Since the Last Glacial Maximum, ∼20 k.y. ago, Alpine glaciers have retreated and thinned. This transition exposed bare bedrock surfaces that could then be eroded by a combination of debuttressing or local frost cracking and weathering. Quantification of the respective contributions of these processes is necessary to understand the links between long-term climate and erosion in mountains. Here, we quantified the erosion histories of postglacial exposed bedrock in glacial valleys. Combining optically stimulated luminescence and terrestrial cosmogenic nuclide (TCN) surface exposure dating, we estimated the erosion rate of bedrock surfaces at time scales from 101 to 104 yr. Bedrock surfaces sampled from the flanks of the Mer de Glace (Mont Blanc massif, European Alps) revealed erosion rates that vary from 3.5 ± 1.2 ⋅ 10−3 mm/yr to 4.3 ± 0.6 mm/yr over ∼500 m of elevation, with a negative correlation between erosion rate and elevation. The observed spatial variation in erosion rates, and their high values, reflect morphometric (elevation and surface slope) and climatic (temperature and snow cover) controls. Furthermore, the derived erosion rates can be used to correct the timing of deglaciation based on TCN data, potentially suggesting very rapid ice thinning during the Gschnitz stadial.
To understand the long-term evolution of Alpine landscapes, the respective contributions of surface erosion, sediment production, and sediment transport must be quantified. During the Quaternary period, the alternation between glacial and interglacial periods has modulated the efficiency of glacial, fluvial, and hillslope processes (Koppes and Montgomery, 2009). In that context, changes in bedrock morphology and corresponding sediment delivery have been related to glacier extent, and glacial erosion is often thought to be the most efficient erosional and sediment transport mechanism in mountain environments (e.g., Hallet et al., 1996; Brozović et al., 1997; Montgomery, 2002; Mitchell and Montgomery, 2006; Egholm et al., 2009; Herman et al., 2013; Herman and Champagnac, 2016). Recent studies have also revealed the importance of periglacial processes during interglacial periods (Burbank et al., 1996; Ballantyne, 2002; Scherler, 2015). Yet, the rate at which bare-bedrock surfaces weather and erode during interglacials remains poorly quantified (e.g., Colman, 1981; Zimmerman et al., 1994; André, 2002; Nicholson, 2008; Kirkbride and Bell, 2010).
The erosion of hillslopes in periglacial environments is governed by a combination of landsliding, rock shattering, and weathering (e.g., Anderson and Anderson, 2010). During the last decades, the development of terrestrial cosmogenic nuclide (TCN) methods, mainly using in situ–produced 10Be in quartz crystals, has improved our ability to quantify bedrock surface erosion over time scales from 104 to 106 yr, assuming that erosion occurs steadily through time (Balco et al., 2008; von Blanckenburg and Willenbring, 2014; Hippe, 2017). Over modern time scales, geomorphologists working on frost cracking have also highlighted the feedbacks between temperature variation and snow cover and their effects on the evolution of bedrock surfaces over diurnal to decadal time scales (e.g., Łoziński, 1909; Matsuoka and Murton, 2008). However, bridging the temporal gap between these erosion estimates remains challenging, in part because of the stochastic nature of geomorphic processes (Koppes and Montgomery, 2009; Ganti et al., 2016).
To address these issues, we adopted a new method that combined optically stimulated luminescence (OSL) and TCN surface exposure dating (Sohbati et al., 2018; Lehmann et al., 2019). In this study, we applied this methodology to investigate how erosion rates have evolved over time scales of 101–104 yr on bedrock surfaces of the Mer de Glace (Mont Blanc massif, European Alps) and how morphometric and climatic factors control their evolution. Then, we addressed how the variability in erosion rates can be used to correct TCN exposure ages, leading to very different possible scenarios of ice thinning during the Gschnitz stadial (a period of regionally extensive glacier advance in the European Alps, temporally between the breakdown of the Last Glacial Maximum piedmont lobes and the beginning of the Bølling warm interval).
We collected samples along two elevation profiles at the Mer de Glace (Fig. 1). Six bedrock surfaces were sampled below the Mont Blanc Tête de Trélaporte (MBTP sample sites, west side of the glacier, from 2545 to 2094 m above sea level [masl]; Fig. 1), and three bedrock surfaces were sampled below the Aiguille du Moine (MBAM sample sites, east side of the glacier, ranging from 2447 to 2259 masl; Fig. 1). All surfaces were from the same phenocrystalline granitic lithology of the Mont Blanc massif, and selected sampling sites can all be classified as glacially eroded bedrock surfaces (see the GSA Data Repository1). The surfaces are rough and exhibit a weathered texture without glacial striations. All studied bedrock surfaces were located between the ice-surface elevations of the Little Ice Age (LIA) and the Last Glacial Maximum (LGM; Coutterand and Buoncristiani, 2006; Vincent et al., 2014) and were therefore likely deglaciated sometime between ca. 20–18 ka (Wirsig et al., 2016) and A.D. 1850.
We measured both 10Be concentrations (e.g., Gosse and Phillips, 2001; Ivy-Ochs and Briner, 2014) and OSL profiles (Sohbati et al., 2012) on exposed granitic rock samples (see the Data Repository). The 10Be concentrations provided us with constraints on the time since the rocks were exposed to cosmic rays and with a temporal framework for the possible erosion histories since exposure. OSL profiles constrained the erosion history since the rock exposure to light following ice decay (Lehmann et al., 2019). Note that it is the difference in sensitivity between 10Be and OSL that makes it possible to quantify surface erosion rate histories over short (<102 yr) and long (>104 yr) time scales (Sohbati et al., 2018; Lehmann et al., 2019).
The evolution in time of the OSL bleaching front into a rock surface depends on exposure age, surface erosion, electron trapping and detrapping (bleaching) rates, and athermal loss (Lehmann et al., 2019, their equation 1). While the latter three variables can be constrained from laboratory measurements, the erosion history is unknown. Here, we constrained the erosion history by performing a joint inversion of the OSL and 10Be data for each sample. In this inversion, we assumed that erosion rates evolved as step functions from no erosion to erosion going forward in time (Fig. 2; Lehmann et al., 2019). A reference OSL profile was established by taking the TCN exposure age with no erosion correction (t0). The inferred erosion history included an erosion rate ( in Fig. 2), and two times: tS = the time at which erosion began, and tC = the corrected exposure age. Note that tC was estimated by combining OSL and TCN data. In turn, various solutions for the erosion rate history (tS- pairs) and tC were inferred; high erosion rates and durations that did not fit the observed 10Be concentration data (Lal, 1991) were excluded from the parameter search (forbidden zone in Figs. 3D–3F; see the Data Repository).
RESULTS AND DISCUSSION
In Figures 3A–3C, we show the OSL measurements (i.e., infrared stimulated luminescence at 50 °C) for samples MBTP1, MBTP2, and MBTP11. Three individual cores per sample were sliced such that a depth and an OSL signal could be attributed to each rock slice (green dots in Figs. 3A–3C; Tables DR3–DR5 in the Data Repository). The OSL signal was bleached near the surface and reached a plateau at depth. As a reference profile, a model was computed (Lehmann et al., 2019, their equation 2) using t0 and considering no erosion (dashed black line in Figs. 3A–3C). Erosion rate histories were inferred from paired bedrock surface OSL signals and 10Be concentrations (Figs. 3D–3F and 4A). The most likely solutions (red lines in Figs. 3A–3C and yellow areas in Figs. 3D–3F) were determined by testing 104 pairs of both and tS (combination of 100 values of both parameters) in log space.
The inferred erosion rates varied between 3.5 ± 1.2 × 10−3 mm/yr and 4.3 ± 0.6 mm/yr (Fig. 4A), with high erosion rates (>0.1 mm/yr) over time scales from 101 to 103 yr and low erosion rates (<0.1 mm/yr) over longer time scales, 103 to 104 yr. Such a variation is not due to lithological changes, since the bedrock is uniform. Biological controls are also likely to be minor because of the high elevation; vegetation is not a major component of the environment, and lichen cover does not differ significantly along the profiles. We assumed that these surfaces were mainly affected by grain-by-grain erosion because the character of the glacially eroded bedrock surfaces is preserved; there is no evidence of rockfall scars or surface spalling.
The inferred erosion rates are one to two orders of magnitude greater than previously observed rates for bedrock surfaces globally. Portenga and Bierman (2011) compiled 10Be erosion rates of outcropping bedrock surfaces and found a mean erosion rate of ∼1.2 × 10−2 mm/yr. Studies using in situ 10Be and 26Al exhibited a maximum mean bare-bedrock erosion rate of ∼7.6 × 10−3 mm/yr in arid western North American summits (Small et al., 1997). Part of this disagreement is likely due to the time scale over which TCN erosion rates are averaged, which is typically 104–106 yr. Using reference surfaces such as ice-polished quartz veins preserved on glacially eroded bedrock surfaces, erosion rates of 0.1 × 10−3 to 10 × 10−3 mm/yr have been measured, depending of the lithology and the location (André, 2002; Nicholson, 2008; Kirkbride and Bell, 2010). In contrast, the intensities of erosion rates observed in this study are comparable to other erosion processes such as debris flows and rockfalls (Norton et al., 2010) or subglacial erosion (e.g., Herman et al., 2015, 2018).
Although the observed trend may suggest that the most recently exposed surfaces are more prone to erosion, we note that the apparent decrease in erosion rates with increasing time is common to most techniques that are used to constrain erosion rates (e.g., Koppes and Montgomery, 2009; Ganti et al., 2016). Such a relationship is thought to reflect a bias caused by the stochastic nature of geomorphic processes (e.g., Koppes and Montgomery, 2009; Schumer and Jerolmack, 2009; Sadler and Jerolmack, 2015; Ganti et al., 2016).
A second result is the apparent decrease in erosion rate with elevation (r2 = 0.53; Fig. 4B). The highest erosion rates were observed at the lowest elevation. This decrease in erosion rate is opposite to what is expected for frost cracking for this elevation range (Anderson, 1998; Hales and Roering, 2007). Frost cracking predicts high erosion rates within the frost cracking window (FCW), which would be at high altitude here. Using modern records, a rock surface at 1800 masl spends 8% of the year in the FCW (−3 °C to −8 °C; Matsuoka and Murton, 2008), or 14% and 21% at 2400 and 3200 masl, respectively. This trend also holds for the Younger Dryas–early Holocene transition, assuming the temperature was 4.5 °C less than today (3.6–5.5 °C, Protin et al., 2019) and that the summer-winter difference was similar to today. The observed decrease in erosion rates with elevation is also too pronounced to be explained by chemical weathering alone. Assuming a 3 °C difference over the 451 m of elevation, the difference in reaction rates accommodates only 2% of the observed change (using an activation energy of 60 kJ/mol; Lasaga et al., 1994).
In Figure 4C, we investigated the relationship between the erosion rates and surface slope measured at the outcrop scale. A positive correlation between erosion rates and surface slopes was observed, although the correlation was weaker (r2 = 0.22) than with elevation. This implies that stagnant water on the bedrock surface may not have a primary effect on setting erosion rate. An alternative possibility is snow cover. The surfaces at high elevation with flatter slopes experience higher solid precipitation and periods of snow cover during yearly cycles, which maintain the rock surface at 0 °C, and in turn suppress the efficiency of frost cracking. Surfaces at lower elevations with steeper slopes are less shielded by snow cover and thus experience more time in the FCW and are exposed to more freeze-thaw cycles. The snow cover pattern surely has different effects on TCN and OSL signals. By considering a snow cover of 50 cm for 6 mo per year (Wirsig et al., 2016), the TCN correction accounts for a change of between 5.4% and 6.6% (for MBTP5, 20.2 ± 0.9–21.3 ± 0.9 ka; and MBAM2, 12.1 ± 0.5–12.9 ± 0.5 ka). The snow cover effect for OSL dating is accounted for in the calibration, assuming that calibration surfaces have experienced the same snow cover as the other surfaces (Lehmann et al., 2018).
We also speculate that surfaces in the vicinity of the glacier are influenced by cold katabatic winds coming from the glacier (∼50 m for alpine glaciers; Oerlemans and Grisogono, 2002), which in turn promote frost cracking at lower elevation. A final explanation could be that bedrock surfaces are damaged under the glacier in a way that decreases with the depth in the rock due to the difference in ice load (e.g., Leith et al., 2014). This area of inherited damage facilitates rapid subaerial weathering until much less susceptible rocks with fewer microcracks are encountered, which in turn erode much less rapidly.
We show in Figure 4D that TCN exposure ages increase with elevation (Table DR2). If erosion is ignored, we observe a correlation between age and elevation on the Trélaporte profile with age from 16.6 ± 0.6 ka to 6.6 ± 0.9 ka (MBTP1 and MBTP6, respectively). The outlier MBTP5 (20.2 ± 0.8 ka) is potentially located in a glacial erosion shadow of the topography, exhibiting inheritance from previous exposure. For the Moine profile, the three sites were freed from the glacier at the same time, with noncorrected TCN ages of ca. 12 ka (Fig. 4D; Table DR2). These results agree with known deglaciation scenarios for the Alps (Ivy-Ochs, 2015; Wirsig et al., 2016; Protin et al., 2019). Indeed, ice-surface lowering starting at 16.6 ± 0.6 ka at an elevation of 2550 masl coincides with TCN dating from the southern side of the Mont Blanc massif (Wirsig et al., 2016). However, TCN exposure ages are different when the full range of estimated erosion is included. Any correction of TCN exposure ages older than the corrected age at the highest elevation (16.6 ± 0.6 ka to 134.1 ± 5.8 ka; Fig. 4D; Table DR2) is not physically plausible because a low-altitude surface cannot be exposed before a high-altitude one (gray area in Fig. 4D). In the most extreme scenario, the ages may indicate extremely rapid, near-instantaneous deglaciation at these sites.
By combining OSL and TCN surface exposure dating, we quantified erosion-rate histories of postglacial exposed bedrock in glacial valleys at time scales from 101 to 104 yr. Bedrock surfaces sampled from the flanks of the Mer de Glace revealed erosion rates that vary from 3.5 ± 1.2 × 10−3 mm/yr to 4.3 ± 0.6 mm/yr over 451 m of elevation, with an anticorrelation between erosion rate and elevation. We conclude that a combination of climatic (temperature and snow cover) and morphometric (elevation and surface slope) factors control the variation in erosion rates. Finally, we found that the OSL-derived erosion rates can be used to correct the timing of deglaciation age based on TCN data. Without such a correction, our data would have suggested progressive thinning after 16.6 ± 0.6 ka. In some cases, erosion is so large that one can accommodate a scenario in which thinning was very rapid and thus deglaciation was essentially instantaneous. The difference between these two end-member scenarios leads to important implications for paleoclimate reconstructions, and the potential controls of precipitation and temperature on the regional climate of the western European Alps.
This work was supported by the Swiss-AlpArray SINERGIA project (CRSII2_154434/1) and project PP00P2_170559 (Valla) funded by the Swiss National Science Foundation (SNFS). King acknowledges support from project PZ00P2_167960. We thank P.-H. Blard for sharing the code of the CREp calculator, and D. Six and C. Vincent for GLACIOCLIM Alps data availability. We thank A. Cogez, N. Gribenski, F. Mettra, D. Fabel, and A. Lang for their constructive inputs; S. Coutterand for expertise in the Quaternary of the Mont-Blanc massif and for help during the sampling campaign; and N. Stalder, J. González Holguera, G. Bustarret, U. Nanni, and S. Vivero for their support during field excursions. J. El Kadi, M. Faria, and K. Häring are thanked for laboratory support. We thank K. Hippe, R.S. Anderson, and one anonymous reviewer, whose relevant remarks helped to improve this manuscript.