Spatio-temporal variability and controlling factors for postglacial erosion dynamicsdenudation rates in the Dora Baltea catchment (western Italian Alps)

. Disentangling the influence of bedrock erodibilitylithology from the respective roles of climate, topography and tectonic forcing on catchment denudation is often challenging in mountainous landscapes due to the diversity of geomorphic processes in action and of spatial/temporal scales involved. The Dora Baltea catchment (western Italian Alps) appears the ideal setting for such investigation, since its large drainage system, extending from the Mont Blanc Massif to the Po Plain, cuts across different major litho-tectonic units of the western Alps, whereas this region has experienced relatively homogeneous 15 climatic conditions and glacial history throughout the Quaternary. We acquired new 10 Be-derived catchment-wide denudation rates from 18 river-sand samples collected both along the main Dora Baltea river and at the outlet of its main tributaries. The inferred denudation rate resultsrates vary between 0.2 and 0.9 mm/yr, consistent with previously-published values obtained across the European Alps by previous studies. Spatial variability in denudation rates was statistically compared with topographic, environmental and geologic metrics. 10 Be-derived denudation records dodoes not correlate with the distribution 20 of modern precipitation and rock geodetic uplift. We find, rather, that catchment topography, in turn conditioned by bedrock structuration and erodibility (litho-tectonic origin) and glacial overprint, has the main influence on 10 Be-derived denudation ratespatterns. We calculated the highest denudation rate for the Mont Blanc Massif, whose granitoid rocks and long-term tectonic uplift support high elevations, steep slopes and high relief and thus favour intense glacial/periglacial processes and recurring rock -fall events. Finally, our results, in agreement with modern sediment budgets, demonstrate that the high sediment 25 input from the Mont Blanc catchment dominates the Dora Baltea sediment flux, explaining the constant low 10 Be concentrations measured along the Dora Baltea course even downstream the multiple junctions with tributary catchments.


Introduction
The use of in-situ 10 Be concentrations measured in river sediments to quantify catchment-wide denudation rates over centennial to millennial time scale is now well-established (e.g. Brown et al., 1995;Granger et al., 1996;Bierman and Steig, 1996;Granger and Schaller, 2014). 10Be concentrations are measured at the outlet of the studied basin and are inversely correlated to mean catchment denudation rate (von Blanckenburg, 2005).Widespread research investigation has used this technique to estimate catchment denudation around the globe (see reviews in Portenga and Bierman, 2011;Willenbring et al., 2013;Codilean et al., 2018) and specifically in mountain belts such as the European Alps (Delunel et al., 2020 and references therein), with the aim of illustrating the controlling mechanisms on recent (10 2 -10 5 years) erosiondenudation dynamics and assessing the respective roles of climate, and tectonics or even anthropogenic forcing.
In the European Alps, the large-scale compilation of catchment-wide denudation rates by Delunel et al. (2020) highlighted (1) the first-order correlation between topographicdenudation rate and mean catchment slope (derived from glacial impactimprint on Alpine topography) and denudation rate,), (2) the absence of control ofrelationship between modern climate onand denudation patterns and (3) a significant correlation between rock uplift and denudation for >100-km 2 catchments.This compilation also pointed at a rather weak control of bedrock lithology on denudation, with the lowest rates in the low-elevation foreland areas (with clastic sedimentary lithologylithologies) and highest rates in the high-elevation crystalline parts (with gneissic, granitic or metamorphic lithologies) within the core of the Alps.This trend, however, was not investigated further, since, at the scale of the European Alps, it appeared difficult to disentangle the relative influence of bedrock erodibility, topography and tectonic forcing on denudation rate as these are closely interrelated.Our study thereby aims to further explore the potential links and controls between climatically-driven topography, tectonic uplift and bedrock erodibility on the efficiency of erosiondenudation processes by investigating spatial variability of 10 Be-derived denudation rates within the Dora Baltea (DB) catchment (western Italian Alps; Fig. 1).The DB catchment appears the ideal setting for this investigation, since its large drainage system, extending from the Mont Blanc Massif (4808 m a.s.l.) to the Po Plain (around 200 m a.s.l.), cuts across the main litho-tectonic units of the western Alps (Fig. 2).Relatively similar climatic conditionsgradients and glacial history but variable bedrock lithology and geodetic uplift within the DB catchment and its tributaries allow us to assess how spatial variability in bedrock erodibility between litho-tectonic units may participate in controlling catchment topography and 10 Be-derived denudation rates.

Study area
The Dora Baltea (DB) catchment is a large drainage system of ~3900 km 2 located in the western Italian Alps (Fig. 1).Over a 170-km long distance, the DB river flows NW-SE from the Mont Blanc Massif to the Po Plain, and drains several tributary valleys (13 of which are investigated in this study)tributaries connected to major >4000-m Alpine peaks (e.g.Mont Blanc, Monte Rosa, Matterhorn and Gran Paradiso; Fig. 1).Present-day mean annual temperatures range from -10°C (in highelevation zones) to 15°C (at valley bottoms) within the DB basin (Regione Autonoma Valle d'Aosta, 2009).Precipitations are spatially variable acrossPrecipitation varies between the semi-arid conditions prevailing at low elevations in the DB catchment, with higher central part of the DB valley (mean annual precipitation of 400-500 mm/yr) and the wet conditions in the highelevation internal valleys (Isotta et al., 2014).Higher mean annual precipitation values are observed in the Mont Blanc Massif (around 1800 mm/yr) compared to the north-western and southern sectors of the DB catchment (around 1400 mm/yr for Matterhorn and Monte Rosa area, and around 1150 mm/yr in the Grand Paradiso), and semi-arid conditions prevailing in the central part of the DB valley (mean annual precipitation of 400-500 mm/yr; Isotta et al., 2014).Present-day glaciers cover 3.6% of the total DB area, and are distributed within the upstream high-elevation parts of DB tributary catchments (Fig. 1), with terminus glacier elevations ranging from 2601 to 2800 m a.s.l. in (data from 2005;Diolaiuti et al., 2012).
The geology of the DB catchment is complex, since the DBits drainage network cuts across the main litho-tectonic units of the western Alps, recording the long-term collisional history between the European and Adriatic plates (e.g.Dal Piaz et al., 2008;Perello et al., 2008;Polino et al., 2008;Fig. 2).West of the Penninic Frontal Thrust, the European basement is exposed in the granitoid of the Mont Blanc External Massif and its Helvetic sedimentary cover.Bedrock units belonging to the thinned European crust (gneisses and schists of the Briançonnais basement and its terrigeneous to carbonate metasedimentary cover, high-pressure gneisses of the Internal Massifs), the Tethyan oceanic crust (meta-ophiolite and calcschists of the Piedmont units) and the Adriatic margin (Austroalpine gneisses and eclogitic micaschists) are exposed roughly from NW to SE across the axial belt (, delimited by the Penninic Frontal Thrust to the NW and the Insubric Fault to the SE; (Fig. 2).Long-term (10 6 -10 7 years) exhumation rates estimated from bedrock apatite fission-track datingdata are higher in the western sector of the DB catchment (0.4-0.7 km/Myr for the External zones, west of Internal Houiller Fault; Fig. 2) than in the east (0.1-0.3 km/Myr for the Internal zones, i.e. between the Internal Houiller and the Insubric Faults, Malusà et al., 2005; Fig. 2).Similarly, modern geodetic rock upliftA similar pattern has been illustrated by short-term sediment budgets inferred from detrital apatite fissiontrack (Resentini and Malusà, 2012) and sediment gauge (Bartolini et al., 1996;Vezzoli, 2004;Bartolini and Fontanelli, 2009) data.Modern geodetic rock uplift also appears spatially variable within the DB catchment, with rates up to 1-1.6 mm/yr in the Monte Rosa, Mont Blanc and Ruitor areas, around 0.6-0.7 mm/yr in the axial belt and in the Gran Paradiso Massif, while decreasing to 0.2 mm/yr in the Po plain (Sternai et al., 2019).For the entire DB catchment, a 10 Be-derived denudation rate of 0.6 mm/yr was obtained by Wittmann et al. (2016, sample T12).
The DB catchment was repeatedly glaciated during the Quaternary, with the extensive DB glacial systemmajor glaciers covering most of the catchment with the exception of the highest peaks (~3000 km 2 , >1000 m thick; Serra et al., in revision) abandoningpress) and extending into the Po Plain afterduring the Last Glacial Maximum (LGM, ca. 26-19 ka;Clark et al., 2009), and the).The tributary glaciers already retreated in their upper valley catchments during the Lateglacial climatic oscillations (14-12 ka; Baroni et al., 2021;Serra et al., in revisionpress).As shown by the present-day topography (Fig. 1), postglacial fluvial dissection and hillslope processes following deglaciation have locally re-shaped the glacial landscape, with the development of V-shape valleys profiles and the spread of large alluvial fans and sediment deposits along the main valleys.
Figure 2. Simplified litho-tectonic map of the study area with output catchment-wide denudation rates ((±1σ, mm/yr) reported at sampling locations (catchment boundaries in solid black lines).Major litho-tectonic domains and structural features (dashed lines) of the Westernwestern European Alps are shown (modified map after Resentini and Malusà, 2012).Output catchment-wide denudation rates are corrected for topographic, LIA-glacier and snow shielding, and for quartz-content (see text and Table 1 for details).

10 Be-derived catchment-wide denudation rates
Eighteen river-sand samples were collected within the DB catchment, 5 along the main DB river and 13 at the outlet of the main tributaries (Fig. 1).One sample (DB06) was collected at the same location as sample T12 from Wittmann et al. (2012) to assess for the possible temporal variability of the in-situ 10 Be signal exported by the DB river.Around 20-50 g of pure quartz were extracted from the 250-400 µm grainsize fraction, following sieving, magnetic separation and leaching in diluted HCl, H3PO4 and HF (detailed protocol reported in Akçar et al., 2017).The purified quartz was dissolved in concentrated HF after addition of around 200 μg of 9 Be carrier (Table S1), and Be extraction was performed through anion and cation exchange column chemistry (Akçar et al., 2017).Measurements of 10 Be/ 9 Be ratios were performed at ETH Zürich with the MILEA AMS system (Maxeiner et al., 2019), and normalized to the ETH in-house standards S2007N and S2010N (isotope ratios 28.1 ×10 - 12 and 3.3 ×10 -12 , respectively; Christl et al., 2013).Calculated 10 Be concentrations (Table 1) were corrected using a full process blank 10 Be/ 9 Be ratio of 2.96±0.32×10 -15 .
In order to compute catchment-wide denudation rates, catchment spatiallySpatially-averaged 10 Be production rates were calculated, usingderived from pixel-based calculations conducted with the Basinga 'Production rates' GIS tool (Charreau et al., 2019), withusing a 35-m resolution DEM from Regione Autonoma Valle d'Aosta and Regione Piemonte as input for catchment topography.For production rate calculations, we used the Lal/Stone time-dependent scaling model (Lal, 1991;Stone, 2000), which integrates corrections for atmospheric pressure and geomagnetic field fluctuations according to the ERA-40 reanalysis database (Uppala et al., 2005) and the Muscheler's VDM database (Muscheler et al., 2005), respectively.TheThe total 10 Be surface production rate at each DEM cell of the studied catchments was calculated based on a 10 Be production rate at sea-level and high-latitude (SLHL) of 4.18±0.26at g -1 yr -1 (Martin et al., 2017) considering relative contributions of 98.86, 0.87 and 0.27% by neutrons, slow muons, and fast muons, respectively (Charreau et al., 2019, after Braucher et al., 2011, and Martin et al., 2017).Attenuation length values of 160, 1500 and 4320 g cm -2 were used for neutrons, slow muons, and fast muons, respectively (Charreau et al., 2019, after Braucher et al., 2011), and a rock density of 2.7 g cm −3 was assumed.
Denudation-rate uncertainties (one-sigma external) were estimated only based on values and relative errors of 10 Be concentrations and cosmogenic production rates from neutron and muons (Eq. 5 in Charreau et al., 2019).scaledwith the Lal/Stone time-dependent scaling model (Lal, 1991;Stone, 2000), integrating corrections for atmospheric pressure and geomagnetic field fluctuations according to the ERA-40 reanalysis database (Uppala et al., Integration times associated to the denudation rates (i.e. the time needed to erode the uppermost ~0.6 m of bedrock and ~1 m of soil; von Blanckenburg, 2005) were calculated as mean estimates (no uncertainty propagated).2005)and the Muscheler's VDM database (Muscheler et al., 2005), respectively.
Catchment-averaged production rates were corrected for (1) topographic shielding, (2) quartz-content, (3) LIA-Little Ice Age (LIA, 1250-1860 CE) glacier cover, and (4) snow shielding (Charreau et al., 2019).Catchment topographic shielding was computed with the 'toposhielding' Topotoolbox function (Schwanghart and Scherler, 2014), following the method of Dunne et al. (1999) and Codilean (2006).We acknowledge the recent publication by DiBiase (2018) suggesting no need to correct for topographic shielding when calculating catchment-wide 10 Be denudation rates.Our 10 Be production rates were however corrected for topographic shielding to follow a conservative approach similar to the recent Alpine compilation study by Delunel et al. (2020).As reported in Table 1, mean topographic shielding values obtained within the DB catchment are all very similar (~0.95), implying that neglecting the topographic-shielding correction would result in similar output rates generally within error estimates.
Based on the 1/100,000-and 1/250,000-scale digital geological maps from Regione Autonoma Valle d'Aosta and Regione Piemonte, respectively, we mapped and excluded from the 10 Be production-rate calculation the catchment areas covered by mafic and non-siliceous sedimentary (carbonate) bedrocks (Fig. S1), based on the assumption that they do not provide (or to a minor extent) quartz grains to the fluvial routing system.Crystalline bedrocks and Quaternary deposits (Fig. S1) were instead considered as quartz-bearing lithologies in our approach.In addition, we excluded areas with slope < 3°, assuming that they are likely not linked to the stream network or act as storage/transfer areas and therefore do not reflect catchment denudation (Fig. S1; Delunel et al., 2010).In order to estimate shielding correction due to glacier cover, 10 Be production rates were set to null for areas covered by Little Ice Age (LIA, 1250-1860 CE) glaciers (GlaRiskAlp Project, http://www.glariskalp.eu;Fig. S1), this).This conservative approach assumingassumes sufficient ice thickness for complete cosmic-ray shielding (e.g.Delunel et al., 2010;Wittmann et al., 2007).Shielding correction factors for snow cover were calculated as function of the average elevation for each individual catchment, by applying an empirical model reported in Delunel et al. (2020) that allows to predictand predicting snow-shielding factors as a function of elevation for the European Alps.The obtained average snowshielding correction factors vary between 0.82 and 0.87 and were then combined to the topographic-shielding corrections in a single rasteras scaling factors for the DB each sub-catchment.
Catchment-wide denudation rates were then obtained using the previously-calculated catchment-averaged 10 Be production rates and the measured 10 Be concentrations (Table 1), usingwith the Basinga 'Denudation rates' GIS tool (Charreau et al., 2019).

Topographic, environmental and geological metrics
In order to investigate We investigated potential drivers conditioning the observed spatial variability in DB catchment-wide denudation rates within the DB catchment,, and to this aim we performed topographic analyses, and extracted environmental and geological variables of the quartz-bearing areas (Fig. S1) for each investigated tributary catchment through an ArcGIS-Matlab routine (Delunel et al., 2020).
Topographic analyses were conducted using a 35-m resolution DEM (Regione Autonoma Valle d'Aosta and Regione Piemonte).We calculated drainage area, mean elevation, mean slope, percentage of slopes steeper than 40°, geophysical relief, and hypsometric integral for eachthe quartz-bearing areas of the individual catchmentcatchments (Table 2).For slope analyses, the 'gradient8' Topotoolbox function was used (Schwanghart and Scherler, 2014), returning the steepest downward gradient of the 8-connected neighbouring cells of the DEM.The percentage of catchment slope steeper than 40° was calculated as indicative of the areal proportion of oversteepened threshold landscape (DiBiase et al., 2012).The geophysical relief (i.e.averaged elevation differences between a surface connecting highest topographic points and the current topography; Small and Anderson, 1998) was calculated in ArcGIS using a 5-km radius sampling window, and can be used as an indicator of past landscape change (i.e.high geophysical relief may indicate increased relief fromor potential for locally increased erosion; (see Champagnac et al., 2014 for discussion).The hypsometric integral was computed based on Eq. 1 from Brocklehurst and Whipple (2004) and is inversely related to the stage of landscape evolution (i.e. more evolved landscapes, whose high-elevation areas have been eroded, have lowerpresent low hypsometric integrals).
In addition, we extracted catchment-averagedaverage values of the following environmental variables.Averaged for the quartz-bearing areas of each individual catchment.Average annual precipitation for each catchment was obtained from the 5km resolution grid of mean annual Alpine precipitation from Isotta et al. (2014), in order to investigate the potential influence of modern precipitation/runoff on erosiondenudation dynamics.Percentage of bare-rock area was estimated from the extent of class 30 ("bare bedrock") of the 100-m resolution CORINE Land Cover Inventory (2018), to consider if catchment areas with null to low soil/vegetation cover are more subjectsubjected to physical erosion and chemical weathering.LIA-glacier areal cover was calculated based on the LIA-glacier extent mapped within the GlaRiskAlp Project (http://www.glariskalp.eu), in order to investigateassess the influence of modern to historical glacial and periglacial processes on output 10 Be-derived denudation rates (Delunel et al., 2010).Mean LGM ice-thickness and areal percentage of each catchment above the LGM Equilibrium Line Altitude (ELA) were estimated by using the LGM paleo-glacier reconstruction of the DB system (70-m resolution, LGM ELA at 2103 m a.s.l.; Serra et al., in revision), bothpress).Both metrics potentially givinggive indication on the LGM glacial imprint on topography and subsequent potential for postglacial erosion response (Norton et al., 2010;Salcher et al., 2014;Delunel et al., 20202014;Delunel et al., 2020).Catchment average temperature was not estimated since, at the relatively constant latitude of the investigated catchments, temperature variability directly follows catchment hypsometric distribution and thus relates to catchment elevation which is already investigated in the present study (topographic metric).
Lastly, we extracted geological variables for the studied catchments.Based on the simplified litho-tectonic map of the DB catchment (Fig. 2), modified after Resentini and Malusà (2012), we estimated the relative proportion of the different lithotectonic units within the quartz-bearing areas of each catchment.Catchment-averagedaverage geodetic uplift rates were as well considered using the 30-km resolution interpolation grid from Sternai et al. (2019), here downscaled to 600-m resolution grid (Delunel et al., 2020).

Spatial variability in catchment-wide denudation rates
Calculated catchment-wide 10 Be production rates and derived denudation rates vary according to the applied production-rate correction factors (Table S2).Uncorrected denudation rates (i.e.including only mean catchment topographic shielding and excluding areas with slope <3°) range between 0.27±0.02and 1.49±0.13mm/yr, while rates obtained by applying all corrections vary between 0.21±0.02and 0.91±0.08mm/yr (Table 1 and Fig. 2).Significant production-rate corrections were obtained when taking into account snow shielding and LIA-glacier cover (up to 17 and 42% reduction compared to uncorrected 10 Be production/denudation rates, respectively), especially for catchments with high mean elevations and associated highlarge LIA-glacier coverage (Table 2).Lower corrections were obtained when considering quartz-contentbearing areas (maximum 10% reduction in output 10 Be production/denudation for catchments DB08 and 11, where relatively abundant sedimentary and 235 mafic bedrocks occur; Fig. S1).All corrections combined together lead to reduction in 10 Be production/denudation rates of 16-53% compared to the uncorrected estimates.(Table S2).Hereafter, we consider 10 Be production/denudation rates obtained by applying all corrections (Table 1 and Fig. 2), in order to maintain a conservative approach as in the recent Alpine compilation study (Delunel et al., 2020) Be measurements were calibrated against ETH in-house standards S2007N and S2010N (isotope ratios 28.1 x 10 -12 and 3.3 x 10 -12 , respectively; Christl et al., 2013).Calculated 10 Be concentrations were corrected for full process blank 10 Be/ 9 Be ratio of 2.96 ±0.32 × 10 −15 .Additional analytical data are reported in Table S1.b Catchment topographic shielding was computed with the 'toposhielding' Topotoolbox function (Schwanghart and Scherler, 2014, after Dunne et al., 1999and Codilean, 2006).
c Catchment-averaged 10 Be production rates were calculated with Basinga (Charreau et al., 2019), based on SLHL total 10 Be production rate of 4.18±0.26at g -1 yr -1 (Martin et al., 2017) 2016) is also shown for discussion.Samples DB03 and DB14, 14 and 18 are omitted since they are in turn tributaries of catchments DB04 and DB13, respectively, and do not directly connect to the main DR river.
In order to quantify at first-order the relative contribution of the Mont Blanc Massif (represented by the lowest 10 Be concentration of sample DB02; Fig. 3) to the 10 Be signal measured along the DB river, we followed the approach reported in Delunel et al. (2014).River-sediment 10 Be concentrations from tributaries and along the DB river have been first normalised to the SLHL 10 Be production rate (i.e.4.18±0.26at g -1 yr -1 ), implying that variations in normalised 10 Be concentrations represent the variability in denudation rates only.We then estimated the respective contributions of the Mont Blanc Massif and different tributaries through a mixing model (mass-balance model involving catchment 10 Be concentrations and contributing areas; Delunel et al., 2014) considering (A) the normalised 10 Be concentration for river materials exported from the Mont Blanc catchment, (B) the averaged normalised 10 Be concentration from the upstream tributaries contributing to each sampling points along the main DB river and (C) the normalised 10 Be concentration at the sampling points along the main DB river (DB12, 10, 06).Between our two most upstream DB river samples (DB01 and DB02), we based our model on DB02, which provides a more conservative estimate of the contribution of the Mont Blanc Massif to the 10 Be signal measured along the DB river (i.e. the potential contributions of the tributaries are maximized).By applying this simple model, we find that the Mont Blanc Massif (upstream catchment DB02) contributes to 90, 87 and 77% of the river-sediment 10 Be signal measured respectively at locations DB12, 10 and 06, in line with overall constant 10 Be concentrations measured along the DB course (Fig. 3).

Catchment metrics and denudation rates 290
Results of catchment topographic analyses, along with estimates of environmental and geological metrics, are reported in Table 2.As for calculated catchment-wide denudation rates, DB01 (Mont Blanc Massif) also appears as an end-member with maximum values in most of the reported metrics (   Sternai et al. (2019).riverlocations for 10 Be analysis).Names of sample collected along the main DB river are given in italics.With the exception of the total drainage area, all the metrics were calculated using the quartz-bearing area of each catchment.See text for details.
We compared the 10 Be-derived catchment denudation rates against topographic, environmental and geological metrics and evaluated the statistical significance of investigated linear correlations (p-value and R2r 2 ; Figs. 4 and 5, Table S3).Samples along the main DB river (downstream of DB01, i.e.DB02, 12, 10, 06; Fig. 2) were excluded from the investigated correlations since their apparent denudation rates are potentially affected by cumulative drainage and sediment mixing along the DB course.
Correlations were calculated both including and excluding sample DB01,.Cook's distance values were also calculated in order to assess whether DB01 strongly influences the derived correlations as a potential outlier.SignificantWe selected a threshold value of 3 times for DB01 Cook's distance compared to the data mean Cook's distance.As a consequence, DB01 appears as an outlier in different investigated correlations between catchment denudation rates and topographic, environmental or geological metrics (Table S3).Nevertheless, significant linear correlations (i.e.p-value <0.05) both with and without DB01 were obtained between catchment denudation rates and topographic metrics, including mean elevation (Fig. 4A) and), 5-km geophysical relief (Fig. 4C), the relative abundance of bare bedrock (Fig. 5B), and the percentage of area above the LGM ELA (Fig. 5D).4C).Significant linear correlations between catchment denudation rates and mean slopes (Fig. 4B) or), proportions of oversteepened slopes (Table S3) and relative area covered by LIA glaciers (Fig. 5C) were instead only found when including DB01.In addition, we found statistical linearNon-significant correlations for environmental metrics such as the relative abundance of bare bedrock (Fig. 5B), and the percentage of area covered by LIA glaciers (only including DB01; Fig. 5C(pvalue ≥0.05) were observed between catchment denudation rates and drainage areas (Table S3), hypsometric integrals (Table S3), mean annual precipitation values (Fig. 5A), mean geodetic uplift rates (Fig. 4D) and mean LGM ice-thickness (Table S3).S3).
Finally, only weak linear correlations (p-value ~0.06-0.08,including DB01) can be observed between catchment denudation rates and LGM glacial metrics (mean LGM ice-thickness and catchment proportion above LGM ELA, Table S3 and Fig.  ) are reported for each plotlinear regression with significant trendstatistical significance (p-value < 0.05).R 2 r 2 is not reported for nonsignificant correlations (p-value > 0.05).Asterisks indicate linear correlations for which DB01 has been considered as an outlier (based on Cook's distance, Table S3).

Litho-tectonic units and denudation rates
In addition to catchment metrics, we explored the potential influence of bedrock properties on the efficiencypostglacial evolution of geomorphic processes andthe DB catchment denudation rates by analysing the correlation between tributarycatchment denudation rates and the spatial distribution of litho-tectonic units within the DB area (Figs. 2 and 6; only quartzbearing areas considered, Fig. S1).The highest denudation rates are observed for tributaries with widespread bedrock exposure of granites of the Mont Blanc External Massif and its Helvetic terrigeneous to carbonate sedimentary cover (8285%; DB01: 0.68±0.05mm/yr), or with ).Moderate denudation rates around 0.4-0.5 mm/yr are observed for catchments with abundant gneisses of the Gran Paradiso Internal Massif (40-7041-73%; DB03, 04, 05: average denudation rate of 0.51±0.02mm/yr).
Formatted: Font: Not Bold   S2, the different correction factors for quartz-content, LIA-glacier cover and snow-shielding lead to 16-53% decrease in catchment 10 Be production and inferred denudation rates compared to uncorrected estimates (i.e.including only catchment-averaged topographic shielding and excluding areas with slope <3°).Such correction factors build on several assumptions and have different implications for our catchment-wide denudation rate results that are discussed hereafter.
However, we should also note that the investigated correlations between denudation rates and topographic, environmental and geological metrics (Fig. 4-5) remain similar when using non-corrected denudation rates.
First, assuming Quaternary deposits as quartz-bearing lithologies is a first-order approximation, since deposits derived from mafic and carbonate-sedimentary bedrocks would bring no or minor quartz to the sediment routing system.For some tributaries dominated by these lithologies (DB09, 11 and 16, as reported in section 4.3), we nevertheless considered that Quaternary deposits may bear quartz given than the upper part of their catchments drain crystalline bedrock (Fig. S1).However, distinguishing deposit provenance/lithology in this Alpine environment, with complex glacial/periglacial systems, would require detailed field investigation and mapping, which is beyond the scope of this work.Moreover, our calculations show that correction for quartz-contentbearing area has only a minimal effect on catchment-averaged 10 Be production and denudation rates, with only up to 10% difference between uncorrected and corrected results thus overlapping within uncertainties.(Table S2).
Second, correction factors for LIA-glacier cover and snow shielding lead instead to significant decrease in catchment-averaged 10 Be production and thus denudation rates (up to 42 and 17%, respectively).Since sediments in sub-/proglacial environments can derive from periglacial erosion from bedrock walls/peaks and/or re-mobilization of previously exposed material (with nonzero 10 Be concentration, e.g.moraine deposits; Wittmann et al., 2007;Delunel et al., 2014;Guillon et al., 2015), assuming null 10 Be concentration input from areas covered by LIA glaciers might lead to overcorrections of our denudation rate resultsestimates.Uncertainties are related also to the snow-shielding correction approach.The snow-shielding vs. elevation model reported by Delunel et al. (2020) has been calibrated on snow-water equivalent records of the Swiss and French Alps, which are wetter regions compare to the DB catchment (Isotta et al., 2014).Therefore, LIA-glacier cover and snow-shielding corrections may be overestimated for the DB catchments, especially for high-elevation tributaries.In particular, catchment DB01 shows the maximum corrections for both LIA-glacier cover and snow shielding (42 and 17% respectively; Table S2) and consequently relatively low output denudation rate compared to estimates obtained for catchments downstream along the main DB river (DB02, 12, 10; Fig. 2), despite similar 10 Be concentrations (Fig. 3).
We therefore acknowledge that our corrected catchment-averaged 10 Be production and denudation rates (Table 1 and Fig. 2) should be considered as minimum estimates, given the correction factors for LIA-glacier cover and snow shielding, in line with the recent compilation over the entire European Alps (Delunel et al., 2020).

Controlling factors and processes on 10
Be-derived catchment denudation ratesFinally, we need to assess the impact of LGM glacial erosion on our 10 Be-derived denudation rates (Glotzbach et al., 2014;Dixon et al., 2016), since our study area has been largely glaciated during the LGM (Serra et al., in press).Deep glacial erosion may have largely to completely zeroed 10 Be concentration on bedrock surfaces, with non steady-state 10 Be concentration depth profiles during postglacial surface exposure leading to apparent overestimate in denudation rates from 10 Be concentrations in river sands (Glotzbach et al., 2014).However, given the deglaciation history of the DB catchment (i.e.largely deglaciated by 14-12 ka; Baroni et al., 2021;Serra et al., in press) and the range of our 10 Be-derived denudation rates (0.2-0.9 mm/yr, Table 1 and Fig. 2), we can estimate an overestimate of our 10 Be-derived denudation rates by maximum 10-15%, similar to the proposed estimate of Dixon et al. (2016) in the Eastern Alps, with a maximum 9% overestimate for slower 10 Be-derived denudation rates (~0.2 mm/yr).We thus are confident in the validity of our 10 Be-derived denudation rates (Table 1), and can exclude any potential strong bias influencing the spatial pattern (Fig. 2) and interpretation with regards to topographic, environmental and geological metrics (Figs.4-5).

Propagation of 10 Be signal along the DB course
Our results highlight the strong 10 Be-dominance of the Mont Blanc Massif (represented by sample DB02, see section 4.1 for discussion) on downstream sediment samples collected along the DB course, below the tributary junctions (Fig. 3, Table 1).
The relatively constant low 10 Be concentrations measured for samples DB01, 02, 12, 10, 06 (around 1.2 x10 4 at/g, Fig. 3) compared to the tributaries (2.0-4.9 x10 4 at/g), and the outcomes of our mixing model indicate unequal sediment contribution (non-balanced sediment budget; Savi et al., 2014) between the main DB stream and its tributaries.The Mont Blanc Massif appears to govern the sediment yield along the main DB river, contributing to >77% of the river-sediment 10 Be signal carried all along the DB river.
A key factor governing the mixing and flux balance of 10 Be concentrations between river streams is the quartz flux from each stream, which is in turn influenced by (1) catchment denudation rate, (2) drainage area, (3) catchment quartz content (Carretier et al., 2015), (4) sediment storage (e.g.dams, lakes, floodplains reducing mass flux but not changing the 10 Be concentration; Wittmann et al., 2016).Our results show significantly higher denudation rate for catchment DB01 compared to other DB tributaries (Fig. 2, Table 1).While the Mont Blanc Massif (upstream DB02 catchment) represents only a minor fraction of the total DB catchment area (~18%), its quartz-bearing surface area appear 5-90% larger than for other tributaries (Table 2).
Likewise, the sediment-provenance studies of Vezzoli et al. (2004) and the sediment-yield estimates of Vezzoli (2004) highlighted that river sands from the Mont Blanc catchment (analogous catchment to DB02) have up to ~20% higher quartz content compared to some other DB tributaries (analogous to DB09, 11, 16; Table S4) and contribute to ~62% of the quartz flux of the entire DB catchment (analogous to DB06; Table S4).Since the occurrence of dams is limited to few catchments (Fig. 1), the high quartz flux and 10 Be-signal dominance of the Mont Blanc Massif along the DB course could derive from (1) its high denudation rate (Fig. 2 and Table 1), (2) its large quartz-bearing drainage area and (3) the high quartz content of the Mont Blanc granitoid (Vezzoli, 2004).Between this three potential causes, we propose that the 10 Be-signal dominance of the Mont Blanc Massif along the DB course is mainly driven by its high denudation rather than quartz fertility or area coverage, as illustrated by the similar trend of modern denudation rates derived from sediment gauging (Hinderer et al., 2013; see also discussion in section 5.4).The high rock-slope instability and glaciogenic sediment production in the Mont Blanc Massif supply abundant low 10 Be concentration quartz to the river system, being therefore efficient in diluting the 10 Be concentration in the downstream course of the DB river.Controlling factors explaining the high denudation rate of the Mont Blanc Massif are further discussed below (section 5.3).
For the entire DB catchment, we can note that the 10 Be concentration is ~30% higher for sample T12 (1.99±0.14x10 4 at/g; Wittmann et al., 2016) compared to DB06 (1.52±0.08 x10 4 at/g; Fig. 3), both collected at the same location (DB catchment outlet, Fig. 1) but at different time periods.The observed difference is probably related to a stochastic change in sediment sources (Lupker et al., 2012), with potentially a temporary dominant sediment input from a DB tributary catchment with higher 10 Be concentration (e.g.DB07, close location and similar 10 Be concentration as T12; Fig. 3) than from the Mont Blanc catchment.By comparing our results to the Po catchment (Wittmann et al., 2016), which drains several main river systems from the south-western Alps in addition to the DB river basin, it emerges that the low 10 Be concentration signal deriving from the Mont Blanc Massif, remaining overall constant along the DB course, increases significantly soon after the DB flows into the Po river.The high 10 Be concentrations measured by Wittmann et al. (2016) in Po river-sediment samples, immediately downstream the DB confluence (samples P1 and P3: around 3.6 x10 4 at/g), show that the Po river is dominated in its initial lowland flow by the high 10 Be concentration inputs from other south-western Alpine catchments (Wittmann et al., 2016).

Controlling factors and processes on 10 Be-derived catchment denudation rates
Our 10 Be-derived denudation rates, varying between 0.2 and 0.9 mm/yr, fit broadly within the values obtained over the European Alps, where 95% of the considered catchments yield denudation rate values <1.2 mm/yr and rates for the Western European Alps range between 0.1 and 1.2 mm/yr (Delunel et al., 2020).Correlations with topographic, environmental and geologic metrics allowed us to identify and discuss potential controlling mechanisms for denudation-rate variability within the DB catchment, that we discusspresent here in comparison with studies conducted in other Alpine sectors.
While precipitation and rock uplift have been recognized as main drivers for Alpine denudation rates, especially for the Central Alps (e.g.Chittenden et al., 2014;Wittmann et al., 2007, respectively), their respective influence on denudation-rate variability within the DB catchment is not significant (Figs.5A and 4D).Interestingly, it can be observed that catchment geodetic rock uplift is overall higher (20-80%) than 10 Be denudation in all the investigated tributary catchments, suggesting a net surface uplift of the DB area for recent timescales, in line with other observations across the European Alps (Norton et al., 2011;Delunel et al., 2020).
Catchment topography, in turn conditioned by both bedrock erodibility and glacial overprint, appears instead to have a major role in controlling the observed spatial variability in DB denudation rates.DenudationFirst, denudation rates are indeed positively correlated with catchment-averaged elevation and 5-km geophysical relief (Figs.4A-C), and to a lesser extent with catchment averaged slopes (i.e. when DB01 is included; (Fig. 4B), similarly to what has been identified by previous studies.
First, elevation4A).Elevation influences denudation rates through periglacial (i.e.frost-cracking; Delunel et al., 2010) and glacial erosive processes, both increasing with elevationaltitude due to their temperature dependency, as well as by modifying soil and vegetation cover, with bare-rock exposure being positively correlated with denudation rates (Fig. 5B).Second, correlations with slope and geophysical relief need to be considered (Figs.4B and C).It has been previously proposed that topographic slope and geophysical relief are positively correlated to catchment denudation until a threshold slope angle of 25-30° (Montgomery and Brandon, 2002;Champagnac et al., 2014;Delunel et al., 2020).Below this threshold, denudation was shown to respond to a slope-dependent equilibrium between regolith cover production through weathering and its downslope diffusion.In oversteepened catchments, denudation rates are instead controlled by mass wasting processes (i.e.rockfalls, debris flows, landslides) which stochastically influence river-sediment 10 Be concentrations.All the DB tributaries catchments have average slope comprised in the threshold range of 25-30°, with the exception of DB01 whose average slope is higher than 30°.While the potential effect of slope alone is here challenging to evaluate as all the tributaries exhibit similar averaged slope values, between ~25-30° (with the exception of DB01 with average slope of ~32°, Fig. 4B), denudation rate exhibits a clear correlation with geophysical relief (Fig. 4C), which is function of both slope and elevation difference (Small and Anderson, 1998;Champagnac et al., 2014).We suggest that slope differences between the investigated catchments, while not significant, are nevertheless still close to threshold values (Fig. 4B; Delunel et al., 2020), which, when combined with elevation differences between catchments, would explain the significant relationship observed between geophysical relief and denudation rates (Fig. 4C).
Our results also show a correlation between catchment denudation rates and bedrock litho-tectonic classification (Fig. 6), which has been proposed to govern erosiondrive denudation through rock mechanical strength (erodibility; Kühni and Pfiffner, 2001).Similar to what has been suggested forbased on DB modern sediment provenance (Vezzoli et al., 2004), we observe a counter-intuitivegeneral trend with the highest denudation rates in catchments dominated by apparent "low erodibility" bedrocks (granite and gneiss), and the lowest rates in catchments with apparent "high erodibility" bedrocks (carbonatesedimentary and terrigeneous rocks; erodibility classes according to Kühni and Pfiffner, 2001).This trend has already been observed locally in the Eastern and Southern Alps (Norton et al., 2011) as well as at the scale of the entire European Alps (Delunel et al., 2020).Such observations were interpreted to be related to the influence of bedrock resistance on catchments morphometry (in turn connected to erosiondenudation dynamics), with the most resistant lithologies located at highest elevations and sustaining the steepest slopes/highest reliefs (Kühni and Pfiffner, 2001;Stutenbecker et al., 2016).To test this hypothesis at the scale of the DB catchment, we evaluated the distribution of elevation, slope and 5-km geophysical relief for each individual litho-tectonic unit (Fig. 7).While the slope distributions appear similar for all the different lithotectonic domains (median of 26-31°; Fig. 7B), higher elevations and reliefs are observed for the External and Internal Massifs (median elevation of 2500-2700 m a.s.l., median relief of around 2016).Our results are in line with this interpretation, with the "low-erodibility" granite of the Mont Blanc External Massif supporting the highest elevation and reliefs and slightly steeper slopes (Fig. 7), where efficient geomorphic processes promote the highest catchment denudation rate (Fig. 6).On the other hand, the "high erodibility" rocks of the Briançonnais cover and of the Piedmont units present low elevation, relief and slope values, and are associated with low denudation rates.High elevation sustained by gneisses and granite of the Internal Massifs (2700 m a.s.l.; Fig. 7A) and slightly steeper slopes supported by gneisses and micaschists of the Austroalpine units and of the Briançonnais basement (30-31°; Fig. 7B) would also drive the moderate denudation rates observed in these three litho-tectonic domains (Fig. 6).Moreover, the different long-term tectonic histories between1800 m; Fig. 7A) compared to the other lithotectonic units (median elevation of 2000-2200 m a.s.l., median relief of 1000-1200 m; Fig. 7C).We tentatively suggest that the lithological control on DB denudation-rate variability (Fig. 6) is connected to the influence of bedrock erosional resistance on topography, with "low-erodibility" rocks supporting high-altitude and high-relief catchments where erosion processes' efficiency promote high catchment denudation rates (Figs.4A and C).Moreover, the different long-term tectonic histories of the litho-tectonic domains could also explain some of the observed variability in catchment denudation between areas west and east of the Penninic Frontal Thrust (Fig. 2).Bedrock tectonic fracturing (Molnar et al. 2007) may influence subsequent erodibility and denudation, facilitated by the exhumation of more fractured bedrock units such as the crystalline units of the Mont Blanc External Massif and its Helvetic sedimentary cover (no deep Eocene subduction during Alpine orogeny), compared to deeply-exhumed rocks of the Internal Massifs and Piedmont units (Schmid et al., 2004).Additionally, higher late-Miocene uplift rates in the Mont Blanc Massif compared to the rest of the DB catchment (Malusà et al., 2005) could have sustained high-elevations in the Mont Blanc Massif, which in turn would also promote efficient geomorphic processes and high denudation rates.
Lastly, we consider the potential connection between landscape glacial imprint and catchment denudation rates.Our correlations Statistically significant correlation between catchment denudation rates and LGM glacial metrics (mean LGM icethickness and catchment area proportion above LGM ELA, Table S3 and (Fig. 5D) appear non-significant, suggesting no direct controlsuggests an impact of LGM glacial metrics/large Quaternary glaciations on our calculated 10 Be-derived denudation rates.However,Most of the catchments have >50% of their area above the LGM-ELA, indicating that large glaciers persisted during the LGM (and potentially older Quaternary glacial stages), with a significant impact on catchment topography characterized by steep Alpine slopes and high reliefs are clear witnesses of Quaternary glacial erosion,(Fig.4A-C , Pedersen and theirEgholm, 2013).We hence tentatively interpret the significant correlation tobetween denudation rates and high elevation / pronounced geophysical relief (Figs.4B and C) is thereforeto be indicative of a long-term glacial topographic control on the postglacial erosional response, as suggested by previous studies (Norton et al., 2010a;Glotzbach et al., 2013;Dixon et al., 2016).The glacial pre-conditioning of the topography has been also enhanced during postglacial times with coupled fluvial incision and hillslope processes increasing Alpine valley slopes and reliefs locally (Korup and Schlunegger, 2007;Valla et al., 2010;van den Berg et al., 2012).Over thea shorter term, the positive correlation between catchment denudation rates and LIA glacial cover (only when including DB01, Fig. 5C) suggests also an important role of Holocene to modern glacial processes in influencing catchment denudation, by contributing to high-sediment delivery (Stutenbecker et al., 2018).By considering the above-mentioned controlling mechanisms for catchment denudation, we propose anthe following interpretation for the high denudation rate obtained for catchment DB01 compared to other DB tributaries (FigFigs.2).Such Formatted: Font: 9 pt Formatted: Font: 9 pt Formatted: Font: 9 pt an observation has already been suggested based on modern sediment provenance (Vezzoli et al., 2004;Vezzoli, 2004 and 4-5).Catchment DB01 has maximum values for most of the investigated metrics (Figs. 4 and 5, Table 2).Its location in the highelevation core of the Alps (Mont Blanc Massif, long-and short-term high uplift rate) was the site of intense Quaternary glaciations (large catchment area above the LGM ELA), which deeply modified the landscape as illustrated by the high geophysical relief of this catchment.Thanks to the highly-resistant granitoid lithology, steep slopes and high reliefs deriving from glacial erosion could be maintained, in turn promoting high millennial to present-day denudation rates in this catchment.
Finally, the supply of sediments by retreating glaciers and active periglacial processes, and the contribution of frequent rockfall events triggered by abundant precipitations (Fig. 5A) and present-day permafrost degradation (Ravanel et al., 2010;Akçar et al., 2012;Deline et al., 2015) participate to the significant sediment yield in the DB01 catchment.ItCatchment DB01 thus supplies material with highly depleted 10 Be concentrations to the river system, which is in turn capable ofto significantly dilute the 10 Be signal along the DB course (Fig. 3, see following section for discussion3).

Propagation of 10 Be signal along the DB course
Our results highlight the strong 10 Be-dominance of catchment DB01 on downstream sediment samples collected along the DB course, below the tributary junctions (Fig. 3, Table 1).The constant low 10 Be concentrations measured for samples DB01, 02, 12, 10, 06 (around 1.2 x10 4 at/g) indicate unequal sediment mixing (non-balanced sediment budget; Savi et al., 2014) between the main DB stream and the tributaries ( 10 Be concentrations of 2.0-4.9 x10 4 at/g).
A key factor governing the mixing and flux balance of 10 Be concentrations between river streams is the quartz flux from each stream, which is in turn influenced by (1) catchment denudation rate, (2) drainage area, (3) catchment quartz content (Carretier et al., 2015), (4) sediment storage (e.g.dams, lakes, floodplains reducing mass flux but not changing the 10 Be concentration; Wittmann et al., 2016).Our results show significantly higher denudation rate for catchment DB01 compared to other DB tributaries (Fig. 2, Table 1).Moreover, the sediment-provenance study of Vezzoli (2004) highlighted that river sands from the Mont Blanc catchment (analogous catchment to DB01) have up to ~20% higher quartz content than other DB tributaries.Since the upstream catchment area of DB01 is comparable to most other DB tributaries (Table 2), and the presence of dams is limited to few catchments only (Fig. 1), we propose that the 10 Be-signal dominance of DB01 along the DB course comes from (1) its high denudation rate (Fig. 2 and Table 1) and ( 3) the high quartz content of the Mont Blanc granitoid (Vezzoli, 2004).The high rock-slope instability and glaciogenic sediment production in the Mont Blanc Massif supply low 10 Be concentration material to the river system, and are therefore efficient in diluting the 10 Be concentration in the downstream course of the DB river.
Following the approach reported in Delunel et al. (2014), we can obtain a first-order estimate of the relative contribution of the Mont Blanc Massif to the river-sediment 10 Be signal measured along the DB river.River-sediment 10 Be concentrations from tributaries and along the DB are first normalised to the SLHL 10 Be production rate (i.e.4.18±0.26at g -1 yr -1 ), implying that variations in normalised 10 Be concentrations represent the variability in denudation rates only.We then estimate the respective contributions of the Mont Blanc Massif and different tributaries through a mixing model considering 1) the normalised 10 Be concentration for river materials exported from the Mont Blanc catchment and 2) the averaged normalised 10 Be concentration from the upstream tributaries contributing to each sampling points along the main DB river.Between our two most upstream DB river samples (DB01 and DB02), we base our model on DB02, which provides a more conservative estimate of the contribution of the export from Mont Blanc catchment (i.e. the potential contributions of the tributaries are maximized while that of Mont Blanc catchment is minimized).By applying this simple model, we find that the Mont Blanc catchment contributes to >77% of the river-sediment 10 Be signal carried all along the DB river, while it only represents around 15% of the total DB catchment area (i.e. at DB06 sampling point).This first-order estimate further exemplifies the significant role of the Mont Blanc Massif in governing the sediment yield along the main DB river.
For the entire DB catchment, we can note that the 10 Be concentration is ~30% higher for sample T12 (1.99±0.14x10 4 at/g; Wittmann et al., 2016) compared to DB06 (1.52±0.08 x10 4 at/g; Fig. 3), both collected at the same location (DB catchment outlet, Fig. 1) but at different time periods.The observed difference is probably related to a stochastic change in sediment sources (Lupker et al., 2012), with temporary dominant sediment input from a DB tributary catchment with higher 10 Be concentration (e.g.DB07, close location and similar 10 Be concentration as T12; Fig. 3) than from the Mont Blanc catchment.
By comparing our results to the Po catchment (Wittmann et al., 2016), which drains several main river systems from the southwestern Alps in addition to the DB river basin, it emerges that the low 10 Be concentration signal deriving from the Mont Blanc Massif, and remaining overall constant along the DB course, increases significantly soon after the DB flows into the Po river.
The high 10 Be concentrations measured by Wittmann et al. (2016) in Po river-sediment samples, immediately downstream the DB confluence (samples P1 and P3: around 3.6 x10 4 at/g), show that the Po river is dominated in its initial lowland flow by the high 10 Be concentration inputs from other south-western Alpine catchments (Wittmann et al., 2016).

Long-and short-term DB denudation rates
The general trend emerging from ourOur 10 Be-derived denudation rates, show a general trend of higher millennial denudation rates in the Mont Blanc Massif compared to other DB tributaries (Fig. 2), which is overall in agreement (, albeit with different absolute values), with erosiondenudation rate estimates on different timescales.
Long-term (10 6 -10 7 yr) exhumation rates estimated from bedrock apatite fission-track datingdata (Malusà et al., 2005) show higher values (0.4-0.7 km/Myr) in the western sector of the DB catchment (west of Internal Houiller Fault, Fig. 2) than in the east (around 0.2 km/Myr, between the Internal Houiller and the Insubric Faults, Fig. 2).Likewise, results from detrital apatite fission-track dating (Resentini and Malusà, 2012) indicate that short-term (10 2 -10 5 years) erosiondenudation rates are higher (around 0.5 mm/yr) in the Mont Blanc External Massif and its sedimentary cover (west of the Penninic Frontal Thrust) than in the axial belt, east of the Penninic Frontal Thrust (around 0.1 mm/yr; Fig. 2  Be-and modern river yield-derived catchment -wide denudation rates.For all tributary catchments and DB locations, modern denudation rates are derived from river-bedload empirical estimates of river bedload are available (Vezzoli et al., 2004; red and yellowlight blue dots for DB and tributary catchments, respectively).For catchment DB06 and DB011, also modern denudation rates derived respectively from sediment gauging and sediment trapping are plotted inwith green diamonds (Hinderer et al., 2013, after Bartolini et al., 1996and Bartolini and Fontanelli, 2009).Errors are represented only for 10 Be-derived denudation rates, since they are not reported for the modern rates.
Modern denudation rates, obtained from sediment-yield estimates (all DB catchments; Vezzoli et al., 2004;based on Gavrilovic empirical formula, Gavrilovic, 1988) and measurements (sediment gauging for DB06 and sediment trapping for DB11; Hinderer et al., 2003, after Bartolini et al., 1996and Bartolini and Fontanelli, 2009), display higher values for samples along the DB (0.07-0.73 mm/yr) compared to DB tributaries (0.01-0.08 mm/yr).While such a pattern is consistent with our 10 Be-derived records, millennial denudation rates are 2 to 50 times greater than modern denudation ratesestimates, with the exception of sample DB01 for which modern and 10 Be-derived denudation rates are roughly similar (Fig. 8).Equivalent order of discrepancy between modern sediment-yield and 10 Be-derived denudation rates has been observed by several studies (e.g.Kirchner et al., 2001;Schaller et al., 2001;Wittmann et al., 2007Wittmann et al., , 2016;;Stutenbecker et al., 2018).Among other factors,; Pitlick et al., 2021).Among other factors (including the exclusion of non quartz-bearing areas for 10 Be denudation rates, see Table 2 and discussion in section 5.1), this discrepancy was interpreted to point to the separate or combined effects of (1) incorporation of high-magnitude low-frequency erosion events in the 10 Be-derived but not in the sediment-yield denudation rates of high-magnitude low-frequency erosion events, (2) contribution of bedload and carbonate dissolutionchemical weathering to 10 Be-derived but not to sediment-yield denudation rates, (3) linear dissection of the landscape by fluvial erosion and subglacial sediment export, leading to preferential postglacial erosion of material with low 10 Be concentration, overestimatingincreasing 10 Be-derived denudation rates, through fluvial linear dissection of the landscape and subglacial sediment export, (4) sediment traps (e.g.lakes, dams), changing the flux measured by sediment gauging but less probably the 10 Be concentrationconcentrations which isare averaged over longer timescales.The first and third hypotheses could be the most plausible for our results.Modern denudation rates are potentially not capturing the occurrence of large sporadic erosional events (Kirchner et al., 2001;Schaller et al., 2001), with the exception of catchment DB01 (and therefore DB02,06,10,12 along the main DB course), where massivemajor erosional events have been occurring during the Holocene towards present-day (i.e.rockfall events; Deline et al., 2012Deline et al., , 2015) ) and therefore potentially included in the 10 1 -10 2 yr integration time of the modern denudation rates.Alternatively, low 10 Be-concentration sediment input in the river system, coming from linear fluvial incision and subglacial sediment export, could explain the mismatch between modern and millennial denudation rates, with 10 Be-derived denudation rates being potentially overestimated (Stutenbecker et al., 2018).

Conclusions
Our 10 Be-derived catchment-wide denudation rates obtained in the Dora Baltea (DB) catchment (western Italian Alps) vary between 0.2 and 0.9 mm/yr and fit within literature values across the European Alps (Delunel et al., 2020).Correlation of output denudation rates with topographic, environmental and geologic metrics excludes any significant control of precipitation and rock uplift on the observed variability in denudation rates within the DB catchment.Our results instead highlight the main influence of catchment bedrock structuration and erodibility (litho-tectonic origin) and associatedresulting topographic metrics on denudation rate variability among the 13 main tributaries.As previously supposed for some other parts of the Alps, our study shows that the most resistant lithologies (granite and gneiss) support high-elevation and high-relief catchments where glacial and slope processes are more intense and denudation rates are higher than in low-elevation/relief catchments, dominated by "high erodibility" bedrocks (carbonate and terrigeneous rocks).
This litho-tectonic control on catchment denudation is exemplified by the tributary catchment draining the Mont Blanc Massif, which has the highest 10 Be-derived denudation rate from our dataset and appears as an end-member for most of the investigated metrics.Located in the long-term actively-uplifting core of the European Alps, the Mont Blanc Massif also experienced intense Quaternary glaciations which deeply modified the landscape.Steep slopes and high reliefs could be supported by the highlyresistant granitoid lithology, which in turn have been influencing the millennial to present-day high denudation of the catchment, governed by intense glacial/periglacial processes and recurring rockfall events.In addition, our results also show that the high sediment input from the Mont Blanc catchment dominates the DB sediment flux, contributing to >77% of the 10 Be signal carried by river sediments along the DB main river, even downstream of multiple tributary junctions.This suggests poorunequal sediment mixing and balancecontribution between tributary fluxes along the DB catchment.
Finally, our 10 Be-derived denudation rates allow for comparison with long-term (10 6 -10 7 yr, from thermochronology) and modern (10 1 -10 2 yr, from sediment budget) erosiondenudation rates, showing that, albeit different absolute values, the spatial trend in catchment denudation is overall in agreement over different timescales, with higher millennial denudation rates in the Mont Blanc Massif compared to the rest of the DB catchment.

Figure 1 :
Figure 1: Study area with investigated Dora Baltea (DB) and main tributary river catchments (mosaic DEM from Regione Autonoma Valle 75 and the Lal/Stone time-dependent scaling model(Lal, 1991; Stone,   2000).Neutron, slow and fast muons are assumed to contribute respectively 98.86, 0.87 and 0.27% to the total 10 Be production rate(Charreau     et al., 2019, after  Braucher et al., 2011, and Martin et al., 2017).d 10 Be-derived catchment denudation rates were calculated with Basinga (Charreau et al., 2019), using default attenuation length values of 160, 1500 and 4320 g cm -2 , for neutrons, slow muons, and fast muons, respectively (Charreau et al., 2019, after Braucher et al., 2011), and assuming a rock density of 2.7 g cm −3 .Denudation-rate uncertainties are estimated only based on values and relative errors of 10 Be concentrations and cosmogenic production rates from neutron and muons (Eq. 5 in Charreau et al., 2019).e Apparent ages represent the time needed to erode one absorption depth scale (~0.6 m in bedrock; von Blanckenburg, 2005) and are given as mean estimates (no uncertainty propagated).

Figure 3 :
Figure 3: Downstream evolution of river-sand 10 Be concentrations in the Dora Baltea (DB) catchment.Data are plotted versus distance from the main DB source (upper Val Veny, right tributary upstream DB01).In red are samples collected along the main DB river, in yellowlight blue are samples at the outlet of tributaries (Fig. 1 for locations). 10Be concentration of sample T12 (red square) from Wittmann et al. (2016) is also shown for discussion.Samples DB03 and DB14, 14 and 18 are omitted since they are in turn tributaries of catchments DB04 and DB13, respectively, and do not directly connect to the main DR river.

Figure 4 :
Figure 4: CorrelationsRelationships between tributary-catchment denudation rates and mean catchment (A) elevation, (B) slope, (C) 5-km geophysical relief, and (D) geodetic uplift.CorrelationsLinear correlations have been calculated including or not sample DB01 (red and black lines, respectively; see main text for discussion).Correlation coefficients (p-value and R 2 r 2 ) are reported for each plotlinear regression with significant trendstatistical significance (p-value < 0.05).R 2 r 2 is not reported for non-significant correlations (p-value > 0.05).Asterisks 325 330

Figure 5 :
Figure 5: CorrelationsRelationships between tributary-catchment denudation rates and catchment (A) mean annual precipitation, (B) relative bare-bedrock area, (C) relative area covered by LIA glaciers, and (D) relative area above LGM ELA (2103 m a.s.l.).Correlations have been calculated including or not sample DB01 (red and black lines, respectively; see text for discussion).Correlation coefficients (p-value and R 2 r 2 ) are reported for each plotlinear regression with significant trendstatistical significance (p-value < 0.05).R 2 r 2 is not reported for nonsignificant correlations (p-value > 0.05).Asterisks indicate linear correlations for which DB01 has been considered as an outlier (based on Cook's distance, TableS3).

Figure 6 :
Figure 6: Tributary-catchment denudation rates (A) and relative proportion of litho-tectonic units within the quartz-bearing areas of the individual catchments (B).Colour code in (A) refers to the most abundant litho-tectonic unit in each individual catchment (see Figure 2 for spatial distribution of the different litho-tectonic units).

Figure 7 :
Figure 7: Box-and-whisker plots for the spatial distribution of elevation (A), slope (B), and 5-km geophysical relief (C) within the entire DB catchment, classified by individual litho-tectonic unit.Red horizontal line represents the median of each distribution, bottom and top of each box are the 25 th and 75 th percentiles.Whiskers extend up to 1.5 times the interquartile range, outliers (red crosses) are observations 375 factors for catchment-wide 10 Be production and denudation rates As reported in Table

Figure 7 :
Figure 7: Box-and-whisker plots for spatial distribution within the entire Dora Baltea catchment of elevation (A), slope (B), and 5-km geophysical relief (C), classified by individual litho-tectonic unit.Red line represents the median of each distribution, bottom and top of each box are the 25 th and 75 th percentiles.Whiskers extend up to 1.5 times the interquartile range, outliers (red crosses) are observations beyond the whiskers.
). Similarly to what has been shown by Glotzbach et al. (2013), the external Alps catchments (west of the Penninic Frontal Thrust; Fig. 2) appear to have equivalent long-term (apatite fission-track derived) and short-term ( 10 Be-derived) erosiondenudation rates, while internal Alps catchments (east of the Penninic Frontal Thrust; Fig. 2) haveshow higher short-term than long-term erosiondenudation rates.This has been tentatively explained by potential differences in driver mechanisms of denudation before and during the Quaternary (Glotzbach et al., 2013).Tectonic forcing dominated Neogene denudation rates, with fast exhuming External Massifs having steeper rivers and higher reliefs and therefore eroding faster than the slowly-exhuming Internal Alpine Massifs.During the Quaternary, instead, climate fluctuations and associated glaciations modified both the Internal and External Alps morphology and topographic reliefs, also resulting in increasingincreased denudation rates for the Internal Alps.Our 10 Be-derived catchment denudation rates for the DB catchment are therefore not totally reflecting long-term exhumation rates over Myr timescales but most probably highlight Quaternary erosion dynamics.

Figure 8 :
Figure8: Comparison of 10 Be-and modern river yield-derived catchment -wide denudation rates.For all tributary catchments and DB locations, modern denudation rates are derived from river-bedload empirical estimates of river bedload are available(Vezzoli et al., 2004;

Table 1 :
. River-sediment sample locations, measured 10 Be concentrations, calculated mean catchment 10 Be production rates, and output denudation rates and apparent ages.Sample coordinates are given in decimal degrees (dd), and sample names collected along the main DB river are written in italics.