A narrow extensional basin on the Zavkhan terrane of Mongolia exposes a >1.8-km-thick succession of basalt flows within the Teel Formation, along with rhyolites and interflow sediments. We present new U-Pb zircon ages of 446.03 ± 0.21 Ma (chemical abrasion–isotope dilution–thermal ionization mass spectrometry) on a rhyolite in the Teel Formation and 286 ± 5 Ma (laser ablation–inductively coupled plasma–mass spectrometry) on a nearby granitic intrusion (Tonkhil Complex). New paleomagnetic data yield a magnetite remanence that is likely primary, acquired during cooling of flows. The mean direction is statistically improved after tilt corrections; however, the tilt test significance is limited given the low variation in tilt between flows. We interpret a second remanence, held by hematite, as an overprint that was likely acquired later in the Paleozoic Era. The tilt-corrected magnetite direction implies a paleolatitude of ∼20°, while the hematite overprint is equatorial in both geographic and tilt-corrected coordinates. The ca. 446 Ma Teel remanence is consistent with an Ordovician paleogeographic position near Siberia; however, the hematite direction requires subsequent drift to the equator, indicating that these Mongolian terranes were not continuously connected to Siberia, which moved away from the tropics during the Paleozoic Era. This result is consistent with biogeographic constraints and a previously proposed model wherein Amuria traveled with North China during the Permian Period and collided with Siberia during the Jurassic to Triassic closure of the Mongol-Okhotsk Ocean. In this model, continental growth occurred through the collision and oroclinal buckling of a ribbon continent rather than long-lived accretion on the margin of a major craton.


The ∼5000-km-long Central Asian orogenic belt (Fig. 1) is considered the largest area of Phanerozoic crustal growth preserved in the geological record, and yet, the tectonic history and geological context of terrane assembly remain controversial and underconstrained (e.g., Khain et al., 2003; Kröner et al., 2010; Mossakovsky et al., 1994; Şengör and Natal’in, 1996; Şengör et al., 1993; Wilhem et al., 2012; Windley et al., 2007; Yakubchuk, 2008). Detailed geological mapping for vast regions of the Central Asian orogenic belt and the generation of more precise geochronological constraints and Paleozoic paleomagnetic poles are necessary to constrain the history of terrane assembly. Here, we provide new Paleozoic geological, geochronological, and paleomagnetic constraints on the Zavkhan terrane of Mongolia and compare the implied paleolatitudes to those of the major cratons that border the Central Asian orogenic belt—Siberia and North China (Fig. 1). These data provide new insight into the paleogeographic history of the Zavkhan terrane and conjoined terranes (grouped as “Amuria” in Van der Voo et al., 2015; Fig. 1).

Paleomagnetic data from the Siberia and North China cratons provide constraints on the closure of the Mongol-Okhotsk Ocean, which constituted the final stages of crustal amalgamation within the Central Asian orogenic belt (Cogné et al., 2005; Van der Voo et al., 2015). These data track the convergence of Siberia and North China through the Jurassic Period and amalgamation by the earliest Cretaceous Period (Cogné et al., 2005; Van der Voo et al., 2015). In the model of Van der Voo et al. (2015), this closure is associated with an arc-system that folded in on itself, leading to a scissor-like collision between Siberia and North China. This reconstruction reconciles the history of paleolatitudinal convergence, tomographic evidence for a slab below Siberia, and the interpretation that the Mongol-Okhotsk Ocean terminated at the Tuva-Mongolia orocline (Van der Voo et al., 2015). In the models of Cogné et al. (2005) and Van der Voo et al. (2015), Amuria is purportedly associated with North China, while terranes north of the proposed Mongol-Okhotsk Ocean suture are interpreted to have been associated with Siberia. These models also allow the Mongolian terranes to be connected along the western margin of the Mongol-Okhotsk Ocean, effectively connected with both Siberia and North China during the Mesozoic Era. But how far back in time can we connect Mongolian terranes to Siberia and/or North China?

Biogeographic constraints link many of the Mongolian terranes with Siberia or peri-Siberian terranes during the early Paleozoic Era, and some (Cocks and Torsvik, 2007) have extended that connection to the present by maintaining a peri-Siberian location for Mongolian terranes since the Ordovician Period. Paleomagnetic constraints indicate far different paleolatitudes for North China and Siberia until the Permian Period (Kravchinsky et al., 2002). Consequently, it has remained unclear how the Mongol-Okhotsk Ocean formed and where the fringing (Mongolian) terranes, such as the Zavkhan terrane, originated.

The Zavkhan terrane is a Precambrian cratonic fragment that is mantled with Neoproterozoic and early Paleozoic sedimentary successions and embedded in the Central Asian orogenic belt (Fig. 1). New geochronologic data from the Zavkhan terrane in conjunction with geologic mapping (Bold et al., 2016) lead to the terrane grouping shown in Figure 1B. The ca. 510–500 Ma outline shows the terrane association we interpret to have existed following the late Ediacaran to Early Cambrian collision of the Zavkhan and Tuva-Mongolia terranes (Tuva-Mongolia massif of Kuzmichev et al., 2001; Tuva-Mongolian microcontinent of Windley et al., 2007) with the Baidrag and Lake terranes (Fig. 1B). Note that distinct differences in the tectonostratigraphy of the Zavkhan and Tuva-Mongolia terranes in comparison to the Baidrag and Lake terranes are inconsistent with their association prior to the late Ediacaran Period (Bold et al., 2016). Tectonic and paleogeographic syntheses have previously referred to various groupings of the terranes in central and southeastern Mongolia as the Central Mongolia (e.g., Ilyin, 1990) or the Amuria block (e.g., Van der Voo et al., 2015). The outlines of Amuria used in recent reconstructions (e.g., Domeier and Torsvik, 2014) and that proposed by Van der Voo et al. (2015), who interpreted it to have collided with Siberia and peri-Siberian terranes, are shown for reference in Figure 1C. Our preferred terrane grouping significantly extends northward the boundary of terranes that we consider to have been part of Amuria. There is a strong tectonostratigraphic basis for this grouping (Bold et al., 2016). Such a terrane outline is consistent with the model of Van der Voo et al. (2015), where subduction of the Mongol-Okhotsk Ocean became doubly vergent and resulted in oroclinal bending of the terranes on the northern margin of Amuria. However, this modification reassigns some of the peri-Siberian crustal material of Van der Voo et al. (2015) (i.e., terranes assumed to be on the northern margin of the Mongol-Okhotsk Ocean) and interprets them as having a shared history with Amuria, namely, Tuva-Mongolia. Note that the Amuria outline we use within paleogeographic reconstructions (outline shown in Fig. 1) marks the current geography, but that there was Paleozoic to Mesozoic oroclinal bending within Amuria between the Tuva-Mongolia, Zavkhan, and Baidrag terranes.

Constraining the mechanisms of crustal growth associated with the Central Asian orogenic belt necessitates an understanding of the origin of crust within the orogeny, as well as constraints on when and where collision and accretionary events occurred. The origin and the tectonic history of the Tuva-Mongolia, Zavkhan, Baidrag, and Lake terranes (Fig. 1) are currently underconstrained. Of particular interest is whether or not these terranes were in close proximity to Siberia or other landmasses, such as North China, during the Paleozoic Era. Şengör et al. (1993) proposed that these Mongolian terranes had a close association with Baltica and Siberia as a Proterozoic continental arc, which separated from the cratonic margin via back-arc extension during Ediacaran to Cambrian times. In this model, the distended peri-Siberian arc, referred to as the Kipchak arc, was oroclinally buckled during the Devonian Period (Şengör et al., 1993; Şengör and Natal’in, 1996), and Amuria protruded south (present coordinates) as a peninsula through the Permian Period. A Siberian origin of the Tuva-Mongolia, Zavkhan, Baidrag, and Lake terranes was also favored by Cocks and Torsvik (2007), Khain et al. (2003), Kuzmichev et al. (2001), Yakubchuk (2008), Kovalenko (2010), and Wilhem et al. (2012). Generally, in these models, the Mongolian terranes are thought to be underlain by Siberian crust and to have occupied a Neoproterozoic to Paleozoic arc to extensional back-arc system on the margin of Siberia that closed and reopened multiple times during and between episodes of accretion.

Other studies have argued that the Tuva-Mongolia, Zavkhan, Baidrag, and Lake terranes may be Gondwana-derived fragments that accreted to Siberia—these interpretations are based on paleomagnetic data from the Neoproterozoic Zavkhan Formation (e.g., Levashova et al., 2010) and Paleozoic detrital zircon provenance data (e.g., Rojas-Agramonte et al., 2011). Paleontological data suggest that some of these Mongolian terranes had a close association with Siberia or “peri-Siberian” terranes during the Cambrian and Silurian Periods (e.g., Cocks and Torsvik, 2007); however, it remains unclear if these terranes were near Siberia during later Paleozoic time (for a more complete review of Neoproterozoic to early Paleozoic tectonic models of Mongolia, see Bold et al., 2016). Driven by paleomagnetic and tomographic data, recent studies have suggested that the Mongolian terranes grouped with Amuria (Fig. 1) remained separate from Siberia and traveled with North China until the Jurassic Period (Edel et al., 2014; Van der Voo et al., 2015). Here, we use new paleomagnetic data from the Zavkhan terrane and compile existing paleomagnetic data from the adjacent terranes to develop a refined Paleozoic paleogeographic model of Mongolian terranes now embedded within the Central Asian orogenic belt.


The Neoproterozoic volcanic and sedimentary rocks of the Zavkhan terrane were buried by ∼1.5 km of early Cambrian foreland basin deposits and then deformed and metamorphosed during subsequent Paleozoic orogenesis (Bold et al., 2016). After late Ediacaran to Ordovician accretion of arc terranes to the south of the Zavkhan terrane (Jian et al., 2014; Macdonald et al., 2009), the Ordovician to Silurian record of the Mongolian terranes is marked by sinistral transtension, extensional magmatism, and basin formation (e.g., Gibson et al., 2013; Kröner et al., 2010; Lamb and Badarch, 2001).

In our study area within the Zavkhan terrane (Fig. 2), Ordovician to Silurian transtension resulted in narrow rift basins that accommodated volcanic and minor sedimentary rocks of the Teel Formation. The Teel Formation is composed of bimodal series of rhyolite and basalt with intervals of siliciclastic sedimentary rocks (Togtokh et al., 1995). In the Khukh Davaa region, more than 1.8 km of basalt, rhyolite, and siliciclastic sedimentary rocks of the Teel Formation were erupted and deposited (Fig. 2). The stratigraphy is dominated by basaltic lava flows that texturally vary from aphyric to plagioclase porphyritic to ophitic, with variable amounts of secondary oxidation (Fig. 3). The excellent preservation of some of the basalt flows and their iron oxides, such as in the ophitic Z31 flow (Figs. 3A and 3B), suggests that alteration, where present, is associated with localized hydrothermal fluid flow rather than regional metamorphism. This oxidation results in variably present hematite, which is apparent on the macroscopic scale as Liesegang banding and on the microscopic scale as hematite staining and replacement near and within iron-oxide and iron-silicate grains (Figs. 3D–3H). Felsic eruptive centers are preserved approximately halfway through the stratigraphy (Fig. 2). Interbedded red beds were previously mapped as the Devonian Tsagaanshoroot Formation, which was paleontologically dated in the Lake terrane (Togtokh et al., 1995). New U-Pb geochronologic data from a Teel Formation rhyolite within the succession (Figs. 2 and 4) in conjunction with measured stratigraphic sections demonstrate that at least some of these sedimentary rocks on the Zavkhan terrane are associated with the Ordovician to Silurian Teel Formation.

The Zavkhan and neighboring terranes were intruded by Permian granitic plutons. On the Zavkhan terrane, these plutons are mapped as the Tonkhil Complex, but U-Pb zircon geochronology has not been previously presented in support of this age assignment. The Tonkhil Complex granites are alkaline and characterized by coarse crystalline syenite-porphyry. The widespread nature of these plutons across the Precambrian cratonic fragments (e.g., Byamba, 2009; Jahn et al., 2009; Kröner et al., 2010; Yarmolyuk et al., 1999; Zacek et al., 2016) supports the interpretation that the regions on both sides of the Mongol-Okhotsk Ocean originally formed a continuous arc that was later oroclinally buckled. The proximity of some of these intrusions to the studied Teel Formation sections (Fig. 2) has the potential to have influenced the paleomagnetic remanence of the basalts and may have resulted in oxidation in some of the samples that can be seen petrographically (Fig. 3).


A sample of Teel Formation rhyolite (U1105) was collected in the Khukh Davaa region in stratigraphic section U1308 (Fig. 2; 47.0387°N, 95.4504°E). A granite from the Permian Tonkhil Complex (U1127-5) was sampled in the same region at the opening of the Salaa Gorge (Fig. 2; 46.8786°N, 95.7105°E). Zircon grains from the samples were first analyzed by laser-ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS), and grains of sample U1105 were analyzed by chemical abrasion–isotope dilution–thermal ionization mass spectrometry (CA-ID-TIMS) at Boise State University.


Zircon grains were separated following the protocol outlined in Bold et al. (2016) and Macdonald et al. (2014). The grains were analyzed by LA-ICP-MS using a ThermoElectron X-Series II quadrupole ICPMS and New Wave Research UP-213 Nd:YAG ultraviolet (213 nm) laser-ablation system. In-house analytical protocols, standard materials, and data reduction software were used for acquisition and calibration of U-Pb dates and a suite of high field strength elements (HFSE) and rare earth elements (REE). A weighted-mean date was calculated using Isoplot 4.15 (Ludwig, 2008) from errors on individual dates that do not include the standard calibration uncertainties. However, the error on the date includes the standard calibration uncertainty within the experiment and is given at 2σ.

Zircon grains from sample U1105 were removed from epoxy mounts after LA-ICP-MS and subjected to a modified version of the CA-ID-TIMS method of Mattinson (2005), with analyses conducted on single grains. Details of the analytical techniques are described in Bold et al. (2016). All common Pb in analyses was attributed to laboratory blanks and subtracted based on the measured laboratory Pb isotopic composition and associated uncertainty. The weighted-mean 206Pb/238U date was calculated from equivalent dates and plotted using Isoplot 4.15 (Ludwig, 2008). The error is given at 2σ and reported as ± x/y/z, where x is the internal error based on analytical uncertainties only, including counting statistics, subtraction of tracer solution, and blank and initial common Pb subtraction; y includes systematic uncertainty associated with tracer calibration; and z additionally includes systematic uncertainty associated with the 238U decay constant. When comparing these dates with those from other U-Pb laboratories not using the EARTHTIME tracer, ± y should be used. Comparisons with other chronometers should utilize ± z.


Five of the six zircon grains from the Teel Formation rhyolite (U1105) yielded equivalent CA-ID-TIMS 206Pb/238U dates with a weighted mean of 446.03 ± 0.21/0.30/0.55 Ma (mean square of weighted deviates [MSWD] = 0.8, probability of fit = 0.52; Table 1; Fig. 4). This date is the interpreted magmatic age. One date is slightly older and interpreted to reflect the presence of an inherited component.

Thirty-nine of the 42 zircon grains from the granite pluton (U1127-5) yielded equivalent LA-ICP-MS 206Pb/238U dates with a weighted mean of 286 ± 5 Ma (MSWD = 1.2, probability of fit = 0.15; Table 2; Fig. 4). This date is interpreted as the magmatic age.


During a field season in 2014, we sampled 28 lava flows from the Teel Formation with the goal of developing a paleomagnetic pole to constrain the position of the Zavkhan terrane in the Late Ordovician Epoch. All samples were collected within the context of volcanostratigraphic sections (Fig. 2). Cores were obtained with a gas-powered drill and oriented using a Pomeroy orienting device. Whenever possible, sun compass measurements were taken in addition to magnetic orientations and were used preferentially to constrain sample orientation. Each individual lava flow sampled for paleomagnetism is considered a site, with at least eight samples (cores) collected from each site (flow); one specimen was measured from each sample.

Paleomagnetic samples were analyzed at the UC Berkeley Paleomagnetism Laboratory using a 2G Enterprises DC-SQUID superconducting rock magnetometer, which is within a magnetostatic shield with fields ≤500 nT (Scott and Frohlich, 1985). The quartz glass sample holder rod used to bring samples into the magnetometer typically has a magnetic moment that is less than 5 × 10–12 Am2. After measuring the natural remanent magnetization (NRM), samples were immersed in liquid nitrogen while in a low-field environment (<10 nT) in order to preferentially demagnetize remanence associated with multidomain magnetite by cycling through the Verwey transition and the isotropic point (Verwey, 1939; Feinberg et al., 2015). Samples were then thermally demagnetized through step heating in a magnetically shielded ASC thermal demagnetizing furnace. Following acquisition of thermal demagnetization data, the PmagPy software package (https://github.com/PmagPy; Tauxe et al., 2016) was used for principal component analysis of magnetization directions (Kirschvink, 1980). The measurement level data, as well as sample level interpretations, are available in a Github repository associated with this study (https://github.com/Swanson-Hysell-Group/2016_Teel_Basalts) and as a contribution within the MagIC database (http://earthref.org/MAGIC/doi/10.1130/L552.1). Details regarding site level interpretations are also available in the GSA Data Repository1. Examples of specimen demagnetization results are shown in Figure 5. Site means are summarized in Figure 6 and Table 3, with mean calculated poles reported in Table 4.

Three distinct directions were revealed through thermal demagnetization (Fig. 5):

  • (1) Low-temperature cycling and thermal demagnetization steps to ∼150 °C, and sometimes continuing to 300 °C, removed a magnetization for which the direction in geographic coordinates (i.e., no correction for bedding tilt) corresponds to the present local geomagnetic field (Fig. 5). A bootstrap tilt test (Tauxe and Watson, 1994) on these directions revealed the vectors to be significantly better clustered in geographic versus tilt-corrected coordinates. These data therefore fail a tilt test, as expected for a viscous magnetic overprint acquired in recent times (see GSA Data Repository).

  • (2) Starting as low as 300 °C, and continuing up to the Curie temperature of magnetite (∼580 °C), a component was removed in most sites that plots in the upper hemisphere of the southwest quadrant in tilt-corrected coordinates (Figs. 5 and 6). Given the unblocking temperatures over which this component was removed, we interpret it to have been held by (titano)magnetite. Through the stratigraphic succession, there is a progressive decrease in the dip of units (from 58° to 24°). Bedding measurements on thin interflow units of sandstone and siltstone between basalt flows were used for tilt corrections. This decrease in bedding tilt was likely associated with syneruptive tilting during basin development and enables a tilt test to be conducted on the directions from different flows. These directions are better grouped at high levels of untilting than in geographic coordinates, suggesting a pretilting acquisition of remanence (Fig. 6). In the bootstrap tilt test, the best concentration of the data occurs at intermediate levels of untilting, with 95% of the pseudosample maxima lying between 50% and 95% unfolding, and 99% of the pseudosample maxima lying between 39% and 102% unfolding. The small variation in dip between flows throughout the section and the relatively small sample size (N = 23) limit the robustness of the tilt test, and the few flows with directions furthest away from the mean have an oversized impact on the results. Overall, we consider synfolding magnetization unlikely and interpret the distinct improvement in precision upon tilt correction to indicate that the magnetite-held direction is a pretilting primary thermoremanent magnetization (TRM) acquired at the time of eruption and cooling (Fig. 5).

  • (3) For some sites, there was no remaining remanence after removal of the magnetite component (e.g., Figs. 5A and B). However, in many lava flows, samples had remanence that unblocked through temperatures characteristic of hematite (∼610–680 °C; Figs. 5C–5F). The magnetization directions associated with this remanence are well grouped as low-inclination south-directed vectors (Fig. 6). In five flows, this same direction was also removed at lower unblocking temperatures within the magnetite range (e.g., Fig. 5F). All hematite directions were near the strike orientation of the bedding and as a result were minimally affected by the applied tilt correction (Fig. 6). As a result, the tilt test conducted on the population of fits made to hematite-held remanences is ambiguous, with no dominant population of pseudosample maxima, and consequently the test is statistically insignificant (Fig. 6). These results allow for remanence acquisition before or after folding, although the precision of the mean direction is higher in geographic coordinates. Given the distinct direction of the hematite remanence compared to that of magnetite, and also the fact that this southerly and horizontal remanence is dominantly held by hematite, we consider this component to be secondary and acquired as a chemical remanent magnetization (CRM) sometime after the eruption of the succession and most likely related to the emplacement of the 286 ± 5 Ma Tonkhil Complex. However, given the inconclusive tilt test and lack of a direct geochronologic link, the hematite remanence may have been acquired in middle to late Paleozoic times.

Petrography reveals that samples dominated by magnetite remanence are very well preserved, whereas samples with appreciable hematite remanence have undergone variable degrees of oxidative alteration. Figure 3 shows petrographic photomicrographs for samples for which paleomagnetic data are shown in Figure 5. For example, sample Z31 has a strong magnetite remanence and no indication of hematite remanence (Fig. 5A), and in thin section, the plagioclase, pyroxene, and Fe-Ti oxides are all well preserved. In contrast, sample Z53-4 has a remanence that is dominated by magnetite with some hematite (Fig. 5C), and there is evidence of secondary hematite formation at outcrop and microscopic scale (Figs. 3D, 3E, and 3F). The petrography is consistent with the interpretation that the magnetite remanence is a TRM, and the hematite remanence is a subsequently acquired CRM.

The demagnetization behavior and resulting remanences through magnetite and hematite unblocking temperatures from six sites (three in the northern outcrop and three in the southern) were different from all other sites. Flows Z44, Z45, Z50, Z56, and Z57 all contained one remanence direction oriented to the south and removed through unblocking temperatures characteristic of both magnetite and hematite (Fig. 5F); the hematite remanence typically dominated. If we consider the hematite as secondary, then remanence removed at lower demagnetization temperatures in these flows with the same direction likely resulted from alteration during the fluid flow that produced the hematite. Vertical-axis rotations cannot account for this variation, given that flows immediately above and below contain magnetite remanences that point southeast. Flows Z44, Z45, and Z50 were likely more susceptible to alteration given their mineralogy, and we consider them to be remagnetized. Sites Z56, Z57, and Z58 are from a small section (N1404; Fig. 2) of flows ∼7 km to the ESE of all other sites (N1402 and N1403; Fig. 2) with no clear correlation between the outcrop panels. The tilt correction used for this outcrop was measured from flow banding rather than interbedded sedimentary units, making the tilt-corrected data for Z56, Z57, and Z58 unreliable. These flows are also much closer to late Paleozoic intrusions, including the Tonkhil Complex (Fig. 2). Therefore the hematite remanence, which we interpret as secondary, is the only constraint that can be taken from these basalt flows.


Age of Paleomagnetic Directions

In the results, we interpreted the magnetite-held remanence direction as being acquired as a thermal remanence when the basalt flows erupted on the basis of the increase in precision upon tilt correction and the excellent preservation of flows in which the remanence was dominated by magnetite (with primary igneous textures including magnetite grains; Figs. 3A and 3I). This interpretation places the age of the tilt-corrected magnetite pole (Table 4) as being the age of the Teel Basalts (ca. 446.03 ± 0.21 Ma). The tilt-corrected mean of the magnetite remanence is unique when compared to available Paleozoic paleomagnetic data from the nearby Mongolian terranes. The magnetite mean is similar to the mid-temperature overprint direction found in Ediacaran to Cambrian rocks from the Zavkhan terrane, both in geographic (in situ) coordinates (Kravchinsky et al., 2001); however, the precision of the Kravchinsky et al. (2001) data increases significantly after tilt correction, suggesting that the geographic direction and its similarity to the Teel magnetite mean is coincidental. Cretaceous results (of reversed polarity) from Mongolia (van Hinsbergen et al., 2008) are also similar in direction to the Teel magnetite geographic mean; however, a Cretaceous remagnetization event that fully overprinted magnetite while leaving the hematite untouched at a distinct direction seems improbable given the primary igneous texture of the magnetite grains with secondary growth of hematite.

The statistically insignificant tilt test for the hematite mean direction makes it difficult to constrain the age of the remanence as post- or prefolding, enlarging the possible range of acquisition ages. However, because the hematite mean direction is identical in both geographic and tilt-corrected coordinates (see common mean bootstrap test in the Data Repository), we are confident that it accurately portrays an equatorial position of the Zavkhan terrane when the overprint was acquired. In the region, there is evidence for Silurian granites, Permian granites, and deformation that postdate the Silurian granites and predate the Permian ones (Bold et al., 2016). Each of these events could have been associated with the observed secondary hematite. The hematite remanence is similar in direction to existing late Paleozoic paleomagnetic data from Mongolia, including the “B component” of Edel et al. (2014) from the Gobi-Altai terrane, which was likely associated with the Tuva-Mongolia, Zavkhan, Baidrag, and Lake terranes since Cambrian times (within red dotted outline in Fig. 1B). Edel et al. (2014) incorporated data from Didenko (1992) that are from the Trans-Altai and South Gobi zones (southern terranes), which were sutured to the Central Mongolian terranes by ca. 280 Ma. Although many lack high-precision radiometric ages or robust field tests, other Paleozoic poles that postdate the Ordovician Period across the terranes imply low paleolatitudes similar to that of the Teel hematite remanence (Fig. 7). Some of these paleomagnetic directions, such as equivalent directions reported by Grishin et al. (1991) and Pechersky and Didenko (1995) from Devonian rocks, may have been reset during late Paleozoic or early Mesozoic metamorphism in the southern terranes, as indicated by recent data from the region (Edel et al., 2014; Lehmann et al., 2010; plot included in Data Repository). Taken together, we interpret the hematite remanence, and other constraints, to indicate a low-latitude position for Mongolian terranes sometime following the Ordovician Period and prior to the Mesozoic Era, when they were in mid- to high-latitude positions (Fig. 7).

Paleontological Considerations

Previous tectonic reviews have cited paleontological data to support an affiliation between the “peri-Siberian” terranes and Siberia throughout the Paleozoic Era (e.g., Cocks and Torsvik, 2007; Wilhem et al., 2012). In contrast, the paleogeographic model described in the next section, which seeks to satisfy the paleomagnetic constraints, predicts that the Zavkhan terrane was situated near Siberia from Cambrian Series 3 to the Late Ordovician Epoch, departed in the Silurian Period, and then returned in the Mesozoic Era. The Paleozoic associations that previous reviews have pointed to are from Cambrian trilobites (e.g., Atashkin et al., 1995; Korobov, 1980, 1989) and Ordovician to Silurian brachiopods (Wang et al., 2011). Although no trilobites have been described from the Zavkhan terrane, early Cambrian trilobites from the associated Tuva-Mongolia terrane are distinct from Siberian trilobites and from those of “peri-Siberian” terranes to the northwest, such as the Altai-Sayan and Gorny-Altai terranes (Atashkin et al., 1995; Korobov, 1980, 1989). The presence of Ordovician to Silurian corals and the low-diversity Silurian Tuvaella brachiopods in Mongolia are consistent with a late Cambrian to Silurian connection with Siberia (Ulitina et al., 2009); however, although the distinctive Tuvaella brachiopod is found throughout southern and northeastern Mongolia and terranes to the northeast, it is not present on the Siberian craton sensu stricto (Wang et al., 2011). After the Ordovician Period, brachiopod assemblages from the Central Mongolian terranes and Siberia diverge (Harper et al., 2013), with Lower and Middle Devonian brachiopods in the Mongolian terranes that constitute Amuria differing from those occurring in Siberia, North China, and South China (Alekseeva et al., 2001; Hou and Boucot, 1990; Boucot and Blodgett, 2001). Webster and Ariunchimeg (2004) described Devonian crinoids in southern Mongolia to be most similar to European and North American fauna. These data, taken together with the geological (Bold et al., 2016) and paleomagnetic data, are consistent with a model in which Late Ordovician to Silurian rifting led to the departure of the Zavkhan terrane (and Lake, Tuva-Mongolia, and Baidrag terranes) from Siberia.

Paleogeographic Model

The contrasting paleogeographic histories of Siberia and North China throughout the majority of the Paleozoic Era enable us to evaluate the paleogeographic position of Mongolian terranes relative to these cratons. Figure 7 shows a compilation of paleomagnetic data for Siberia (Cocks and Torsvik, 2007; Powerman et al., 2013; Kravchinsky et al., 2002; Kamysheva, 1971, 1975; Zhitkov et al., 1994; Davydov and Kravchinsky, 1973), North China (Van der Voo et al., 2015; Huang et al., 1999, 2001; Embleton et al., 1996; Doh and Piper, 1994), and the terranes associated with the Zavkhan terrane that coalesced to form Amuria (additional details and references can be found in the Data Repository). Paleomagnetic data coverage for both Siberia and North China is good throughout most of the Paleozoic and Mesozoic Eras, although there are some portions, such as Silurian and Carboniferous times, where gaps arise (Fig. 7). One approach that has been taken to determine the age of poles from Mongolian terranes that are presumed to be peri-Siberian is to date them through comparison to Siberia’s apparent polar wander path. This method contrasts with our overarching goal of deducing the cratonic neighbor(s) of the Mongolian terranes during the Paleozoic Era without a built-in assumption about the connection. In the compilation in Figure 7, we seek to assign ages to poles based on paleontological or radioisotopic constraints.

Our preferred model (Fig. 8), animated using GPlates (animation included in the Data Repository), was built primarily using the rotation parameters provided by Domeier and Torsvik (2014) and Domeier (2016) for the 420–250 Ma and 500–420 Ma intervals, respectively. The reconstructions from 250 Ma to the present are based on Euler poles given in Torsvik et al. (2012). The majority of large blocks are positioned according to Torsvik et al. (2012, 2014), with the location and paleolatitude of South China and Annamia (Indochina) approximated using data from Cocks and Torsvik (2013) and modified in consideration of additional paleomagnetic data and studies that indicate a close association with northwest Australia and terranes that originated from it (Han et al., 2015; Jing et al., 2015; Zhang et al., 2015). The position of Tarim in the reconstruction is constrained by poles compiled in Zhao et al. (2014), but it should be considered preliminary given the paucity of data. Additional data from Siberia, North China, and the Mongolian terranes that form Amuria were used to create Figure 7 (detailed in the Data Repository) and were used to constrain paleolatitudes in Figure 8. Given the eventual suturing between Siberia and North China with the closure of the Mongol-Okhotsk Ocean (Van der Voo et al., 2015), we considered the possibility that the Zavkhan terrane was associated with North China before suturing, perhaps as early as the Devonian Period. The shape of the Zavkhan and associated terranes used throughout the reconstructions is that of the present day (Fig. 8). This approach is a gross simplification of the history of the terranes, which have undergone oroclinal bending. The history of oroclinal bending and its reconstruction necessitate further work in this portion of the Central Asian orogenic belt. Our approach, at present, is to keep the shape as is and seek to satisfy the paleolatitudinal constraints.

The paleolatitude of the Zavkhan terrane at ca. 446 Ma was 19 ± 5 °N, as constrained by the tilt-corrected magnetite component of the Teel Formation basalts, and is consistent with it being associated with the Siberia craton in the Late Ordovician Epoch (Fig. 7). Ideally, the consistent polarity of the Teel Formation magnetite remanence could be used to provide a hemisphere constraint. However, the global polarity time scale is poorly constrained for Late Ordovician rocks (Pavlov and Gallet, 2005), allowing for the Teel Formation magnetite remanence (TRM) to have been acquired during a period of either normal or reversed polarity. If the geomagnetic field was of normal polarity (as was interpreted by Pavlov and Gallet [2005] on the basis of data from Khramov et al. [1965]), the Zavkhan terrane must have been in the Southern Hemisphere. A Southern Hemisphere position is inconsistent with early Paleozoic paleontological ties with the present-day southern margin of Siberia (see Paleontological Considerations sections) and could point to an association with Gondwana. However, there is only one study that generated normal polarity data from the Late Ordovician Epoch (Khramov et al., 1965), and it is unclear if these rocks are all older than the Hirnantian Stage. Therefore, we do not consider the global polarity time scale for the interval, as it currently stands, to be a useful tool in determining the polarity of the Teel Formation magnetite remanence.

If the geomagnetic field was of reversed polarity at the time of eruption, the Zavkhan terrane could have been in the Northern Hemisphere and near the present-day southern Siberia margin (Fig. 8A). A similar geologic history potentially links the Zavkhan terrane with the southern Siberian margin from ca. 510 to 450 Ma, when the southwestern margin of the Zavkhan and Lake terranes may have accreted to southern Siberia (Bold et al., 2016). Beginning in the Ordovician Period, both southern Siberia and the Zavkhan and Lake terranes hosted rift-related extensional magmatism (Dobretsov and Bulsov, 2007; Yarmolyuk, 2011; Cocks and Torsvik, 2007; Bold et al., 2016), including the volcanics of the Teel Formation.

Due to the low paleolatitude implied by the hematite overprint on the Teel Formation, as well as other compiled data from the associated terranes (Fig. 7), the Zavkhan terrane appears to have not traveled with Siberia through the entirety of the Paleozoic Era. The new paleomagnetic data require a tropical position for the Zavkhan terrane when the chemical remanence (hematite) was acquired, which was likely during the middle to late Paleozoic Era. A low-latitude position is indicated by the data both when they are corrected for bedding tilt and when they are not. In contrast, Siberia appears to have remained in the Northern Hemisphere and traveled to progressively higher latitudes from 440 to 200 Ma, as substantiated by Siberia’s relatively well-constrained apparent polar wander path from the Ordovician to Devonian Period (Fig. 7; Cocks and Torsvik, 2007; Torsvik et al., 2012). A continued connection of the Zavkhan terrane with Siberia would have resulted in a continuous position in the Northern Hemisphere with drift to progressively higher latitudes (Fig. 7). Due to this constraint, we prefer a model wherein the Zavkhan terrane rifted off of a landmass, likely Siberia, during or soon after ca. 446 Ma eruption of the Teel Formation basalts in extensional (transtensional) basins. This model is similar to the one in Domeier and Torsvik (2014) showing Amuria and Siberia at similar latitudes at 410 Ma. However, their model maintains a high-latitude position for Amuria throughout the Paleozoic (often times higher than Siberia, e.g., ∼50°N at 370 and 330 Ma), which conflicts with the Teel hematite paleolatitudinal constraint and some of the Paleozoic paleomagnetic data from other Mongolian terranes (Fig. 7). Even models that extend a linear arc of Mongolian terranes to the south of Siberian cannot accommodate for the large difference in implied paleolatitude.

The linear distribution of Permian peralkaline and alkali-feldspar granitoids that crosscut terrane boundaries in both Mongolia and Siberia has been argued to be a plume-related tie point between the Zavkhan terrane and Siberia (Jahn et al., 2009); however, the wide distribution of pluton ages and the more sparse, less linear distribution of intrusions in Mongolia allow other possible explanations. Given the different paleolatitudes of Siberia and Mongolia during the Permian Period, we believe it is more likely the granitoids formed along a semicontinuous subduction zone on the margin of the Mongol-Okhotsk Ocean before its Mesozoic closure, as has been proposed in tectonic models (Zhao et al., 1990, 2013; Edel et al. 2014; Kravchinsky et al., 2002). A consistent scenario is that the Mongol-Okhotsk Ocean likely closed like a pair of scissors, as proposed by Edel et al. (2014) and Van der Voo et al. (2015). In our version of this model (Fig. 8), the Tuva-Mongolia and northern Lake terranes were contiguous with the Zavkhan and Baidrag terranes, forming the composite Amuria ribbon continent (Bold et al., 2016). We propose that the northern margin of the ribbon continent collided with Siberia during the late Paleozoic Era, as suggested by metamorphic dates along the Charysh-Terekta-Ulagan-Sayan suture-shear zone and Main Sayan fault (Glorie et al., 2011; Buslov, 2011). The long axis of the Amuria ribbon continent collided roughly orthogonal to the Siberian margin, driving a secondary orocline (Johnston et al., 2013) to develop between the Khubsugul and Zavkhan terranes. This orocline drove the closure of the Mongol-Okhotsk Ocean, with the southern arm of the orocline rotating ∼90° relative to Siberia and with the margin and alkali granites folding in on themselves. Constraints for the onset of oroclinal bending and Mongol-Okhotsk Ocean closure may be provided by a 325.4 ± 1.1 Ma date on suprasubduction ophiolites (Tomurtogoo et al., 2005) that presumably formed during subduction initiation. Final closure of the Mongol-Okhotsk Ocean occurred in the Jurassic Period (Van der Voo et al., 2015). This model further implies that the Central Asian orogenic belt did not build exclusively as a long-lived accretionary margin, but instead was greatly modified through the collision of a ribbon continent that was brought into place and oroclinally buckled during a continent-continent collision involving Siberia and North China.


Paleomagnetic data from basalts of the Teel Formation of the Zavkhan terrane provide new constraints on the paleogeographic position of a large composite terrane that comprises much of Mongolia and is a significant component of the Central Asian orogenic belt. The geological and geochronological constraints on this composite terrane significantly increase the northward extent of the Amuria microcontinent. While a remanence interpreted to be primary from the Teel Formation basalt flows is consistent with a peri-Siberian position of the block in the Late Ordovician Epoch, such a position later in the Paleozoic Era is not consistent with the results. Therefore, we favor a tectonic and paleogeographic model wherein the Zavkhan, Baidrag, Tuva-Mongolia, and Lake terranes were separate from Siberia and traveled to the equator subsequently in the Paleozoic Era as the Amuria block. The Mongol-Okhotsk Ocean closed during the Mesozoic Era. During closure, this composite terrane was sandwiched between the Siberia and North China cratons and oroclinally buckled into its present form, which comprises much of present-day Mongolia.

This research was funded by a UC Berkeley Esper S. Larsen Research Grant, a Hellman Fellowship, and National Science Foundation Grant EAR-1547434 awarded to Swanson-Hysell. Macdonald thanks the National Aeronautics and Space Administration Astrobiology Institute at Massachusetts Institute of Technology node for support. We thank M. Domeier, D. Pastor-Galán, and A. Weil for reviews that greatly improved this manuscript. Odbayar Erdenebat and Ariunsanaa Dorj assisted with field work, and Gunnar Speth assisted with paleomagnetic sample preparation.

1GSA Data Repository Item 2016284, a detailed analysis of Teel paleomagnetic data and discussion of the paleomagnetic pole compilation, is available at www.geosociety.org/pubs/ft2016.htm, or on request from editing@geosociety.org.
Taylor Kilian