10Be-inferred paleo-denudation rates imply that the mid-Miocene western central Andes eroded as slowly as today

Terrestrial cosmogenic nuclide concentrations of detrital minerals yield catchment-wide rates at which hillslopes erode. These estimates are commonly used to infer millennial scale denudation patterns and to identify the main controls on mass-balance and landscape evolution at orogenic scale. The same approach can be applied to minerals preserved in stratigraphic records of rivers, although extracting reliable paleo-denudation rates from Ma-old archives can be limited by the target nuclide’s half-life and by exposure to cosmic radiations after deposition. Slowly eroding landscapes, however, are characterized by the highest cosmogenic radionuclide concentrations; a condition that potentially allows pushing the method’s limits further back in time, provided that independent constraints on the geological evolution are available. Here, we report 13–10 million-year-old paleo-denudation rates from northernmost Chile, the oldest 10Be-inferred rates ever reported. We find that at 13–10 Ma the western Andean Altiplano has been eroding at 1–10 m/Ma, consistent with modern paces in the same setting, and it experienced a period with rates above 10 m/Ma at ~11 Ma. We suggest that the background tectono-geomorphic state of the western margin of the Altiplano has remained stable since the mid-Miocene, whereas intensified runoff since ~11 Ma might explain the transient increase in denudation.


Variable canyon incision rate
The paleo-denudation rates reported in the main text are obtained assuming a constant incision rate during the phase of progressive exhumation, which equals the maximum sample depth z below the section's top (1947 m asl), divided by the incision time (5.9 ± 0.5 Ma). Here we show the possible effect of two consecutive 3-Ma-long incision phases at different rates, keeping the total cumulative incision equal to z. Fig. S2 relates to a scenario where the first phase carves 33% of z and the second one completes the remaining 66% (twice as fast). This scenario reduces the Cincision component of the uppermost samples, hence increasing Cinherited and reducing the paleo-denudation rates. Fig. S3 relates to the opposed scenario, where the first phase carves 66% of z and the second one completes the remaining 33%. In this case Cincision becomes more relevant, and the youngest samples only yield minimum paleo-denudation rates. Both scenarios show that the youngest and uppermost samples are the most sensitive to incision rate variations, evidencing that the related paleo-denudation rates can hardly be constrained. Nevertheless, the resulting trend remains analogous to the one reported in the main text, with an intensification of denudation rates above 10 m/Ma at ~11 Ma. Figure S2. Paleo-denudation rates reported in the main text (diamonds with errorbars) compared to those obtained considering two 3-Ma-phases of canyon incision, where the second phase occurs at twice the rate (squares within rose 1σ envelope). Figure S3. Paleo-denudation rates reported in the main text (diamonds with errorbars) compared to those obtained considering two 3-Ma-phases of canyon incision, where the second phase occurs at half the rate (squares within green 1σ envelope).

Shielding factor during canyon incision
The paleo-denudation rates reported in the main text are obtained assuming a constant shielding factor (measured in the field) during exhumation. Although the paleotopography is poorly known, the fieldmeasured shielding factors add a good constraint on the local production rate during the recent history of canyon incision, because the relevance of Cincision increases as the sampled material is progressively exhumed. Nevertheless, using a topographic shielding factor based on the average valley-flanks might better reproduce the average shielding throughout the entire period of canyon incision. We explore this scenario by recalculating Cincision with a shielding factor of 0.9 (as opposed to ~0.75). The effect of this calculation is shown in Fig. S4. Although 3 of the 4 youngest samples would only yield minimum paleodenudation rates, the trend highlighted in the main text remains unaltered, if not strengthened by even higher rates at ~11 Ma. Figure S4. Paleo-denudation rates reported in the main text (diamonds with error bars) compared to those obtained using a Shielding Factor of 0.9 during canyon incision (squares within yellow 1σ envelope).

Production at depth preceding exhumation
The paleo-denudation rates reported in the main text are obtained neglecting possible production at depth between the deposition of the top of El Diablo and the start of incision. This period would have lasted approximately 3 Ma, allowing for relevant muogenic production even at several tens of meters depth. This concentration, however, becomes negligible after 6 Ma of radioactive decay, which is the approximate time of canyon incision. Fig. S5 shows that the results reported in the main text and paleo-denudation rates calculated accounting for these concentrations can be considered almost equivalent. Figure S5. Paleo-denudation rates reported in the main text (white diamonds) compared to those obtained considering 3 Ma of possible production at depth before canyon incision (squares within blue 1σ envelope).
Uncertainty on the paleo-catchment production rate The paleo-denudation rates reported in the main text are obtained neglecting additional uncertainties on the source area production rates, possibly due to fluctuations of the magnetic field strength. Such variations are not reported for times preceding ~2 Ma, but assuming that the mid-Miocene magnetic field was subject to similar fluctuations, our estimates should take into account the uncertainty deriving therefrom. In Fig. S6, the red curve shows the Lifton et al. (2014) 1-ka-Scaling Factor variations, whereas the blue curve is a 100ka-moving average. The blue curve is more appropriate for our study, because at denudation rates <10 m/Ma (comparable to our estimates), 100 cm of eroded material would yield an apparent age >100 ka. This implies that short-term variations in production rate due to magnetic field strength need to be averaged over such integration time. Here, we therefore explore two conservative scenarios, in which the source area production rate is either increased or decreased by 11%, as suggested by the standard deviation of the blue curve in Fig.  S6. The results of this sensitivity test are shown in Fig. S7, where it appears that the paleo-denudation rates would substantially remain unchanged.  . Paleo-denudation rates reported in the main text (diamonds with error bars) compared to those obtained increasing the source area production rate by 11% (squares in yellow envelope), and those obtained decreasing the source area production rate by 11% (triangles in blue envelope).

Magnetostratigraphic ages
The chronology of the Francia section was initially established by von Rotz et al. (2005) and later only reassessed by Schlunegger et al. (2017), who roughly bracket it between 13-10 Ma based on new geochronological evidence (Jordan et al., 2014). Starting from the work of the latter authors, we refine the age-correlation for each sampled layer, using the Python-based software Cupydon (Lallier et al., 2013). This tool allows an objective evaluation of the possible correlations between the reference GPTS chart (Gradstein et al., 2012) and the measured magnetic polarity zone thicknesses (von Rotz et al., 2005). In particular each sample has been assigned the minimum and maximum age of the related polarity interval. In cases where more samples belong to the same magnetopolarity zone, the time span of the interval has been divided for the number of samples involved. Figure S8 shows this correlation and the inferred accumulation rates.  Table S1 displays the compiled and recalculated 10 Be-inferred current erosion rates for the western central Andes (Abbühl et al., 2011;Kober et al., 2007;Kober et al., 2009;Reber et al., 2017 Table S1. Compiled topographic and cosmogenic nuclide data and relative recalculated current erosion rates.