remanent magnetization (IRM), hysteresis loops, IRM acquisition, and DC field demagnetization of the saturated IRM (SIRM), were made on representative specimens.
 The composite IRMs were performed using a 2G Enterprises Pulse Magnetizer (2G660). The hard, medium, and soft components were treated in DC fields of 2.7 T, 0.5 T, and 0.15 T along three mutually orthogonal axes [Lowrie, 1990]. The specimens were then subjected to progressive thermal demagnetization to 685°C at 10–50°C intervals using a thermal demagnetizer (ASC TD-48).
 Hysteresis loops were measured using a Princeton Measurements Corp. MicroMag 3900 Vibrating Sample Magnetometer (VSM). The magnetic field was cycled between ±1.5 T. Saturation magnetization (Ms), saturation remanence (Mrs), and coercivity (Bc) were determined after the correction for paramagnetic contribution identified from the slope at high fields. Specimens were then demagnetized in alternating fields (AFs) up to 1.5 T, and an SIRM was imparted from 0 to 1.5 T also using the MicroMag 3900 VSM. The SIRM was then demagnetized in a stepwise backfield up to −1.0 T to obtain the coercivity of remanence (Bcr). The IRM acquisition curves of the selected specimens were then analyzed using the MAG-MIX package of Egli  to determine the magnetic coercivity distributions.
3.2 Demagnetization of the Natural Remanent Magnetization
 Remanence measurements were measured using a three-axis cryogenic magnetometer (2G Enterprises, USA) installed in field-free space (<300 nT). A total of 1226 selected specimens with ~1 m intervals were subjected to thermal, AF, or hybrid demagnetization. First, 775 specimens were subjected to progressive thermal demagnetization up to a maximum temperature of 585°C or 685°C with 25–50°C intervals below 585°C and 10–15°C above 585°C, using a Magnetic Measurements thermal demagnetizer with a residual magnetic field less than 10 nT. Sixty additional specimens were subjected to progressive AF demagnetization up to a maximum field of 60–70 mT, with 5–10 mT intervals. Finally, 391 new specimens were first subjected to an 80°C and 120°C thermal demagnetization, and then by AF demagnetization at peak fields up to 70 mT.
 Demagnetization results were evaluated by orthogonal vector diagrams [Zijderveld, 1967], and the principal components directions were computed by least-squares fits [Kirschvink, 1980]. Data analysis was completed using PaleoMac [Cogné, 2003] and pmag.py [Tauxe, 2010] software packages. The mean directions were computed using Fisher statistics [Fisher, 1953].
4.1 Rock Magnetic Results
4.1.1 Thermal Demagnetization of Three-Component IRM
 Stepwise thermal demagnetization of the three orthogonal components IRM [Lowrie, 1990] is a powerful method to distinguish the magnetic minerals with different coercivities in a sample [Liu et al., 2010; Tauxe, 2010]. Figure 5 shows the thermal demagnetization curves of the three-component IRM of selected specimens, which show distinct behaviors between different lithologies. For the specimens from gray clays or silts, or dark gray peaty clays, the low-coercivity component shows unblocking temperatures of 300–450°C (e.g., Figures 5f, 5g, and 5k) or 550°C (e.g., Figure 5h), indicative of the presence of titanomagnetite or magnetite. For the specimens from red silts or sandy silts, the high- and medium-coercivity components or all the three components show an unblocking temperature of 680°C (e.g., Figures 5a and 5c), signifying the dominant presence of hematite. We also found a subtle inflection at the temperature of ~550°C (e.g., Figures 5e, 5i, and 5j), which indicates the presence of magnetite. These behaviors suggest that titanomagnetite, magnetite, and hematite are the predominant remanence carriers of the Sanying Formation sediments.
4.1.2 Hysteresis Properties and IRM Acquisition Curves
 Hysteresis loops and IRM acquisition curves (Figure 6) are also used to assess the magnetic mineralogy. These measurements also show distinct behaviors between different lithologies. The specimens from gray clays and silts show hysteresis loops that close above 0.5 T (Figures 6a and 6j) and have high S ratios of >0.8 (Figures 6b and 6k), indicating the dominance of soft magnetic components. These observations are consistent with the derivative curves of IRM component analyses (Figures 6c and 6l), which show a pronounced low-coercivity component and one or two much subdued medium- or high-coercivity components. However, the specimens from red silts and sandy silts show distinctive goose-necked hysteresis loops [Roberts et al., 1995; Tauxe et al., 1996] with an open nature of the loops up to fields of 1.2–1.5 T (e.g., Figures 6d and 6g) and much lower values of the S ratio than the other lithologies (Figures 6e and 6h). Accordingly, the IRM component analyses show a two-humped distribution, with a pronounced high-coercivity component and a substantially subdued medium-coercivity component (Figures 6f and 6i). These observations suggest the coexistence of low- (e.g., magnetite) and high-coercivity (e.g., hematite) phases.
 In summary, the synthetic rock magnetic experiments suggest that the Sanying Formation sediments contain a mixture of magnetic minerals, such as titanomagnetite, magnetite, and hematite.
4.2 Anisotropy of Magnetic Susceptibility
 The AMS is a fast and powerful technique that provides valuable information on the origin and subsequent deformation about rocks. The magnetic fabric of sediments strongly depends on the depositional processes. For the undeformed fine-grained sediments, the AMS ellipsoid reflects the effect of water flow and compaction, so the primary AMS should have an oblate shape with the maximum principle susceptibility (K1) axes parallel to the bedding plane and the minimum principle susceptibility (K3) axes perpendicular to the bedding plane [Tarling and Hrouda, 1993]. Figure 7 shows stereonet projections of the maximum principle (K1) and minimum principle (K3) susceptibility axes before and after tilt correction. The K1 axes generally strike north to south, with a low inclination in both geographic and stratigraphic coordinates, and are approximately parallel to the bedding attitude (the mean K1 before and after tilt correction is D/I = 181.0°/1.3° and 0.2°/3.7°, respectively). The K3 axes are tilted to the west in geographic coordinates but are subvertical in stratigraphic coordinates (the mean K3 before and after tilt correction is D/I = 273.3°/60.2° and 226.0°/84.6°, respectively). This shape of the susceptibility ellipsoid, combined with the relatively low value of PJ (corrected anisotropy degree) (ranging from 1.001 to 1.068, with an average of 1.02), clearly suggests a primary sedimentary fabric that has been undisturbed since its deposition. The clustering of the minimum susceptibility axes near vertical following restoration of bedding to horizontal indicates that the fabric was acquired prior to deformation.
4.3 Paleomagnetic Results
 Representative demagnetization diagrams are shown in Figure 8. Thermal and AF demagnetization isolated several distinct directions. A low temperature or low field component was removed below ~300°C or ~10 mT. The mean direction (in situ coordinates) of this component is indistinguishable from the present Earth field or geocentric axial dipole field direction, and we interpret this component to be a recent viscous overprint. The characteristic remanent magnetization (ChRM) component was separated between 300–350°C and 585°C (e.g., Figures 8a, 8b, 8d, 8e, and 8p) or between 30 mT and 60 mT (e.g., Figures 8i and 8j). However, for some specimens, the high-stability ChRM component persists up to 685°C (e.g., Figures 8c, 8h, 8k, and 8l); the ChRM directions were determined using only the points above 610°C for these specimens. At least four successive demagnetization steps trending toward the origin were selected to determine the ChRM direction. Specimens with a maximum angular deviation larger than 15° were rejected from further analyses. A total of 669 specimens (54.6%; representing 669 sampling levels) gave reliable ChRM directions, with 424, 38, and 207 specimens demagnetized with thermal, AF, and hybrid techniques, respectively. Although there may be many reasons for the high rejection rate, we suggest two predominant causes. First, the silty clays and sands at the lower part of FA2 are tan and maroon, but the colors cut across the sedimentary beddings, indicating a significant influence of fluids within these sediments [Wang et al., 1998]. The specimens from these sediments usually have strong low-temperature components and show erratic direction behaviors during high-temperature demagnetization process; thus no reliable results can be obtained; second, the coal-bearing peaty sediments at the bottom of the section, which have a very weak magnetization because of rare magnetic minerals in them, do not yield reliable results.
 Virtual geomagnetic pole (VGP) latitudes assuming a 100% geocentric axial dipole field were calculated from the ChRM data to construct the magnetostratigraphy. As shown in Figure 9, 26 magnetozones were identified in the studied section: 13 with normal polarity (N1–N13), and 13 with reverse polarity (R1–R13). Each polarity zone was determined using at least four paleomagnetic sampling levels.
5.1 Reliability of the Paleomagnetic Directions
 The generally homoclinal structure of the sampled intervals prevents us from conducting a paleomagnetic fold test with any statistical rigor. However, the observed dual polarity does make a reversal test possible. After rejecting 248 specimens with VGP latitudes less than 60°, which may signal a transitional field behavior or due to the complex process of sediments to acquire remanent magnetizations [e.g., Verosub, 1977], the remaining 421 specimens have a mean normal direction of D/I = 7.4°/29.7° (α95 = 2.0°, n = 209) and reversed direction of D/I = 183.9°/–26.0° (α95 = 2.1°, n = 212); the overall mean direction is D/I = 5.6°/27.9° (α95 = 1.5°, n = 421) (Table 1). However, the reversal test is negative at the 95% confidence level [McFadden and McElhinny, 1990], which primarily may result from differences in inclination values and the small error ellipses that encompass the mean directions. The steeper inclination of normal direction compared with that of reversed direction is most likely due to an unremoved overprint that will steepen the normal directions and shallow the reversed directions [Gilder et al., 2001]. Such an overprint may be especially problematic for specimens cleaned with AF demagnetization because AF treatment cannot remove high-coercivity components carried by, for example, goethite or hematite. We do observe a slight significant difference between direction populations: the mean direction of thermally demagnetized specimens is D/I = 5.3°/26.1° (α95 = 1.7°, n = 289), whereas for AF and hybrid demagnetized specimens, it is D/I = 6.4°/31.7° (α95 = 3.0°, n = 132) (Table 1). The directions isolated by thermal demagnetization have a mean normal direction of D/I = 6.6°/27.8° (α95 = 2.6°, n = 114) and reversed direction of D/I = 184.5°/-25° (α95 = 2.3°, n = 175); they pass the reversal test at 95% confidence level with A classification [McFadden and McElhinny, 1990]. Thus, we believe that the ChRM directions obtained in this study are primary and reliable. Since there is no significant difference between the mean direction of the thermally demagnetized specimens and the mean direction of all specimens, we use the latter (D/I = 5.6°/27.9°; α95 = 1.5°, n = 421) in our discussion of rotations.
Table 1. Mean ChRM Directions From the Sanying Formation
Dexp ± ΔDexp (°)
R ± ΔR (°)
n, numbers of specimens. D, I, and α95 are the mean declination, inclination, and 95% confidence limit, respectively. Dexp and ΔDexp are the expected declination and its confidence limit at the sampling site calculated from the coeval reference pole. ChRM, characteristic remanent magnetization. TD, thermal demagnetization. AF, alternating field demagnetization. The 10 Ma Eurasian reference pole is at λp = 87.2°N, φp = 125°E (A95 = 2.5°) [Torsvik et al., 2008]. The vertical-axis rotation with 95% confidence limit, R ± ΔR, was calculated following Butler .
1.4 ± 2.9
4.2 ± 2.7
4.4 ± 2.5
5.2 Inclination Shallowing
 We also calculated the mean direction recorded by magnetite and hematite, respectively (see Supplementary Table 2). The inclination of the mean direction is 33.6° (n = 211) for magnetite and 25.5° (n = 210) for hematite. Lower magnitudes of inclination shallowing in magnetite compared to hematite have been observed in several other stratigraphic sections [e.g., Gallet et al., 2012; Deng et al., 2013]. The steeper inclination recorded by magnetite probably originated from a small relative flattening effect between magnetite and hematite grains that occurred during sediment compaction, with the former less prone to flattening. This is consistent with the morphology of the magnetite and hematite, as magnetite is usually equant or elongate, whereas hematite is more platy and, hence, much easier to bias toward the horizontal during depositional and postdepositional compaction [Iosifidi et al., 2010]. However, when compared with the expected inclination derived from the 10 Ma apparent polar wander path (APWP) of Eurasia [Torsvik et al., 2008], ~20° of inclination shallowing is calculated for both of the inclinations.
 The inclination shallowing of sediments, especially for the red beds in central Asia, has been a longstanding problem during the past decades [e.g., Gilder et al., 2001; Gilder et al., 2003; Tan et al., 2003; Huang et al., 2005; Cogné et al., 2013]. Many mechanisms have been proposed to interpret it, including a contribution of nondipole field or local geomagnetic anomaly, the nonrigidity of Eurasia, the inaccuracy of Europe's APWP relative to that of Asia, and the depositional and postdepositional processes (see detailed discussion in, e.g., Dupont-Nivet et al.  and references therein). The steeper inclination recorded by magnetite than hematite may suggest that depositional and postdepositional processes are the main cause of inclination shallowing in this study. Our study is not concerned about inclination shallowing, however, and it does not affect our rotation calculations; therefore, we refrain from discussing the inclination shallowing in the Sanying Formation any further.
5.3 Age Constraints of the Sanying Formation
 On the basis of paleomagnetic, paleobotanic, and sedimentologic constraints, the recognized magnetozones can be correlated with the geomagnetic polarity timescale (GPTS) of Gradstein et al.  (Figure 9). Tao and Kong  described 14 species of plant fossils in the Sanying Formation at the same location as our study. These include Quercus pannosa (H.-M.), Quercus semicarpifolia Sm., Quercus gilliana R. et W, Acer paxii Franch, Pinus yunnanensis Fr., and Celtis bungeana Blume. Fossils of shell and leaf are also found in the section at the thicknesses of 325 and 896 m, respectively (Figures 4l and 4m). A detailed palynoflora study by Kou et al.  identified 58 palynomorphs belonging to 49 families. The fossil floral and sporo-pollen assemblages are very similar to the Quercus pannosa in Shisapangma, indicating a middle to late Pliocene age [Tao and Kong, 1973; YBGMR, 1990]. The lower part of section FA1 (see Figures 3 and 9a) consists of peaty clay or lignite interbedded with sands or silts. These kinds of organic-rich sediments, which contain hominoid Lufengpithecus sp., are widespread in Yunnan, such as in the Kaiyuan and Lufeng Basins [e.g., Qi et al., 2006]. The hominoid-bearing sediments have been assigned to the late Miocene based on biochronology and magnetochronology [Flynn and Qi, 1982; Zhu et al., 2005; Qi et al., 2006]. Finally, Quaternary sediments overlay the Sanying Formation. Collectively, these observations indicate that the age of the sedimentary sequence of the Sanying Formation can be constrained to an interval from late Miocene to Pliocene.
 When correlating to the GPTS, we give the most weight to the three long reverse magnetozones, R2, R9, and R11. These reversed magnetozones may correlate to Chrons C2r, C3r, and C3Ar (Figures 9e and 9f) or to Chrons C1r, C2Ar, and C3r (Figures 9f and 9g). The upper middle of the section is dominated by normal polarity, which is consistent with the Gauss normal chron. Thus, we propose two possible correlations of the recognized magnetozones to the GPTS (Figures 9e–9g). We do not correlate magnetozones N1 and R1 to the GPTS due to the presence of a sedimentation gap at the top of FA4 (Figure 4a). As a result, magnetozones N2–N13 and R2–R13 can be correlated to Chrons C4n.1r through C2n (correlation I, see Figures 9e and 9f) or those from C3Ar to C1r.1n (correlation II, see Figures 9f and 9g). We prefer correlation I because correlation II requires more dramatic variations of sediment accumulation rates (Figure 10) and higher sediment accumulation rates in fine-grained sediments in the lower part of FA3. However, we note that in our preferred correlation, the correlation of magnetozones N7 to N9 with Chrons C3n.1n to C3n.4n is weak due to the absence of the short polarity interval C3n.1r. A hiatus in sedimentation or a secondary overprint related to fluid flow [Wang et al., 1998] may be the most plausible reasons. Furthermore, two short chrons (normal Chrons C3Br.1n and C3Br.2n, or reverse Chrons C3Br.1r and C3Br.3r) are missing. The lack of the two short polarity intervals may be attributable to the low success rate in coal-bearing sediments. Nevertheless, our preferred correlation represents the best available fit in terms of interval number and duration.
 Therefore, the age of the Sanying Formation in the Dali Basin was paleomagnetically constrained to a span from ~7.6 Ma to ~1.8 Ma, representing a late Miocene to early Pleistocene age. This is the first precise age constraint for the late Cenozoic sediments in northwest Yunnan. Our magnetostratigraphy results indicate that the Miocene/Pliocene and Pliocene/Pleistocene boundaries are within the middle part of FA2 and the lower part of FA4 of the Sanying Formation, respectively, with the former being located at the thickness of 385.6 m, and the latter, at the thickness of 772.8 m (Figures 9a–9f).
5.4 Infilling of the Dali Basin and Unroofing of the DCS
 A plot of magnetostratigraphic age versus depth (Figure 10) shows that the sediment accumulation rates vary from 9.4 cm/kyr to 22.8 cm/kyr. The variability in sediment accumulation rates may reflect tectonic or climate change during the infilling process of the Dali Basin.
 The magnetostratigraphy constrains the onset age of deposition in the Dali Basin at ~7.6 Ma, the late Miocene. Since the deposition of the Sanying Formation has a kinematic link with the activity of the left-lateral Dali fault system and the Erhai normal fault, the 7.6 Ma onset of deposition reported here may indicate the inception of the left-lateral Dali fault system and the Erhai normal fault and the exhumation of the DCS at this time. However, this time is significantly earlier than the age of 4.7 ± 0.1 Ma obtained by Leloup et al.  through 40Ar/39Ar dating of K-feldspar. We attribute this difference to two aspects: On the one hand, the age of 4.7 ± 0.1 Ma is an average of six K-feldspar age plateaus. However, as pointed out by the authors, one of the samples shows a slight earlier age than the other samples. Earlier ages of 9.5 ± 0.5, 7.3 ± 0.1, and 6.4 ± 0.3 Ma were also obtained from the coexisting biotite. Thus, the exhumation of the DCS may begin as early as 10 Ma. On the other hand, all the samples in Leloup et al.’s  study for thermochronological analyses are from the southern end of the DCS. Thus, the different ages may represent different exhumation time of DCS between the south and north. The difference is consistent with the stratigraphy deposited in the basins around the DCS. Actually, the late Miocene to Pliocene sediments are only deposited in the Eryuan and Qiaohou Basins [YBGMR, 1990; Leloup et al., 1993], the east and west sides of northern DCS, while only Quaternary sediments are deposited in the Lake Erhai Basin, the northeast side of southern DCS. Furthermore, the Erhai normal fault is not a single fault, comprising at least two segments (Figure 2).
 The infilling of the Dali Basin can be classified into three stages based on the variation in sediment accumulation rates. The first stage spans from 7.6 Ma to 6.4 Ma and has a sediment accumulation rate of 22.8 cm/kyr. The lithology during this stage is characterized by swamp coal-bearing peaty clay interbedded with fluvial sands. It represents the initial fault-controlled subsidence. We conclude that the relative high sediment accumulation rate during this stage was associated with the initial formation of the basin and was high because of tectonic subsidence. The second stage, between 6.4 Ma and 3.6 Ma, has a relatively lower accumulation rate of 9.4 cm/kyr. Sediments during this interval are dominated by red sands, gray silts, and silty clays. The sediment accumulation rates increase abruptly from 9.4 cm/kyr to 21.2 cm/kyr at 3.6 Ma. Meanwhile, the first occurrence of conglomerates in the Sanying Formation, which was clearly derived from Paleogene conglomerates in the field observations, may indicate the activation of the fault that bounds our section in the east (Figure 3a) and exposes the Paleogene conglomerates. The clearest and most prominent feature at the top ~200 m of the section is the development of syntectonic sediments and silts replaced by massive conglomerates. The appearance of syntectonic sediments dominated by conglomerates indicates active deformation at this time. Wang et al.  concluded that sediments in basins related to the Dali fault system mainly consist of coarse-grained strata with abundant conglomerates in contrast to underlying fine-grained, coal-bearing strata. We suggest that the appearance of syntectonic coarse-grained sediments at ~2.5 Ma resulted from a reactivation of the left-lateral Dali fault system. Moreover, the basal age of the sedimentary sequence in the Heqing Basin, another fault-bound basin in the Dali fault system ~35 km north of our section (Figure 2a), is 2.7 Ma [Shen et al., 2010; An et al., 2011]. Thus, we conclude that the Dali fault system experienced an enhanced activity in the last 3 Myr, and normal-sense displacement is dominant throughout this interval. The increasing normal component activity of the Dali fault system at 2.5 Ma resulted in the rapid exhumation of mountains in this area, such as the DCS and the Yulong Snow Mountain (Figure 2b), the rapid infilling of the adjacent basins with coarse-grained sediments in the early of Quaternary, and the deformation of sediments in the basins. The end of the fluvial and lacustrine deposits of the Sanying Formation at ~1.8 Ma marks the transition of aggradation to degradation of the sedimentary basin filling, which may record another phase of fault activity at this time.
5.5 Clockwise Rotation of the Dali Basin
 The clockwise rotation of the SEMTP has been proposed by many previous authors. Based on a plastic block model experiment, Tapponnier et al.  suggested that the SEMTP rotated 25–40° clockwise as a consequence of extrusion. However, England and Molnar  proposed that the crustal blocks between the east-striking left-lateral faults in the eastern Tibetan Plateau rotated clockwise 1–2°/Myr within a N-S trending right-lateral simple shear zone extending along the eastern margin of the Tibetan plateau from the Eastern Syntaxis to northeast Tibet. In a third kinematic model, Wang et al.  suggested that the clockwise rotation is related to displacements along individual structures in the fault network. This network includes the Xianshuihe-Xiaojiang fault, the Red River fault, and the Dali fault system in the SEMTP.
 Although the mechanisms of rotation are different in each of the models listed above, the clockwise rotation of the SEMTP has been confirmed on short timescales by modern GPS observations [e.g., Zhang et al., 2004; Gan et al., 2007] and on long timescales by paleomagnetic studies [e.g., Yang and Besse, 1993; Yang et al., 2001; Sato et al., 2007; Zhu et al., 2008; Kondo et al., 2012]. A recent review of all the previous paleomagnetic studies in the SEMTP by Li et al.  indicates that, since the Cenozoic, with respect to the stable Eurasian block, most of the SEMTP experienced a clockwise rotation of ~20–80°, with some areas experiencing clockwise rotation by as much as 135° (Figure 1 and Supplementary Table 3), indicating significant internal deformation in this area. Notably, however, most of these paleomagnetic studies were conducted on Cretaceous sediments. The lack of paleomagnetic data from upper Cenozoic units prevents us from understanding the distribution of vertical axis rotations in time.
 Our study, however, provides paleomagnetic data from late Miocene sedimentary units. The sample ChRM directions converted to virtual geomagnetic poles (VGPs) for the Sanying Formation give a mean pole position at 78° N, 252°E, A95 = 1.2° (Table 1). These results indicate 4.4 ± 2.5° of clockwise vertical axis rotation of the Dali Basin since ~2 Ma with respect to the 10 Ma Eurasian reference pole [Torsvik et al., 2008]. The observed direction is D/I = 5.8°/29.2°, with ΔD = 1.2°, ΔI = 2°, whereas the expected direction at Dali is De = 1.4°, Ie = 47.8°, with ΔD = 2.9°, ΔI = 2.9°.About 200 km to the east, in the Xianshuihe-Xiaojiang fault system, the Yuanmou Basin (Figure 1) has rotated 12° clockwise since 4.9 Ma [Zhu et al., 2008]. This similar timing but significantly different magnitude of rotation between the Dali and Yuanmou regions suggests the importance of local structures in determining the magnitude of vertical axis rotations, rather than a regionally coherent block rotation. As stated previously, the large-scale movement along the ASRR fault occurred during Oligocene to mid-Miocene; the Plio-Pleistocene right-lateral movement is just tens of kilometers, consistent with the high magnitude of rotation in Cretaceous units and lower magnitude of rotation observed in late Miocene and younger units. The displacement along the Plio-Pleistocene Xianshuihe-Xiaojiang fault is nearly 100 km, and GPS measurements also suggest that the velocity along the Xianshuihe-Xiaojiang fault is 2–5 times faster than that of the Red River fault and Dali fault system [Wang et al., 1998; Zhang et al., 2004; Gan et al., 2007]. Paleomagnetic data, GPS measurements, and geological observations indicate that the different rotations in the Chuandian terrane were accommodated by differential displacement along the Xianshuihe-Xiaojiang fault and the Dali fault system.
 The available paleomagnetic results throughout the SEMTP also provide valuable information about the timing of rotations. As discussed above, paleomagnetic data from the pre-Cenozoic units show that the SEMTP experiences a significant and large clockwise rotation. However, the only two paleomagnetic results from the Neogene units of the Mae Moh Basin in northern Thailand [Benammi et al., 2002; Coster et al., 2010] show conflicting results. Benammi et al.  suggested a counterclockwise rotation of about 13 ± 1.32° with respect to the reference of Besse and Courtillot, 1991, while Coster et al.  proposed no significant vertical axis rotation with respect to the 10 Ma Eurasian reference pole [Besse and Courtillot, 2002]. To solve this conflict, we recalculated the vertical axis rotation using the 10 Ma Eurasian APWP as a reference pole [Torsvik et al., 2008], and results show only minor amounts of rotation as revealed by this study (Figure 1 and Supplementary Table 3). These results, although very sparse, suggest that most of the clockwise rotation accumulated prior to late Miocene and may have occurred predominantly during the Eocene to Miocene. Therefore, most of the observed rotation in the SEMTP appears to be associated with slip along the left lateral ASRR fault. Admittedly, however, robust paleomagnetic data from Cenozoic strata from the SEMTP are still very scarce. Therefore, additional and more detailed paleomagnetic studies of Cenozoic rocks through this region are justified and will help constrain the kinematic history of the SEMTP.
 This paleomagnetic study has constrained the age of the sedimentary sequence of the Dali Basin in SEMTP to an interval from ~7.6 Ma to ~1.8 Ma, a late Miocene to early Pleistocene age. These results are the first precise age constraints for upper Cenozoic sediments in northwest Yunnan. Our main conclusions are as follows:
 The Dali Basin began to accumulate sediments at ~7.6 Ma, the late Miocene, indicating that the Dali left-lateral fault system and Erhai normal fault and the unroofing of the DCS began at this time. The youngest Sanying Formation sediments are ~1.8 Ma.
 Calculated linear sediment accumulation rates, as well as facies changes and syntectonic strata, suggest that the Dali fault system reactivated dominantly as a normal fault at ~2.5 Ma, which in turn resulted in a rapid exhumation of the DCS and infilling of the Quaternary basins at this time.
 The overall mean ChRM direction suggests that the Dali Basin experienced minor but significant (4.4 ± 2.5°) post–late Miocene rotation. Our results from Plio-Pleistocene units, in conjunction with previously published paleomagnetic data from the SEMTP, suggest that most of the clockwise rotation in SEMTP accumulated prior to late Miocene, possibly during the Eocene to Miocene and contemporaneous with the left lateral ASRR fault.
 We thank B. C. Huang, E. Wang, J. M. Sun, Q. R. Meng, and J. Liu-Zeng for the helpful discussions; Y. Q. Zhang and F. Gao for field assistance; and R. Egli for help with the IRM component analysis. S. Gilder and two anonymous reviewers are highly appreciated for their constructive comments and language improvements on an earlier manuscript, which resulted in the great improvement of this manuscript. Paleomagnetic and mineral magnetic measurements were made in the Paleomagnetism and Geochronology Laboratory (PGL), Beijing. This research was supported by the National Key Basic Research Program of China grant 2012CB821900, the Chinese Academy of Sciences grants KZCX2-YW-Q05-02 and KZCX2-EW-117, and the National Natural Science Foundation of China grants 41274073, 40925012, and 41174054.