The burial dating method is based on the relative radioactive decay of two cosmogenic nuclides, ²⁶Al and ¹⁰Be, which accumulate with a known ²⁶Al/¹⁰Be production ratio of 6.61 within the quartz (SiO₂) mineral fraction (in situ production) of rocks exposed at the earth's crust surface due to nuclear reactions induced by the cosmic ray derived energetic particles on silicon (Si) and oxygen (O). Since the cosmic ray flux is efficiently attenuated by matter, the deposition of a few meters of sediments over a previously exposed surface (burial) leads to a sufficient reduction of the effective energetic particle flux to stop the ²⁶Al and ¹⁰Be production. In the absence of production, the initial concentrations of each cosmogenic nuclide consequently start to radioactively decay according to their respective half-life, that is 0.717 ± 0.017 Ma for ²⁶Al (Granger, 2006 and Samworth et al., 1972) and 1.387 ± 0.012 Ma for ¹⁰Be (Chmeleff et al., 2010 and Korschinek et al., 2010). The ²⁶Al concentration thus decreases approximately twice as fast as that of ¹⁰Be, the ²⁶Al/¹⁰Be ratio decreases exponentially with an apparent half-life of 1.48 ± 0.04 Ma. This method thus allows determining quartz mineral burial duration from 100 ka (100,000 years) to approximately 5 Ma (Granger and Muzikar, 2001).

The physico-chemical treatments performed on the four studied quartz pebbles as well as the accelerator mass spectrometry measurements at ASTER of their ¹⁰Be and ²⁶Al concentrations followed the protocols and parameters fully described in Lebatard et al. (2014a) and references therein.

The concentrations measured for these two cosmogenic nuclides in the same quartz sample insure that they both record the same history in term of exposition, denudation and burial. They allow calculating the ²⁶Al/¹⁰Be ratio associated to each sample and, consequently, to determine their corresponding burial duration and denudation rate using the method fully described in the SOM of Pappu et al. (2011).This modeling method is based on the equation (1) describing the evolution of the in situ produced cosmogenic nuclide concentration C(x,ɛ,t) as a function of the depth (x), the denudation rate (ɛ) and the time (t): (1) C ( x , ε , t ) = C ( x , 0 ) ⋅ e − λ t + P n ⋅ e − x Λ n ⋅ ( 1 − e − ( λ + ε Λ n ) t ) ε Λ n + λ + P μ sl ⋅ e − x Λ μ sl ⋅ ( 1 − e − ( λ + ε Λ μ sl ) t ) ε Λ μ sl + λ + P μ ft ⋅ e − x Λ μ ft ⋅ ( 1 − e − ( λ + ε Λ μ ft ) t ) ε Λ μ ft + λ \[ C(x,\epsilon ,t)=C(x,0)\cdot {\text{e}}^{-\lambda t}+\frac{{P}_{n}\cdot {\text{e}}^{\frac{-x}{{\Lambda }_{n}}}\cdot (1-{\text{e}}^{-(\lambda +\frac{\epsilon }{{\Lambda }_{n}})t})}{\frac{\epsilon }{{\Lambda }_{n}}+\lambda }+\frac{{P}_{\mu \text{sl}}\cdot {\text{e}}^{\frac{-x}{{\Lambda }_{\mu \text{sl}}}}\cdot (1-{\text{e}}^{-(\lambda +\frac{\epsilon }{{\Lambda }_{\mu \text{sl}}})t})}{\frac{\epsilon }{{\Lambda }_{\mu \text{sl}}}+\lambda }+\frac{{P}_{\mu \text{ft}}\cdot {e}^{\frac{-x}{{\Lambda }_{\mu \text{ft}}}}\cdot (1-{\text{e}}^{-(\lambda +\frac{\epsilon }{{\Lambda }_{\mu \text{ft}}})t})}{\frac{\epsilon }{{\Lambda }_{\mu \text{ft}}}+\lambda } \] where λ is the radioactive decay constant, P _n is the spallogenic production rate linked to the neutrons, P _μsl and P _μft are production rates linked to slow and fast muons, respectively (Braucher et al., 2011 and associated references), Λ _n, is the neutrons attenuation length, and Λ _μsl and Λ _μft, are the slow and fast muons attenuation lengths, respectively. Modeling were computed using the parameters discussed in Braucher et al. (2011), including the ²⁶Al/¹⁰Be spallogenic production rate ratio of 6.61 ± 0.50. Neutronic production rates have been scaled using (Stone, 2000) and are based on a weighted mean ¹⁰Be spallation production rate at sea level and high latitude (SLHL) of 4.03 ± 0.18 at g⁻¹ a⁻¹ (Molliex et al., 2013).

Uncertainties associated with the ratios, the durations and the denudation rates, reported as 1σ, result from the propagation of the uncertainties previously referred and described (Table 1).

The performed modeling method allows to determine minimum burial durations, based on the sole differential cosmogenic nuclides radioactive decay, and the estimated associated denudation rates (Table 1: “Model without post-burial production”), and maximized burial durations, based on the assumption that the environmental conditions remained relatively stable since the burial, and the associated denudation rates (Table 1: “Model with post-burial production”). Minimum burial durations are also obtained using the exposure–burial diagram (Fig. 5; e.g., Granger, 2006), which aims to reproduce the minimum burial duration required to lead by radioactive decay from an initial ²⁶Al/¹⁰Be concentration ratio conditioned by the denudation rate before burial to the measured ratio. The measured value may, however, also result from different and more complicated scenarios involving repeated burials and exposures which would obviously lead to significantly longer burial duration.

Table 1 summarizes all burial durations, denudation rates and post-burial concentrations calculated or graphically determined. The results are also plotted in the graph ²⁶Al/¹⁰Be versus ¹⁰Be (Fig. 5), also called “exposure–burial diagram” (e.g., Granger, 2006).

Paleomagnetic sampling was performed all along the thickest sedimentary deposit section at Mansu-Ri (Locality IV; Fig. 2). All the sampling was handmade, oriented with a compass and consolidated with plaster tape. In these thus obtained 20–25 cm long and 7 cm large and wide oriented cores, cubes (7–8/core) were cut for natural remanent magnetization (NRM) measurements. The measurements were made using a 2G DC-Squid Superconducting Rock Magnetometer (SRM) at CEREGE. The direction of the characteristic remnant magnetization (ChRM) was retrieved by means of stepwise demagnetization using alternating field (AF) up to 60 or 80 mT.

The results obtained for the two quartz pebbles from Mansu-Ri (Loc. IV; a and b, Fig. 4) sampled 6 m beneath the surface and for the two quartz pebbles from Wondang-Jangnamgyo (c and d, Fig. 4) sampled ∼2 m beneath the surface are presented in Table 1. In Fig. 5, they are plotted on a graph (exposure–burial diagram; e.g., Granger, 2006) presenting the evolution of the ²⁶Al/¹⁰Be ratio as a function of the ¹⁰Be concentration.

Regarding the four studied quartz pebbles, the minimum burial durations obtained with the model without post-burial production range from 331.9 ± 18.0 ka to 1.28 ± 0.19 Ma associated with denudation rates between 3.9 ± 0.6 m Ma⁻¹ to 17.8 ± 0.9 m Ma⁻¹. In the exposure–burial diagram (Fig. 5), the pebbles all located in the “burial area” lead to minimum burial durations ranging from 325 ± 100 ka for the youngest (WJ¹⁰BeS-3) and to 1.28 ± 0.30 Ma for the oldest (WJ¹⁰BeS-1) and denudation rates affecting the overlying surfaces ranging from 4.0 to 17.7 m.Ma⁻¹, similar to the data obtained performing the model without post-burial production. However, it is worth mentioning that in both sites the samples were not deeply buried, especially those from Wondang-Jangnamgyo and that therefore, due to the thin soil cover, post production may have occurred leading to burial durations that should be longer. Performing the model with post-burial production to thus consider possible cosmogenic nuclide production during burial yields maximized burial durations and denudation rates ranging from 382.4 ± 20.8 ka to 2.77 ± 0.41 Ma and from 2.5 ± 0.4 m Ma⁻¹ to 18.8 ± 0.9 m Ma⁻¹, respectively. The denudation rate ranges obtained by the three methods are similar and the rates at the Mansu-Ri (Loc. IV) site are consistent.

Due to the sampling depths, the post-burial cosmogenic nuclide production is significant at both sites, between 3 and 7% at the Mansu-ri (Loc. IV) site and 16 to 67% at the Wondang-Jangnamgyo site.

Due to the data dispersion, no weighted mean burial duration can be calculated. Considering the individual durations and their associated uncertainties at the Mansu-Ri (Loc. IV) site, the burial duration of the two quartz pebbles from the archaeological level 5-c is longer than 235 ka (obtained by subtracting it uncertainty to the youngest age [MS¹⁰Be S-2] derived from the diagram, that is 330–95 = 235 ka), but shorter than 494.6 ka (obtained by adding it uncertainty to the oldest age [MS¹⁰Be S-1] derived from the model with post-burial production, that is 464.2 + 30.4 = 494.6 ka). Similarly, at the Wondang-Jangnamgyo site, the burial duration of the WJ¹⁰Be S-3 quartz pebble coming from the archaeological level V is longer than 225 ka, but shorter than 621.2 ka.

The raw NRM intensity values of the sediment range between 30 and 80 × 10⁻³ A/m (Fig. 6). The maximum values occur in the sandy silt and in the plastic silty clay, at the base of the paleosoils (50 à 80 × 10⁻³ A/m). The minimum values occur in paleosoils and sandy levels (< 10 × 10⁻³ A/m). The high magnetic intensity values below the paleosoils may result from an intense weathering of the sediments followed by a downward migration of the iron oxides which then accumulated in the underlying sedimentary layers. These accumulations of iron oxide suggest that the deposits underwent, at least, four hot and wet climatic periods.

When confronted with the Chinese loess record, the high NRM intensity values measured along the sedimentary deposit section at Mansu-Ri (Locality IV) correspond to the high magnetic susceptibility values measured along the S1 to S5 Chinese soils (Fig. 6).

The inclination of the magnetic components varies between 34° and 66°. Nevertheless, the base of core 29 presents a significantly lower value of 14°. The declination of cores 1 to 29 exhibits values close to 0°.

Generally, all the cores analyzed in this locality present a normal magnetic polarity, similar to the current polarity. The average inclination is of the order of 50°, similar to the current inclination measured at Chengwon (South Korea) equals to 52°.

This normal polarity sequence can be thus undoubtedly awarded to the period of Brunhes younger than 780,000 years, with no hints of an excursional record.