Insights into the early Tibetan Plateau from (U–Th)/He thermochronology

The central Songpan–Ganzi belt, located on the eastern margin of the Tibetan Plateau, has a similar high elevation and low relief to parts of central Tibet. Thermochronological studies from the central Tibetan Plateau reveal a history of slow exhumation (rates <0.05kmMa−1) since 45Ma; however, the exhumation history of the central Songpan–Ganzi belt is unknown. To address this, we conducted an apatite and zircon (U–Th)/He thermochronology study of bedrock samples collected across the central Songpan–Ganzi belt and into central parts of the Tibetan Plateau. Zircon (U–Th)/He ages range from 54.2±7.5 to 146.5±10.0Ma and the majority of apatite (U–Th)/He ages fall between 74.7±19.0 and 35.7±9.4Ma. Thermal history models of these data show rapid cooling in the late Mesozoic and much slower cooling diagnostic of low rates of erosion throughout most of the Cenozoic. The late Mesozoic rapid cooling is consistent with the existence of significant topography and relief at least in some parts of the Songpan–Ganzi belt at that time. We find no evidence for a regional Miocene acceleration in erosion, although three samples from the headwaters of the Salween and Mekong rivers gave younger AHe ages between 15 and 16Ma that reflect an acceleration in river incision. Supplementary material: Laser ablation–inductively coupled plasma mass spectrometry zircon U–Pb data from central and eastern Tibet are available at www.geolsoc.org.uk/SUP18647.

Knowledge about the growth history of the Tibetan Plateau, the largest region of highest elevation on Earth ( Fig. 1; Fielding et al. 1994;Zhu et al. 2013) is critical to understanding the geodynamics of plateau growth and the consequent impacts on regional deformation, and is central to the debate about the role of the Himalayan-Tibetan orogen in the evolution of Cenozoic climate through association with high rates of chemical weathering and drawdown of atmospheric CO 2 (Raymo et al. 1988;Raymo & Ruddiman 1992). Surface uplift and expansion of the eastern margins of the plateau has also been linked to intensification of the East Asian monsoon (e.g. Clift et al. 2008).
To understand the elevation growth history of Tibet researchers have applied a range of geological and geochemical proxies such as the initiation of east-west extension (Coleman & Hodges 1995;Blisniuk et al. 2001) and the timing of potassium-rich volcanism to date surface uplift (Turner et al. 1993;Chung et al. 1998;Guo et al. 2006), as extension and magmatism arguably result from convective thinning of the lithosphere, which can increase the surface elevation (Molnar et al. 1993). However, these interpretations are controversial because slab break-off of the oceanic portion of the Indian plate can also explain the generation of potassium-rich lava (e.g. Kohn & Parkinson 2002), and localization of the stress field from the India-Asia collision across southern Tibet can result in north-south-oriented normal faulting (e.g. Kapp & Guynn 2004).
More direct indicators of elevation change based on palaeoaltimetry studies using oxygen isotopes have indicated that the central Tibetan Plateau reached elevations in excess of 4 km by 35 ± 5 Ma and found no evidence to support plateau-wide uplift in the late Miocene (Rowley & Currie 2006;DeCelles et al. 2007). In contrast, apatite thermochronometry studies of the marginal areas of the northern and eastern Tibetan Plateau show evidence for accelerated erosion in the middle Miocene, considered to be the response to expansion of the topography owing to changes in the style of deformation (Clark et al. 2005Zheng et al. 2010;Lease et al. 2011;Wang et al. 2012;Dai et al. 2013). A recent study (Duvall et al. 2012) based on Bayesian statistical modelling of detrital apatite thermochronological data from modern river sand samples collected from eight catchments that drain the interior of eastern Tibetan Plateau found evidence for a widespread acceleration in river incision and erosion between 11 and 4 Ma following a long period of slow erosion. It was suggested that this was due to regional-scale uplift in concert with eastern expansion of the plateau. Our work aims to build on these previous studies by applying apatite thermochronometry to bedrock samples collected from the central Songpan-Ganzi belt, an area that has so far received little attention.
To understand the relationship between the central Songpan-Ganzi belt and the central Tibetan Plateau we collected bedrock samples for (U-Th)/He thermochronometry as a transect across the Songpan-Ganzi belt extending SW into the Qiantang terrane and Lhasa block (Fig. 2). Both apatite and zircon (U-Th)/He analyses were performed on the majority of samples, with supplementary zircon U-Pb geochronology used to define sample formation ages where unknown. The objectives of these analyses were to (1) define the exhumation history of the central and eastern Tibetan Plateau, (2) test if these two areas shared a common erosion history, and (3) compare our new results with published exhumation studies for other parts of the Tibetan Plateau. J. DAI ET AL. 918

Geological setting
In this paper, the central Tibetan Plateau refers to the Lhasa and Qiangtang blocks, whereas the eastern Tibetan Plateau is composed of the Songpan-Ganzi and the Longmen Shan. The Qiangtang block is bounded by the Jinsha suture zone to the north and the Bangong-nujiang suture zone to the south (Fig. 1). The closure of the Jinsha suture zone between the Songpan-Ganzi and the Qiangtang occurred probably in the Late Triassic-Early Jurassic, whereas the closure along the Bangong-nujiang suture zone is considered to have occurred in the Late Jurassic-Cretaceous (yin & Harrison 2000). The Lhasa block lies between the Bangongnujiang suture zone and the yarlung Zangbo suture zone. The geological characteristics of the northern and southern parts of the Lhasa block are distinctly different. The northern part is largely composed of strongly deformed pre-Cenozoic strata (Leier et al. 2007), whereas the southern part is characterized by the Gangdese batholith (Wen et al. 2008;Chu et al. 2011) and Cretaceous to Tertiary nonmarine volcanic rocks of the Linzizong Group (Mo et al. 2008;Lee et al. 2009).
The Songpan-Ganzi is a triangular tectonic domain between the East Kunlun in the north, the Qiangtang in the south, and the South China block in the east, occupying a large portion of the eastern Tibetan Plateau (Figs 1 and 2; yin & Harrison 2000; Zhang & Santosh 2012, and references therein). The region is characterized by thick Triassic flysch, which covers >200000 km 2 and is 10-15 km thick (nie et al. 1994;Weislogel et al. 2006). The Songpan-Ganzi flysch was intensely deformed by folding and thrusting during the Late Triassic and Early Jurassic (Burchfiel et al. 1995;Worley & Wilson 1996;yin & Harrison 2000;yan et al. 2011). Except for localized Tertiary and Quaternary outcrops, post-Triassic sedimentary rocks are largely absent (Pan et al. 2004).

Apatite and zircon (U-Th)/He analyses
Samples were mechanically crushed and mineral grains were concentrated by using standard heavy liquid and magnetic separation techniques. Apatite and zircon grains were hand-picked under a microscope and photographed. Only optically inclusion-free, euhedral, unfractured apatite and zircon grains were selected (Farley 2002;Ehlers & Farley 2003;. Selected grains were digitally measured and geometric parameters were used to calculate α-ejection corrections. Both apatite and zircon grains were encapsulated in niobium tubes. Helium concentration was measured by 3 He isotope dilution using a Balzers QMS 200 quadrupole mass spectrometer at the University of California, Santa Cruz (UCSC) helium thermochronology laboratory. After He measurements, apatite and zircon grains were spiked using a mixed 229 Th-233 U tracer. Apatites were dissolved in concentrated HnO 3 whereas zircons were dissolved in HF and HnO 3 for isotope dilution inductively coupled plasma-mass spectrometry (ICP-MS) analysis of U and Th. Concentrations of U and Th were determined by ICP-MS using a Thermo Scientific X-series II quadrupole system at UCSC. The analytical details are the same as those presented by Dai et al. (2013). Apatite and zircon ages were corrected for α-ejection using standard procedures of Farley et al. (1996) and Hourigan et al. (2005), respectively. Weighted mean ages were calculated by using Isoplot 3.7 (Ludwig 2008). Six fragments of Durango fluorapatite standard were analysed along with our samples, yielding consistent ages with an average value of 32.7 ± 1.6 Ma. Six  Duvall et al. (2012) and AHe and AFT ages (Clark et al. 2005;Reid et al. 2005;Wang et al. 2008a;Ouimet et al. 2010;Hetzel et al. 2011;Wilson & Fowler 2011;Rohrmann et al. 2012). The blue stars represent the locations discussed in the text.
EARLy TIBETAn PLATEAU 919 replicate analyses of single zircon grains from the Fish Canyon tuff yielded a weighted mean age of 29.0 ± 0.6 Ma.

Zircon U-Pb analyses
Zircon U-Pb dating of seven representative samples (D27, D28, D30, D32, D36, D56, D03) was carried out by laser ablation (LA)-ICP-MS at the Key Laboratory of Continental Collision and Plateau Uplift, Institute of Tibetan Plateau Research, Chinese Academy of Sciences. Laser sampling was performed using a newWave and ATL 193 nm ArF excimer LA system (UP193FX) coupled to an Agilent 7500a ICP-MS system. The zircon Plesovice standard (thermal ionization mass spectrometry reference age 337.13 ± 0.37 Ma; Sláma et al. 2008) were used to correct for instrumental mass bias and depth-dependent inter-element fractionation of Pb, Th and U. The laser ablation spot diameter was 35 μm. Off-line isotope ratios and trace element concentrations were calculated relative to nIST 612 (Pearce et al. 1997) using GLITTER 4.0. Common Pb correction and ages of the samples were calibrated and calculated using ComPbCorr#3.15 (Andersen 2002). Probability density plots and weighted mean calculations of U-Pb ages were made using Isoplot 3.7 (Ludwig 2008).
Both apatite and zircon (U-Th)/He datasets show significant overdispersion amongst the single-grain ages, which is not uncommon for (U-Th)/He datasets. For the AHe ages, only sample D56 shows some positive correlation between AHe age and the effective uranium concentration, which serves as an indirect proxy for radiation damage control on He diffusion (Shuster et al. 2006;Flowers et al. 2009). The  eU, effective U = U + 0.235Th. Ft, α-ejection correction after Farley et al. (1996). Weighted mean calculations of AHe ages were carried out using Isoplot 3.7 (Ludwig 2008). The errors for the weighted mean ages are 95% confidence errors. Longitude and latitude are GPS positions. The computation of α-ejection factor FT for the correction of the ages is made by calculating the surface/volume ratio of the parallelepiped-shaped grains and using the factors proposed by Hourigan et al. (2005). FCT, standard zircon from Fish Canyon tuff. Lithology and GPS location of sample D26: granite; 32°17'25"n, 91°35'10"E. Lithology and GPS locations of other samples are given in Table 1. Weighted mean calculations of AHe ages were carried out using Isoplot 3.7 (Ludwig 2008). The errors for the weighted mean ages are 95% confidence errors.
absence of any correlation for the other samples suggests that radiation damage does not play a significant role in the variation of singlegrain ages (Flowers et al. 2009). The bulk of the overdispersion is therefore attributed to intragrain zonation, especially for the ZHe ages, and the effects of small inclusions, high-lanthanide 'neighbours' and grain boundary phases in the case of apatite. These together may contribute to inaccurate α-ejection correction (Hourigan et al. 2005;Farley et al. 2011;Johnstone et al. 2013).

Interpretation
The multiple chronometric systems employed in this study allowed a basic assessment of sample cooling histories. We employed nominal closure temperature values of 800-900 °C for zircon U-Pb, c. 170-190 °C for ZHe (Reiners et al. 2004), and c. 45-75 °C for AHe Farley 2000) to construct simplified cooling trajectories from these rocks. Because the formation ages of the igneous rocks range from 215 to 37 Ma, the early stage cooling histories of these samples are diverse. However, the sample cooling rates in the Cenozoic show broadly similar cooling histories, most with slow cooling rates <1.5 °C Ma −1 . An example is the biotite granodiorite (DO3), which has a formation age of 215.7 ± 3.3 Ma and an average cooling rate of c. 9.5 °C Ma −1 between 216 and 147 Ma, followed by a very slow cooling rate of c. The above calculations assume a linear and uninterrupted cooling path between each closure temperature, which may not be realistic. Therefore to obtain more detailed information on the low-temperature thermal histories paired apatite and zircon (U-Th)/He ages from four representative samples (D27, D28, D56, and D32) were selected for inverse modelling using the HeFTy program (Ketcham 2005). We applied the annealing model of Flowers et al. (2009) for AHe and the diffusion kinetics of Reiners et al. (2004) for ZHe. The average values of grain radius and U and Th concentrations of each sample were assigned to each population. Modelling results of samples D27, D28 and D56 from Group 1 showed a transition from faster (3-5 °C Ma −1 ) to very slow cooling (<1 °C Ma −1 ) at c. 50 Ma (Fig. 3a-c), whereas those of sample D32 from Group 2 showed a faster cooling rate (c. 5 °C Ma −1 ) between 75 and 65 Ma, a very slow cooling rate (<1 °C Ma −1 ) between 65 and 20 Ma, and finally a faster cooling rate (3-5 °C Ma −1 ) since c. 20 Ma (Fig. 3d). These results are consistent with the cooling rates extracted from multi-system thermochronometers outlined in the previous paragraph. Using a palaeogeothermal gradient of 25-30 °C km −1 , similar to that employed by Clark et al. (2005) in the eastern Tibetan Plateau, and surface temperatures of around 5 °C, the modelled samples of Group 1 yield average exhumation rates of 0.04-0.02 km Ma −1 since c. 50 Ma, whereas sample D32 has average exhumation rates of 0.2-0.1 km Ma −1 since c. 20 Ma.

Discussion
In the following sections, we consider the data-constrained erosion histories in the context of the presence (or not) of significant topography in central Tibetan Plateau and the margins to the north and east through the Mesozoic and Cenozoic to the present.

Mesozoic magmatism and exhumation
Two granitoid samples (D56 and D03) from the central Songpan-Ganzi have similar Late Triassic magmatic crystallization ages around 214 Ma, whereas their ZHe ages are 103.6 ± 6.5 Ma and 146.5 ± 10.0 Ma (Table 2), respectively. Multi-system thermochronometers reveal that these samples have experienced relatively rapid cooling during the Late Triassic and Early Cretaceous. Structural analysis and thermochronological data in the eastern Songpan-Ganzi reveal that folding and thrusting within the Triassic flysch occurred during Late Triassic-Early Jurassic time (Burchfiel et al. 1995;Worley & Wilson 1996;yin & Harrison 2000;yan et al. 2011). Zhang et al. (2006) attributed the adakitic geochemistry of the yanggon and Maoergai granitoids (Fig. 1) in the northeastern part of the Songpan-Ganzi to partial melting of thickened lower crust (>50 km) in the Late Triassic, indicating significant crustal thickening prior to emplacement. Rapid cooling during the Early Cretaceous, long after emplacement-related cooling, is consistent with high rates of erosion and would require the existence of significant topography and relief at least in some parts of the Songpan-Ganzi belt at this time. In contrast, the data of Roger et al. (2010Roger et al. ( , 2011 indicate a very slow Jurassic-Cretaceous cooling history of Mesozoic granites from the southern and eastern Songpan-Ganzi (including the Longmen Shan region).
Four granitoid samples (D27, D28, D30, and D32) from the central Tibetan Plateau possess Jurassic-Cretaceous crystallization ages. Sample D27 has a weighted mean zircon U-Pb age of 179.9 ± 2.0 Ma consistent with previous studies of the same area (e.g. 185-170 Ma; Guynn et al. 2006). This stage of magmatism was related to the development of a continental arc during Bangong ocean northward subduction (Guynn et al. 2006). The other samples dated by U-Pb gave closely similar Early Cretaceous ages between 114 and 111 Ma. Such ages are new to this area but are common elsewhere in the Lhasa block owing to widespread early Cretaceous postorogenic peraluminous to calc-alkaline magmatism (between 135 and 100 Ma). Although magmatism overlapped with extensional deformation and a marine transgression there must have been some topography during this time, as our ZHe data for samples D26, D27, D28 and D32 yielded cooling ages between 74.3 ± 11.0 and 115.1 ± 11 Ma, consistent with previous studies that recognized Cretaceous exhumation in the central-western parts of the Qiangtang and Lhasa terranes (Murphy et al. 1997;Ding & Lai 2003;Hetzel et al. 2011;Rohrmann et al. 2012), which is possibly related to flat slab subduction of the yarlung-Zangpo neo-Tethys that took place between 90 and 78 Ma ).

Low long-term exhumation rates
AHe ages of five samples from the central Tibetan Plateau vary from 65.9 ± 6.4 to 35.7 ± 9.4 Ma and most are older than 50 Ma (Fig. 2), indicating <2 km of exhumation throughout the Cenozoic, similar to the results obtained by a multi-thermochronometer study of central Tibet conducted by Rohrmann et al. (2012). That study also documented rapid to moderate exhumation through the Late Cretaceous to Eocene followed by slow rates of exhumation (<0.05 mm a −1 ) since c. 45 Ma. Here we emphasize that the interior of the Tibetan Plateau experienced long-term low rates of exhumation, suggesting that the central and eastern parts of the Tibetan Plateau have experienced relative tectonic and geomorphological stability since the early Eocene (van der Beek et al. 2009;Wilson & Fowler 2011;Rohrmann et al. 2012), which is in accord with other observations. numerical modelling and structural geological mapping show that more than 50% upper crust shortening of the Lhasa and Qiangtang blocks occurred during the Cretaceous-Early Eocene (England & Searle 1986;Ratschbacher et al. 1992;Murphy et al. 1997;Kapp et al. 2003Kapp et al. , 2005Kapp et al. , 2007Guynn et al. 2006) and suggest that crustal thickness conditions capable of isostatically sustaining a plateau may have existed in central and eastern Tibet by at least 50-40 Ma. Geochronological and geochemical studies of the Duogecuoren adakitic rocks from the central-western Qiangtang block imply that the onset of surface uplift of central Tibet might have occurred as early as 45-38 Ma because these rocks were derived from a garnetbearing source that represents the crustal thickening of the Qiangtang crust (Wang et al. 2008b). Stable isotope-based palaeoaltimetry of late Eocene deposits of the Lunpola basin and Oligocene palaeosol carbonate of the nima Basin reveals that their current elevations were attained at least before 35 Ma Currie 2006) and26 Ma (DeCelles et al. 2007).

Implications for mechanisms of Tibetan Plateau growth
If the central and eastern parts had constituted an extensive proto-Tibetan Plateau as early as the Eocene, this challenges the existing mechanisms for Cenozoic deformation and surface uplift. Various competing mechanisms, including the following, have been proposed that predict distinct temporal-spatial uplift patterns.
(1) Convective removal of the lower lithosphere. This model would produce a rapid increase in the mean elevation over a broad region, and predicts that the whole of Tibet gained its elevation simultaneously (Molnar et al. 1993). This model is inconsistent with either stepwise growth (Tapponnier et al. 2001) or proto-plateau models (Wang et al. 2008a).
(2) Underthrusting of Indian lithosphere as far as the northern Qiangtang block resulting in crustal thickening and then surface uplift (DeCelles et al. 2002). Although this model suggests a northward expansion of the Tibetan Plateau, the Hi-CLIMB seismic array indicated that underplating of Indian lithosphere is limited to the Lhasa block (nábělek et al. 2009), implying that this mechanism might play a critical role for surface uplift within only the Lhasa block.
(3) Crustal thickening by continental subduction and partial melting associated with Palaeogene magmatism (Spurlin et al. 2005). Continental subduction might have been a major contributor to the early uplift of the Qiangtang block and the Songpan-Ganzi block (Tapponnier et al. 2001).
(4) Crustal thickening by lower crustal flow (Royden et al. 1997;Clark & Royden 2000). Late Miocene (10-15 Ma) lower crustal flow is widely invoked to explain topographic expansion of the northern and eastern margins and the steep topography between Longmen Shan and the Sichuan basin (Duvall et al. 2012). This model is controversial and is unlikely to have been a widespread phenomenon; for example, owing to deep-seated faults that may interrupt flow (yao et al. 2008). Furthermore, c. 45% shortening of upper crust during the Palaeogene in the yushu-nangqian thrust belt (Spurlin et al. 2005), Oligocene crustal thickening along the Longmen Shan (Wang et al. 2012), strong crustal shortening revealed by seismic reflection in the northeastern Tibetan Plateau (Wang et al. 2011), and balanced geological cross-sections across the Longmen Shan and Sichuan basin (Hubbard & Shaw 2009), are not in accord with crustal flow models. However, we should point out that this does not preclude lower crustal flow operating in certain areas during the late Cenozoic. Therefore, surface uplift in central and eastern Tibet is likely to be heterogeneous, and attributed to different mechanisms at different times.

Miocene river incision
Three samples of Group 2 (D32, D36, D45), collected from the headwaters of the Salween and Mekong river gorges (Fig. 1) yielded AHe weighted mean ages of 15.0 ± 0.8 Ma, 16.3 ± 2.5 Ma and 15.0 ± 3.2 Ma, respectively. The thermal modelling result of sample D32 show a three-stage cooling history with fast cooling at c. 76-65 Ma, slow cooling between 65 and 20 Ma, and finally accelerated cooling since c. 20 Ma (Fig. 3d). Thus, the Miocene AHe ages of these samples probably reflect the timing of major river incision and are consistent with other studies in the area. AHe data from three elevation transects from Dadu River, yalong River  (Ketcham 2005). Light grey path envelopes denote acceptable fit to the data (goodness-of-fit parameter (GOF) >0.05); dark grey path envelopes denote good fits to the data (goodness-of-fit parameter >0.5); continuous curves show weighted mean path. Measured and predicted AHe and ZHe ages are given within panels. For each inversion, random paths were generated for 10000 paths. The weighted mean age with 2σ internal errors and the U and Th mean value of each sample were used for modelling. These mean values of sample D32 do not include grain D32-2 and D32-3. and yangtze River gorges indicated that regional river incision began c. 15 Ma (Ouimet et al. 2010) or between 9 and 13 Ma (Clark et al. 2005). For the yarlung Zangbo gorge incision initiated at c. 11-8 Ma (Dai et al. 2013), and for the Lhasa River gorge at 15 Ma (Rohrmann et al. 2012). Duvall et al. (2012) applied a novel inverse modelling technique (Avdeev et al. 2011) to the interpretation of detrital apatite fissiontrack and (U-Th)/He data, and suggested that the entire Tibetan Plateau experienced an abrupt increase in erosion rate between 11 and 4 Ma following a long period of slow erosion. The vast majority of AHe ages in the Duvall et al. study are clustered around the late Mesozoic-early Cenozoic and the wide distribution of ages in their dataset was interpreted to represent a long period of slow cooling or no erosion with a relative increase in erosion rate toward the present. In contrast, the present study, based on bedrock samples and multiple thermochronometers, has found a different history marked by rapid cooling through the late Mesozoic-early Cenozoic, followed by a decrease in cooling rate through most of the Cenozoic. Group 1 AHe ages of five samples from the central Songpan-Ganzi vary between 74.7 ± 19.0 and 45.2 ± 18.0 Ma (Table 1; Fig. 2). Inverse modelling results of AHe and ZHe ages from one representative granitoid sample (D56) show that it underwent fast cooling (3-5 °C Ma −1 ) between c. 100 and 60 Ma followed by a much lower rate (<1 °C Ma −1 ) to the present, corresponding to exhumation rates of 0.04-0.02 km Ma −1 . Only samples from the headwaters of the Salween and Mekong rivers record accelerated incision at c. 15 Ma.
Differences between the study of Duvall et al. (2012) and ours relate in part to approaches to sampling and the analytical and interpretative strategies. For example, Duvall et al. (2012) treated each apatite fission-track and AHe age as a true age rather than recognizing that each grain age is a random value from a Poisson distribution. Furthermore, extraction of cooling histories from detrital datasets assumes that erosion is uniformly distributed within each catchment, which might not be the case if base levels have changed as a result of recent local deformation events in contrast to a regional surface uplift (e.g. the multiple terraces of the yellow River). The susceptibility of apatite to weathering in the near-surface environment, particularly under conditions of slow erosion, may also result in a detrital grain-date population that is naturally skewed toward apatite samples from faster eroding parts of the catchment. In light of the assumptions and potential biases of the detrital method, it remains difficult to assess the regional significance of relatively few young dates.
It is tempting to link timing of incision with intensification of the East Asia monsoon. Study of core taken at Ocean Drilling Program (ODP) Site 1143 used upwelling-related radiolarian abundances to suggest that the summer monsoon intensified c. 12-11 Ma and reached a maximum strength at c. 8.2 Ma (Chen et al. 2003), and a multiproxy sedimentary study of Site 1146 highlighted changes at 15, 8 and 3 Ma (Wan et al. 2007). However, this is tied to understanding the effects of the orography of the Tibetan Plateau on the Asian monsoon systems (Liu & yin 2003). A wider 'chicken or egg' question is how much of the acceleration in river incision is due to local river base-level change compared with the enhanced erosion associated with the increase in precipitation owing to monsoon intensification. If there was a late Miocene plateau-wide increase in erosion associated with monsoon intensification, our results show that the depths of erosion have not been sufficient to affect the apatite chronometers, except in the river gorges near the plateau margins. Fig. 4. Integrated low-temperature thermochronological AHe and AFT data from references (Clark et al. 2005;Reid et al. 2005;Wang et al. 2008a;Ouimet et al. 2010;Hetzel et al. 2011;Wilson & Fowler 2011;Rohrmann et al. 2012) and this study.