Results for AFT dating are presented in Table 1. Most samples yielded 60–90 single grain ages. Central ages ranged from 43.7 ± 1.9 Ma to 78.0 ± 2.8 Ma. The sample in the bottom of the section (HLG100) has an AFT age younger than its depositional age (53 Ma). All other upper samples have central ages older than corresponding depositional ages. AFT central ages do not show regular changes with depositional ages (Table 1). Most samples fail the χ² test with P(χ²) <5%, indicating heterogeneous age components (Green, 1981). This can occur if grains in these samples were not totally reset during post-depositional burial and still record mixed cooling signals of a variety of source terranes. For samples with heterogeneous grain ages, the observed age distributions of each sample were decomposed into grain age components using RadialPlotter and DensityPlotter programs (Vermeesch, 2012). The dispersion of single-grain ages are shown in Figure 3 as radial plots. The single grain age-density distributions and mixture model peak ages for all samples are plotted in Figure 4. In total, 21 age components were obtained, which have peak ages ranging from 21.2 ± 2.9 Ma to 98 ± 10 Ma (1σ). These decomposed age components were divided into three age populations (P1, P2, and P3) based on correlation of their peak ages (Table 1). A lag-time plot shows the relationship between decomposed AFT component peak ages and depositional ages of corresponding samples is present in Figure 5. Notably some components have peak ages younger than corresponding depositional ages and show negative lag-time. Both indicate these samples may be partially annealed after deposition (Figure 5).

Most of our samples yielded 80–120 confined track length measurements. The c-axis corrected mean track lengths range from 12.83–13.28 μm (with an exception of 14 measurements giving a length of 11.39 μm). Average Dpar values of these length-measured grains range between 1.22 and 1.65 μm. Histogram plots of the track length distribution of each sample are presented in Figure 6. Integrating the heterogeneity of grain ages and the relatively long track lengths of our samples, their post-depositional burial annealing may not intensity.

The mean track lengths of our samples are gradually shortening with sampling position down in the section (Figure 7). It is more obvious when integrated with published track length data in strata overlying the sampled section (He et al., 2021). The mean track lengths were nearly constant in the upper half of the section, but become shortened in the lower half of the section (Figure 7). The lower sample has shorter length suggests the shortening of track length in the lower half of the section should be caused by burial annealing. Analogous to track length variation, the component ages show younger trend down-section in the lag-time plot (Figure 5), which conform to the law of burial annealing. Therefore, AFT ages and length data together illustrate that samples experienced post-depositional partial annealing.

Detrital rocks can be complex sedimentary mixtures derived from a variety of source terranes with variable cooling histories. During post-deposition burial, grains were reset by partial annealing. The apparent AFT ages and lengths in our data are the partially reset cooling signals of mixtures of different sedimentary sources. Broadly, grains in one detrital sample in this study experienced the same thermal history after deposition. This coherence permits the modeling of the post-depositional thermal history of our samples based on the AFT age and length data. Our samples experienced deposition, burial, and exhumation, we attempted to decipher the time constraints of these processes.

We used the HeFTy program (Ketcham, 2005) to model the thermal history. Our data of AFT ages, lengths, and Dpars are included in the annealing model of Ketcham et al. (2007). Eight of ten samples with numerous confined track lengths were modeled. We defined prior constraints for the thermal modeling as follows: each sample, using their corresponding depositional age with ±5 Ma error, was included in a 20 ± 20°C palaeo-surface temperature; each sample experienced the upper AFT partial annealing zone at 70 ± 10°C after deposition; each sample, an incipient condition of one to two times the AFT age, was included in a temperature of partial or full annealing (60–140°C). We give each constraint large ranges in time and temperature to model freely depend on the program. All models run with 500,000 iterations.

Because thermal evolutions of grains in a detrital sample are a mixture of pre-depositional conditions and a unity of post-depositional conditions, only the post-depositional thermal history is credible. The inversion results show thermal histories with two stages during post-deposition with a thermal transition in age ranges from 10–3 Ma (Figures 8, 9). After deposition, samples experienced a gradual heating phase in the early Cenozoic up to 10–3 Ma. By this time, samples reached the upper range of the AFT partial annealing zone (60–80°C). After 10–3 Ma, samples experienced a rapid cooling phase and were exhumed to the surface (Figures 8, 9). Modeled thermal histories are in accord with the inferred deposition-burial-exhumation process for the samples, as constrained by geological framework. The heated phase corresponding to the gradually buried, and the thermal condition converted to cooling at ∼10–3 Ma indicated the commenced exhumation. The exhumation since ∼10 Ma may be caused by intensified deformation of the sampled strata (Yin et al., 2008), which finally inclined and exposed the section at the surface by the present.

Although we adopt the stratigraphic chronology of Fang et al. (2019) in the HLG section as depositional age constraints, another rather younger stratigraphic chronology assignment of Wang et al. (2017) in the Honggou section (see section Geological Setting for details) is tried for thermal modeling. We changed the prior constraints of depositional ages to Wang et al. (2017) for modeling the representative samples (Figure 10). The modeled post-depositional thermal histories show transition from heating to cooling after ∼10 Ma (Figure 10), which are nearly the same with that using constraints of Fang et al. (2019). This result suggests that the different depositional age constraints of the two schemes do not affect the post-depositional thermal histories of our samples significantly.

The early Cenozoic tectonic deformation of the northern Qaidam Basin is demonstrated by growth strata in many seismic profiles (Yin et al., 2008; Yu et al., 2017; Cheng et al., 2019). This syndepositional and post-depositional tectonic deformation, which inclined the sampled strata, resulted in the burial depth much shallower than the strata thickness. Thus, our AFT samples, even that in the ∼5,000 m thickness, experienced moderate annealing after deposition. The early Cenozoic deformation has been observed across the NQQB (Qi et al., 2016; Jian et al., 2018; Zhuang et al., 2018; Lin et al., 2019; Li et al., 2020; He et al., 2020, 2021). The AFT thermal modeling results indicate the northern Qaidam Basin experienced exhumation since ∼10 Ma. The exhumation should be caused by intensified thrusting and folding in this region (Yin et al., 2008; Cheng et al., 2019). Although the deposition continued through ∼10 Ma across the northern Qaidam Basin, the strata considered in this study was gradually exhumed. This exhumation-related tectonic deformation possibly resulted from the enhanced activity of thrust faults in the Lulehe anticline (Figure 2B) since ∼10 Ma. As the HLG section locates between the Saishiteng Shan and Luliang Shan, which is the outmost thrust belt of the NQQB, we suggest tectonism transmitted to this structure belt at ∼10 Ma and caused persistent deformation since then.

Published in situ thermochronology data revealed widespread middle-late Miocene rapid exhumation in the NQQB (Figure 1A). The intense tectonic deformation of the northern Qaidam Basin that commenced at ∼10 Ma is also supported by synchronous tectono-sedimentary records in surrounding basins. Several tectonic traces at ∼10–8 Ma both recorded in the northern Qaidam Basin and Jiuquan Basin that surrounding the NQQB, such as: the beginning accumulation of coarse clasts (Fang et al., 2005, 2007; Li et al., 2014); the syndepositional deformation shown in seismic or exposure profiles (Yin et al., 2008; Li et al., 2014); abruptly increased accumulation rates (Fang et al., 2005, 2007; Ji et al., 2017); initiation of sedimentary recycling in basins margin (Wang et al., 2017; Zheng et al., 2017; Pang et al., 2019a). In addition, the cease of rotation of the Jiuquan Basin suggests that the domination of shortening and uplift since then (Yan et al., 2012).

Thermochronology data together with the tectono-sedimentary data suggest the sedimentary source uplifted synchronous with the marginal basin. The southernmost Qilian Shan (Qaidam Shan), which locates close to the HLG section, have deformed at 15–10 Ma (Meng et al., 2020). The appearing of tectonic deformation in the northern Qaidam Basin since ∼10 Ma delayed the deformation of the southern Qilian Shan, indicating a southward extension of deformation in the southernmost Qilian Shan-northern Qaidam Basin. However, the spatial and temporal distribution of the Miocene tectonic events in the NQQB does not shows any regular migration (Figure 1A). We suggest the NQQB experienced intensified tectonic deformation since the late Miocene and the basin-ward migration of intense deformation is limited to the marginal basin.