Magnitude and timing of transient incision resulting from large‐scale drainage capture, Sutlej River, Northwest Himalaya

Few studies have constrained the magnitudes and timescales associated with large‐scale drainage captures (areas >103 km2), even though these constraints are crucial to reconstruct sediment budgets, assess the potential for drainage reorganization to be preserved in the rock record, and determine the extent to which environmental signals (i.e., structures, composition and fossil assemblages within sedimentary rocks that are influenced by sediment supply and transport) are representative of conditions during deposition. In this work, we characterize the Pleistocene capture of the Zhada Basin, an ~23 000 km2 extensional basin in southern Tibet, by the Sutlej River, a prominent tributary to the Indus River. We quantify the magnitudes and timescales of capture‐driven erosion using knickpoint celerity modelling, paleotopographic reconstructions, 10Be‐derived denudation rates, and topographic analyses of drainage divides. We find that capture has removed 2010 ± 400 km3 of sediment from the Zhada Basin, increasing sediment supply to the Sutlej network by 17%–29% since 735 ± 269 ka. This work represents a crucial step towards reconstructing the Pleistocene sediment budget of the Indus sedimentary system and identifying potential impacts from sediment redistribution. We also identify several plausible tectonic or autogenic mechanisms that may have facilitated capture of the Zhada Basin, including: (1) preferential erosion of weak lithologies along active faults, (2) headward erosion in response to prior capture of the Spiti River and (3) headward erosion generated by breaching of a structural culmination downstream (the Kullu‐Rampur Window). This provides a framework to assess the mechanistic links between arc‐parallel extension, large‐scale drainage capture, landscape evolution and orogenic wedge deformation.

timescales.Rearrangement of river networks through drainage capture, wherein catchment area is transferred from one network to an adjacent network (Bishop, 1995), alters discharge and sediment flux for both expanding and contracting networks (sensu Scheingross et al., 2020).This area transfer can occur under either constant or changing climatic or tectonic forcings (e.g., Beeson et al., 2017;Pelletier, 2004;Willett et al., 2014), which can complicate interpretations made from morphological features that reflect spatiotemporal changes in erosion and deposition (e.g., knickpoints and channel steepness).
In this work, we characterize the Pleistocene capture of the Zhada Basin of southern Tibet by the Sutlej River and quantify associated magnitudes and timescales of capture-driven erosion.Specifically, we (1) identify transient knickpoints to test the capture hypothesis and resolve propagation rates using digital elevation model (DEM) analyses and knickpoint celerity modelling, (2) quantify the volume removed by capture-driven erosion using reconstructed, pre-capture basin topography, (3) estimate the basin-averaged denudation rate for the Zhada Basin using a combined dataset of new and published 10 Be-derived denudation rates, (4) refine capture age estimates using volume and denudation rate estimates and (5) determine whether capture was driven primarily by headward erosion or overspill mechanisms by comparing modern divide elevations with reconstructed paleotopography.Our analyses produce several key findings.First, we estimate that the transient response to capture within the Zhada Basin has removed 2010 ± 400 km 3 of sediment, implying an increase in sediment supplied to the Sutlej network within the orogen by 17%-29% on average since capture occurred at 735 ± 269 ka.Second, capture was likely driven by headward erosion rather than a top-down mechanism (Bishop, 1995;Smith, 2013), suggesting basin integration resulted from tectonic forcings and/or autogenic processes.Lastly, we posit new mechanisms for large-scale drainage capture within the Himalayas in the absence of clear tectonic or climatic forcings, which has important implications for interpreting changes in erosion and deposition rates.

| STUDY AREA
The Sutlej is the largest trans-Himalayan river and a prominent tributary to the Indus River (Figure 1).Although locally referred to as the Satluj River, it is widely referenced as the Sutlej River in the scientific literature.From its headwaters, the Sutlej flows into the arid and unvegetated Zhada Basin of southern Tibet, then to the WNW towards the Leo Pargil dome before bending sharply as it exits the basin and enters the deeply incised Sutlej Valley across the Himalayan orogenic core (Figures 1 and 2).The Sutlej exits the Himalayas onto the Indo-Gangetic Plain, transitioning from a primarily bedrock river to an alluvial river (Figure 2).Within the Sutlej drainage network are several features consistent with past drainage reorganization (Bishop, 1995).Tributaries to the Sutlej contain numerous barbed drainages, suggesting rerouting of flow paths in response to drainage capture (Figure 2; Bishop, 1995).Furthermore, a windgap north of the Qusum Range has been interpreted as the former river channel of an ancient river rerouted in the late Miocene (Figure 2; Saylor, DeCelles, Gehrels, et al., 2010).The most dramatic evidence of recent drainage capture comes from the Zhada Basin, where a low-relief landscape has been deeply incised to bedrock by the Sutlej network, exposing the overlying basin fill sequence along dramatic cliffs (Figure 2; Kempf et al., 2009;Saylor, DeCelles, Gehrels, et al., 2010;Saylor, DeCelles, & Quade, 2010).
The interpretation of the geomorphic surface as the latest paleodepositional surface implies it represents a datum recording the transition from deposition to incision.We interpret the age of the geomorphic surface as a maximum age of the hypothetical capture event integrating the Zhada Basin into the Sutlej network.Four studies have attempted to constrain the age of the Zhada Formation and the geomorphic surface using magnetostratigraphy paired with biochronologic correlations (Qian, 1999;Saylor et al., 2009;Wang et al., 2013;Wang, Zhang, et al., 2008).Wang et al. (2013) reinterpreted these previous studies with additional mammalian faunal assemblages to produce a reconciled chronology of the basin fill and an estimated depositional age of $400 ± 200 ka for the uppermost Zhada Formation.We interpret this to be the best age estimate of the geomorphic surface.Before attempting to quantify the impacts of large-scale drainage capture on the Sutlej sedimentary system, we provided quantitative support for the capture hypothesis by recreating observed transient knickpoint locations using an area-dependent knickpoint celerity model (Rosenbloom & Anderson, 1994).We reconstructed the geomorphic surface atop the Zhada Formation to approximate the paleotopography at the time of capture and obtained volumetric estimates of capture-driven erosion.We used these volumes, along with the basin-averaged, modern denudation rate of the Zhada Basin, to resolve a viable age range for the capture event.Finally, we determined whether capture was driven by a top-down or bottom-up mechanism, providing insight into the processes driving capture.In the following section, we document our approach using a combination of area-dependent celerity modelling, surface interpolation, DEM differencing, 10 Be-derived denudation rates and topographic analysis of drainage divides.

| Transient knickpoint identification
Numerous studies have used DEMs to extract longitudinal profiles of bedrock rivers, which can be interpreted using the stream power family of models (e.g., Perron & Royden, 2013;Whipple & Tucker, 1999;Wobus et al., 2006).The simplest form of the stream power model is as follows: where E is incision rate, K is bedrock erodibility, A is contributing drainage area, S is channel gradient and m and n are empirical constants (Howard & Kerby, 1983;Whipple & Tucker, 1999).Any instance where a channel segment experiences base level lowering results in an increase in S immediately upstream from base level.This process increases E (Equation ( 1)) and forms a knickpoint, a discrete change in S that separates a downstream channel segment adjusting to the new base level from an upstream segment that retains the original profile form (e.g., Kirby & Whipple, 2012).Since S and E of the downstream segment are greater than S and E of the upstream segment, the knickpoint migrates upstream (e.g., Crosby & Whipple, 2006;Niemann et al., 2001;Wolpert & Forte, 2021).When this knickpoint encounters a tributary junction, an additional transient knickpoint forms within the tributary channel (e.g., Crosby & Whipple, 2006), leading to a networkwide response to downstream base level change.
Transient knickpoints generated in lithologies with uniform erodibility and uplift rate, and in the absence of diffusion, propagate at a constant and uniform vertical velocity (Niemann et al., 2001).Therefore, a diagnostic feature of channel networks adjusting to an instance of base level fall is transient knickpoints clustered around a common elevation (Armstrong et al., 2021;Berlin & Anderson, 2007;Gallen et al., 2013;Larimer et al., 2019;Miller et al., 2013).Since drainage capture inherently leads to base level fall within the captured network, an expected outcome is a cluster of knickpoints at similar elevations within the network in the Zhada Basin.These transient knickpoints should be radially dispersed, such that knickpoints in tributaries closer to the capture point have propagated further upstream from the Sutlej mainstem than knickpoints in tributaries farther away.By identifying transient knickpoints within the Zhada Basin, we could compare observed locations to predicted locations using a knickpoint celerity model to simulate knickpoint propagation in response to capture (Section 3.2).
We used a resampled 30-m DEM, consisting of 1 arc-second Shuttle Radar Topography Mission (SRTM) tiles supplemented with 1 arcsecond Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER; NASA/METI/AIST/Japan Spacesystems and U.S./Japan ASTER Science Team, 2019) tiles to fill voids, and standard channel analyses using Topographic Analysis Kit (TAK; Forte & Whipple, 2019) and TopoToolbox 2 (TTB2; Schwanghart & Scherler, 2014) to identify knickpoints within the Zhada Basin channel network.We defined the flow network using the D8 algorithm of TTB2 with a threshold A of 1 km 2 and smoothed the network using constrained regularized smoothing with a smoothing parameter value of 20 on the median profile from quantile regression (Schwanghart & Scherler, 2017).
We detected knickpoints from the channel network using the 'knickpointfinder' function of TTB2 with a tolerance value of 70 m and identified knickpoints concentrated at similar elevations, which we interpreted to be clusters of potential transient knickpoints.This tolerance value represented the maximum elevation difference between the 90th and 10th percentile channel elevations derived from a smoothed network using constrained regularized smoothing and quantile carving and approximated the error inherent in the DEM to ensure knickpoints resulting from artefacts in the DEM were not included in the resulting dataset (Schwanghart & Scherler, 2017).
Knickpoints that did not fall within a cluster were removed from the dataset of potential transient knickpoints.To distinguish between clusters of transient knickpoints and knickpoints reflecting structural control or the transition from glacial to fluvially-dominated incision, we examined knickpoints within each cluster using a combination of ESRI Satellite World Imagery and published geologic maps (Kempf et al., 2009;Saylor, DeCelles, & Quade, 2010).Knickpoint clusters that appeared to correspond with the observed extent of glaciers, terminal moraines, lithologic contacts or faults were excluded from the dataset.

| Knickpoint celerity modelling
We used a stream power-based celerity model to simulate knickpoint propagation in response to capture and compared expected and observed locations of transient knickpoints, consistent with previous studies (e.g., Berlin & Anderson, 2007;Crosby & Whipple, 2006;Miller et al., 2013;Mitchell & Yanites, 2019).These studies used celerity models to compare expected and observed locations of transient knickpoints, determine if observed knickpoints could have been generated from a common change in base level, and constrain the timing of such an event.A power law relationship for knickpoint celerity (dx/dt) can be derived from Equation (1) assuming n = 1 and considering E = dz/dt and S = dz/dx: Berlin & Anderson, 2007).Equation (2) can be rearranged as follows: which allows calculation of the time (dt) for the knickpoint to migrate across each stream pixel (dx).
We assumed the capture point is located at the outlet of the Zhada Basin, which coincides with a prominent knickpoint along the Sutlej mainstem (Figure 2b).We also assumed no significant erodibility contrasts within the Zhada Formation and applied a uniform (k sn ) to downstream reaches with higher k sn or were co-located with the visible transition from the low-relief, relict landscape to the deeply dissected downstream landscape.Knickpoints that did not satisfy all three conditions were removed from the dataset.
Streams containing transient knickpoints from the trimmed dataset were then selected from the channel head to the capture point using TAK, and upstream distance and drainage area values were extracted for each pixel along each stream using TTB2.We then implemented the TTB2 function 'mnoptim' to determine the ratio m/ n for the streams we analysed.This function uses Bayesian optimization to determine the optimal value of m/n (Equation ( 1)).We assumed n = 1, supplied the optimal value of m/n as the value of m in Equation ( 3), and ran the model.
To model the Zhada Basin capture scenario, we assumed the duration of the model run corresponded to the time since capture (T cap ), which we considered to be between 200 ka and 1 Ma (Saylor, DeCelles, Gehrels, et al., 2010;Wang et al., 2013).Independent constraints on K within the Zhada Basin do not exist, so we conducted 2550 model runs over a range of K for natural substrates (10 À6 to 10 À3 in 10 0.02 increments; Stock & Montgomery, 1999) and T cap (0.2 to 1.0 Ma in 0.05 Ma increments) to account for uncertainties in K and T cap .We then identified the best fit model for each T cap (n = 17) by finding the K resulting in the lowest misfit, defined as the sum of squared differences in the upstream distance from the outlet of the Zhada Basin between the modelled and observed knickpoints for each of the examined streams.

| Surface reconstruction and eroded volume estimation
To estimate the volume of material removed from the Zhada Basin due to transient capture-driven erosion, we reconstructed the geomorphic surface above the Zhada Formation by fitting polynomial surfaces through elevation points corresponding to surface remnants.
We interpreted the geomorphic surface to be a datum representing the pre-capture elevation of the basin fill and used the surface reconstructions to approximate paleotopography at the time of capture (Ant on et al., 2019;Mather et al., 2002).We then estimated the volume of material removed by capture-driven erosion by differencing the modern topography from the paleotopography.
We first isolated elevation points from the 30-m DEM within the mapped extent of the Zhada Formation (Kempf et al., 2009) and excluded pixels with surface slopes greater than 5 (Pérez-Peña et al., 2009).To avoid including active fluvial features (e.g., channel beds and terraces), we also excluded pixels within 500 m of channels.
From this set of pixels, we eliminated pixels at elevations below 4450 m and above 4850 m to ensure ridgetops at elevations below and above the geomorphic surface were excluded from the fit, as we assumed they are likely modified by hillslope erosion and not representative of the correlative surface.Pixels within the highly dissected northwestern portion of the basin, adjacent to the Qusum Range, were also excluded.Since this portion lies on the basin fringe, we interpreted this region to primarily consist of topography generated prior to capture.
Since the geomorphic surface is correlative, subhorizontal and minimally deformed (Saylor, DeCelles, & Quade, 2010), we reconstructed the surface using polynomial regression to fit both a quadratic and cubic surface to the remaining elevation points corresponding to the modern relict surface.Areas where reconstructed surface elevations were below current elevations were removed, assuming these areas corresponded with paleotopography.
We then created difference rasters consisting of the vertical difference between each surface reconstruction and modern topography.
Since the pre-capture lateral extent of the Zhada Formation is poorly constrained due to incision by the Sutlej network, we considered two possible end-member extents of the geomorphic surface at the time of capture: a minimum extent corresponding only to the mapped area of the Zhada Formation (fig.1b of Kempf et al., 2009) and a maximum extent corresponding to the entire Zhada Basin area upstream from the presumed capture point.The difference raster for each surface reconstruction was clipped to correspond to either the minimum or maximum surface extent, and vertical differences were summed to produce volumetric estimates of capture-driven erosion.

| Cosmogenic radionuclide-derived denudation rates
Eroded volume estimates alone do not reveal the extent to which capture-driven incision has altered the total sediment supplied to the Sutlej network, nor do they provide information on the age of Zhada Basin capture.To address both of these shortcomings, we used a combined dataset of new and published cosmogenic 10 Be-derived denudation rates to (1) characterize the spatial pattern of denudation within the Sutlej drainage to quantify the impact capture of the Zhada Basin had on the sediment budget and (2) estimate modern denudation rate for the Zhada Basin, allowing us to estimate capture age independently from methods used in previous studies (Qian, 1999;Saylor, DeCelles, Gehrels, et al., 2010;Wang et al., 2013;Wang, Zhang, et al., 2008).
We conducted analyses of in-situ 10 Be from quartz sand in stream sediment samples collected at the mouths of 22 small (<300 km 2 ) catchments along the Sutlej mainstem to estimate catchment-averaged denudation rate (Lal, 1988;Schaefer et al., 2022).Stream sediment was sieved to isolate the 250-500 μm fraction, and quartz was purified following procedures described in Kohl & Nishiizumi (1992) using froth flotation to separate feldspars from quartz.Beryllium was extracted using ion chromatography following procedures outlined in Codilean et al. (2023).Samples were spiked with $300 μg of 9 Be from a lowlevel beryl carrier solution added prior to complete HF dissolution.
Denudation rates were calculated using CAIRN v.1 (Mudd et al., 2016).Nuclide production from neutrons and muons was calculated with the four-exponential approximation of Braucher et al. (2009) using a sea-level and high-latitude total 10 Be production rate of 4.3 atoms g À1 year À1 (Mudd et al., 2016) scaled using the timeindependent Lal/Stone scaling scheme (Stone, 2000).Production rate was calculated at every pixel using the SRTM 1-arc DEM, with atmospheric pressure calculated via interpolation from the NCEP2 reanalysis data (Compo et al., 2011).Topographic shielding was calculated from the same DEM using the method of Codilean (2006).All calculations assumed a 10 Be half-life of 1.387 ± 0.012 Myr (Chmeleff et al., 2010;Korschinek et al., 2010).Since only two of the 22 sampled catchments have present-day glaciers, which occupy between $2% and $8% of the catchments' areas, we do not correct 10 Be-derived denudation rates to account for present-day glacial influences (Appendix SA).Similarly, due to the lack of reliable data, we do not account for lithologic variations within sampled catchments (Appendix SA).
To better understand the relative contributions of different portions of the Sutlej drainage to the sediment budget, we combined our new 10 Be denudation rates with rates from previous studies and estimate catchment-averaged denudation rates for the Zhada Basin and two other large (>8000 km 2 ) subbasins in the Sutlej drainage: (1) the Spiti drainage upstream of the confluence with the Sutlej and (2) the orogenic Sutlej subbasin, which comprises the remainder of the Sutlej drainage upstream of the contact between Sub-Himalaya and Lesser Himalaya (i.e., Main Boundary Thrust [MBT]) and primarily drains the rapidly uplifting core of the Himalaya (Figure 2).Denudation rates from previous studies were re-calculated using the same methods as for our new samples (Table 1), so our 10 Be-derived denudation rate dataset is internally consistent and comparable with data published in the OCTOPUS database (Codilean et al., 2018(Codilean et al., , 2022)).
We assumed an average denudation rate of 0.13 and 0.70 mm/ year for the Zhada Basin and Spiti drainage, which correspond to samples RS-7 and RS-4 collected near the outlets of each respective basin (Table 1; Bookhagen B., unpublished data;Olen et al., 2020).Since there are no currently published denudation rates integrating the entire area of the orogenic Sutlej subbasin, we interpolated denudation rates from small (<300 km 2 ) catchments to estimate an average denudation rate for the subbasin.To ensure that denudation rates used in the interpolation were not biased by landscape transience (Reinhardt et al., 2007;Willenbring et al., 2013) or landsliding (Niemi et al., 2005;Yanites et al., 2009), we examined stream profiles and satellite imagery within the area upstream of each 10 Be sample (Figure S1) and excluded samples corresponding to catchments exhibiting major knickpoints that could reflect ongoing transience.We generated a 1-km grid of interpolated denudation rates for the orogenic Sutlej subbasin using an inverse distance weighted (IDW) algorithm to assign values based on proximity to sampled catchment centroids.We then calculated the spatially averaged denudation rate for the orogenic Sutlej subbasin by finding the average pixel value of the grid.To resolve the relative contribution of capture-driven erosion to the sediment budget, denudation rates corresponding to each subbasin were converted to volumetric rates.We assume constant subbasin geometries and that smaller-scale drainage reorganization within the upper reaches of the Spiti drainage (Munack et al., 2016) has minimally impacted the subbasin-averaged denudation rate.
To produce age estimates for the onset of incision in the Zhada Basin (i.e., capture age), we divided our eroded volume estimates by the basin-averaged denudation rate corresponding to the Zhada Basin (0.13 mm/year; Bookhagen B., unpublished data) and assume this rate has remained constant since capture.

| Divide profile analysis
Top-down and bottom-up-driven captures have different implications for the likely driving forces behind drainage integration.Top-down capture occurs when a higher-elevation drainage network is redirected into a lower drainage network (Smith, 2013), typically from overfilling of a basin with sediment or lake overspill (Geurts et al., 2018;Smith, 2013).These events are thought to primarily result from climatically-driven changes in precipitation (e.g., Garcia-Castellanos et al., 2003;Heidarzadeh et al., 2017), but can be tectonically-driven (Geurts et al., 2018) et al., 1996;Kempf et al., 2009;Saylor et al., 2009Saylor et al., , 2016)).
Bottom-up capture is associated with headward erosion within a lower-elevation catchment that breaches the divide with a higherelevation catchment (Bishop, 1995;Smith, 2013).Headward erosion is often interpreted as the result of changes in tectonic forcings (D'Agostino et al., 2001;Smith, 2013;Stokes et al., 2002), however it can be facilitated by groundwater flow across the drainage divide (Brocard et al., 2011;Pederson, 2001).If our analyses did not support basin overfilling or overspilling, we assumed headward erosion was the predominant driver of drainage capture.

| RESULTS
Our work has resulted in the following findings: (1) transient incision following capture of the Zhada Basin is ongoing, (2) approximately 2010 ± 400 km 3 of material has been removed by capture-driven erosion, (3) capture occurred at 735 ± 269 ka, (4) sediment supply within the Himalayan portion of the Sutlej drainage has increased by $17%-29% since capture and (5) capture was driven primarily by headward erosion.In the following sections, we provide a detailed account of these findings.
Knickpoints between 4350 and 4550 m are radially dispersed and coincide with a transition from deeply incised channels to a low-relief landscape we interpret as the geomorphic surface atop the Zhada Formation (Figure 3).Knickpoints between 3800 and 4100 m appear to coincide with the mapped basal contact between the Zhada Formation and Mesozoic Tethyan bedrock (Figure S2; Kempf et al., 2009;Saylor, DeCelles, & Quade, 2010) and do not correspond with any obvious contrast in topographic relief between the upstream and downstream landscapes (Figure 3).We speculate that they are fixedcontact knickpoints resulting from contrasting substrate erodibilities (e.g., Wolpert & Forte, 2021).Satellite imagery reveals that knickpoints between 5000 and 5500 m typically coincide with the termini of glaciers on the basin margins (Figure S2).We assume these knickpoints correspond to the transition in the landscape between being dominated by glacial and fluvial processes.(Figure 4 and Table S1).These values lie within the range of viable K for sedimentary rocks (Stock & Montgomery, 1999).All 17 best fit models produced physically reasonable results, with modelled maximum and minimum celerities (Table S1) lying within the range of measured average knickpoint propagation rates over timescales of 10 5 to 10 6 years (0.1 to 10 m/year; Loget & Van Den Driessche, 2009).Misfits for the 17 best fit K, T cap pairs imply the celerity model reproduced knickpoint locations along their respective channel segments within an average along-channel distance of 3.85-3.93km (Table S1 and Figure 4c) or within $2% of the observed upstream location.

| Estimated volume denuded due to capture
Based on surface reconstruction and differencing, our best estimate for material removed from the Zhada Basin by capture-driven erosion is 2010 ± 400 km 3 .Elevation points corresponding to the modern relict surface comprise relatively continuous, low-relief areas along the NE and SW portions of the Zhada Basin that consistently decrease in elevation towards the basin centre and the Leo Pargil dome (Figures 3   and 5).Assuming a minimum extent of the Zhada Formation, paleosurface reconstructions using quadratic and cubic polynomial regressions produce similar eroded volumes ($1600 km 3 and Table 2).
Assuming a maximum extent, the cubic reconstruction resulted in almost twice as much removed material (4781 km 3 ) than the quadratic reconstruction (2410 km 3 ).Both successfully recreate a subhorizontal paleosurface in the basin centre, with elevations increasing to the NE and SW margins (Figures 5 and S3).While the cubic reconstruction has a lower overall misfit than the quadratic reconstruction (Table 2), it has more local variability and produces physically unreasonable surface elevations with a high degree of curvature that is not representative of the relatively undeformed, subhorizontal Zhada Formation strata (Kempf et al., 2009;Saylor, DeCelles, Gehrels, et al., 2010).For example, surface elevations near Mt.Kailash are greatly overestimated (>7000 m) with an elevation change of 3000 m over $50 km (Figure S3).Taking

| Estimated denudation rates and capture age
Denudation rates from the Sutlej drainage, along with eroded volume estimates, suggest that Zhada Basin capture occurred around 735 ± 269 ka.Our combined denudation rate dataset ranges from 0.09 to 1.95 mm/year with peak rates of >1 mm/year in the orogenic core that decrease to $0.1 and $0.4 mm/year in the downstream and upstream directions, respectively (Figure 6 and Tables 1 and 3).These denudation rates integrate over timescales between $0.3 and 7 kyr, defined as the time required to erode $60 cm of rock or 1 m of soil (i.e., one cosmic ray absorption depth scale; Schaefer et al., 2022;von Blanckenburg, 2005).We used 13 new rates and two published rates from Olen et al. (2020) to generate a grid of interpolated denudation rates over the orogenic Sutlej subbasin (Figure 7).Interpolation resulted in an average denudation rate of 0.65 mm/year within the subbasin (Table 4).
Volumetric denudation rates for the three subbasins suggest the addition of material from the Zhada Basin to the Sutlej drainage network increased sediment supply by 17-29% since capture, with a volumetric denudation rate for the Zhada Basin of 2.9 ± 0.5 km 3 /ka (Table 4).This volumetric denudation rate is similar to analogous cap-  F I G U R E 6 Geology of the study area and upstream basins corresponding to our new and published 10 Be samples used in this study (Tables 2  and 3).

| Style of capture
Our observations rule out basin overfilling and overspilling as viable mechanisms for capture of the Zhada Basin, suggesting that headward erosion was the primary driver of capture.Reconstructed paleosurface elevations along the Zhada Basin drainage divide are consistently lower than modern divide elevations.At the capture point, the estimated elevation of the surface is 4450 m, while divide elevations on either side of the Sutlej range from 5550-6500 m (Figure 9).The >1100-m difference in elevation between modern divide elevations and the reconstructed surface suggests basin overfilling was improbable, and overtopping of the divide at the capture point would have required an unrealistically deep paleolake (e.g., >1 km).Previous studies suggest shrinking of the paleolake in the Zhada Basin started at 3.5 Ma in response to decreasing Indian Summer Monsoon strength (Gasse et al., 1996;Kempf et al., 2009;Saylor et al., 2009Saylor et al., , 2016) ) and ostracodes assemblages of the uppermost Zhada Formation indicate deposition in a shallow water (<10 m) and highly saline environment (Kempf et al., 2009).Additionally, if paleolake elevations exceeded 4950 m, spillover would have occurred across the Himalayan crest into the Dhauliganga River catchment rather than the Sutlej (Figure 9).
Therefore, $1 km of relief must have existed at the capture point before integrating the Zhada Basin, which we assume was likely generated by headward erosion.

| DISCUSSION
Our results represent the first estimates of the magnitudes, rates, and timescales associated with transient incision resulting from the capture of the Zhada Basin.Volumetric estimates of capture-driven erosion and denudation rates provide an independent estimate of capture age that agrees with previous estimates.We also identify potential capture mechanisms by comparing the reconstructed paleodepositional surface with the divide profile of the Zhada Basin.In the following sections, we present a synthesis of the capture event, outline mechanisms that could have led to the capture of the Zhada Basin, and discuss the significance of this event in the context of downstream impacts.

| Synthesis and characterization of capture
Results from knickpoint celerity modelling strongly support the capture hypothesis, allowing us to attribute estimated eroded volumes to capture-driven erosion.While we assume n = 1 in our models, we cannot discount that n ≠ 1.Since transient knickpoints are propagating through similar lithology with regionally consistent surface gradients and minimal relief (Figures 2 and 5), we do not expect spatial variations in channel slope to largely influence the spatial distribution of knickpoints.Rigorous testing of the assumption of n = 1 would require additional constraints on erodibility that are not possible given available data.In the absence of independent measures of erodibility of the Zhada Formation, we cannot use our results to constrain capture age.However, we note the consistent high correspondence in all 17 best fit model scenarios between the observed and modeled patterns of knickpoints suggests knickpoints between 4350 and 4550 m are propagating as a kinematic wave, further supporting the hypothesis of capture-driven erosion (Figure 4; Bishop, 1995;Miller et al., 2013).Since our approach assumes knickpoint propagation is initiated by a discrete lowering of base level, we assume that capture of the Zhada Basin is well-approximated as a single event.Given the lack of significant deformation within the Zhada Formation (Kempf et al., 2009;Saylor, DeCelles, Gehrels, et al., 2010), we infer that remnants of the geomorphic surface have not been significantly modified since capture and that differences between paleotopography at the time of capture and the modern topography correspond primarily to capture-driven erosion (Ant on et al., 2019;Pérez-Peña et al., 2009;Stokes et al., 2002).While the topography within the Zhada Basin has likely been modified to some extent by surface processes since capture (e.g., minor amounts of surface lowering due to deflation, bioturbation and wind erosion), we assume that these effects are relatively minor and that the quadratic surface reconstruction (Figure 5c) provides a reasonable estimate of paleotopography.Since the estimated depositional age of the uppermost Zhada Formation ($400 ± 200 ka; Wang et al., 2013) serves as an upper age bound on basin capture, we assume capture occurred near the younger end of our estimated age range (735 ± 269 ka).While this age estimate depends on the assumption that the Zhada Basin catchment-averaged denudation rate has remained constant since capture, we assume that it is unlikely past rates exceeded the modern rate given the proportion of relict topography in the catchment ($70% by area).Previous models of the landscape response to base level lowering suggest catchmentaveraged denudation rates in a catchment composed of $70% relict topography, like the Zhada Basin, should still be increasing over time (Willenbring et al., 2013).Past basin-averaged denudation rates may have been less than the modern rate, especially immediately after capture.However, we assume rates much lower than the modern rate would have been short-lived and relatively insignificant considering the rapidity of modelled knickpoint celerities ($1 m/year, Table S1) and the highly erodible and relatively undeformed Zhada Formation sediments.

| Potential mechanisms driving capture
Our comparison of divide elevations to reconstructed paleotopography suggests it was unlikely that capture resulted from basin overfilling or overspilling and was, instead, driven by headward erosion.Identifying possible mechanisms for headward erosion is crucial to resolving the spatiotemporal scales associated with setting up the conditions for large-scale drainage capture to occur (Scheingross et al., 2020).Recent work suggests tectonic processes control the denudation pattern across the strike of the Himalaya (S-N), with climatic changes occurring in response to tectonic changes (e.g., Godard et al., 2014;Morell et al., 2017;Olen et al., 2016;van der Beek et al., 2016).We, therefore, assume that climatically driven headward erosion is unlikely to have led to Zhada Basin capture.In the following sections, we present three equally plausible mechanisms that could be responsible for driving headward erosion along the Sutlej: preferential erosion of weak lithologies along fault zones bounding the Leo Pargil dome, transient landscape adjustment resulting from prior capture of the Spiti River by the Sutlej, and headward erosion resulting from breaching of the Lesser Himalaya sequence in the Kullu-Rampur window at $3-7 Ma.

| Preferential erosion along weak Lithologies
Mechanically weak lithologies can be preferentially eroded by rivers (Mudd et al., 2022), constraining channels to zones of high erodibility (e.g., Duvall et al., 2020;Koons et al., 2012;Roy, Koons, et al., 2016;Roy, Tucker, et al., 2016).Thiede et al. (2006) and Zhang et al. (2000) both observed non-cohesive fault gouge within active fault zones along the flanks of the Leo Pargil dome.Preferential erosion and channelization within the Kaurik-Chango fault zone have been observed (Thiede et al., 2006), suggesting active faulting may control planform drainage patterns in the Sutlej network (Styron et al., 2010;Taylor & Yin, 2009;Thiede et al., 2006).Zones of weakness along active faults accommodating exhumation of the Leo Pargil dome (Thiede et al., 2006) could have facilitated capture of both the Spiti and Zhada Basins (Figure 2), with flow paths becoming entrenched, leading to focused incision.Furthermore, fault gouge may form a preferential pathway for groundwater flow and subsurface erosion, which could have contributed significantly to drainage integration (Brocard et al., 2011;Dunne, 1980;Laity & Malin, 1985;Pederson, 2001).Capture aided by preferential erosion provides a mechanism connecting ongoing arc-parallel extension in southern Tibet, which serves as a tectonic control on basin formation (Saylor, DeCelles, Gehrels, et al., 2010;Thiede et al., 2006;Zhang et al., 2000), to large-scale capture in the Himalayan orogen, which can redistribute large amounts of sediment in the foreland basin and orogenic wedge.

| Capture of the Spiti River
Another possibility is that an earlier capture event downstream from the Zhada capture point initiated headward erosion along Sutlej tributaries in response to increased discharge (e.g., Beeson et al., 2017;Giachetta & Willett, 2018;Scheingross et al., 2020).Numerous barbed drainages within the Spiti network and the proximity of the Spiti-Sutlej confluence to the Zhada capture point ($30 km, Figure 2a) imply that capture of the Spiti may have generated sufficient headward erosion to breach the Zhada Basin.Thiede et al. (2006) speculated that an apparent increase in cooling rate between 3 and 4 Ma from apatite fission track cooling ages from the Spiti drainage could reflect an increase in exhumation rate due to the incisional response to drainage reorganization.Channels within the Spiti network also run parallel to and coincide with mapped brittle faults of the Kaurik-Chango fault system (Thiede et al., 2006), suggesting preferential erosion may have aided capture.Importantly, when channels incise through low-erodibility rock overlying high-erodibility rock, an upstream propagating scarp develops within the low-erodibility rock (Forte et al., 2016;Perne et al., 2017).

| Breaching of the lesser Himalaya
This process can lead to the discrete capture of upstream tributaries and basin expansion through headward erosion (Gallen, 2018).2b).Therefore, we assume most eroded sediment from capture has been transported out of the High Himalaya and deposited in alluvial terraces downstream from the Kullu-Rampur window (e.g., Bookhagen et al., 2006;Wulf et al., 2012), in the foreland basin, along the Indo-Gangetic Plain, and in the Indus Fan.
Increased sediment supply from capture-driven transient erosion can impact each of these depositional environments and the surrounding landscapes in various ways.
Increased sediment supply can alter erosional efficiency by providing additional 'tools', which enhance incision by increasing the frequency of grain impacts, or additional 'cover', which armors the channel bed and buffers incision (Sklar, 2001;Sklar & Dietrich, 1998, 2004, 2006).Normalized channel steepness along the Sutlej River remains higher than adjacent river networks through portions of the Lesser and Greater Himalayas (Figure 10).With its arid climate, capture of the Zhada Basin could have contributed to a relatively large increase in supplied sediment and a relatively small increase in discharge, leading to the dominance of the cover effect downstream and a steeper channel profile as the Sutlej adjusted to retain topographic equilibrium (e.g., Forte et al., 2022;Gasparini et al., 2007).The absence of reaches with sustained high steepness in neighbouring drainages supports this hypothesis.A similar pattern is observed in the Arun River drainage downstream from a suspected capture point near the Kharta-Chentang gorge (Wang et al., 2016).We, therefore, speculate there may be a link between capture, increased sediment supply, and channel steepening for 10s of km downstream of the capture point (Figure 10).However, we acknowledge that other mechanisms may be responsible for these steepness patterns.These include, but are not limited to, locally elevated uplift rates resulting from river anticline development (e.g., Montgomery & Stolar, 2006) and greater contributions of coarse sediment from hillslopes along high steepness reaches (e.g., Lai et al., 2021).
Downstream increases in sediment supply may also influence the structural evolution of orogens.DeCelles & Carrapa (2021) outline a mechanical link between anomalous erosion within the orogen, foreland sedimentation, and orogenic wedge kinematics using critical taper models in central Nepal (e.g., Davis et al., 1983;Suppe, 2007).
According to critical wedge theory, anomalous erosion within an orogenic wedge should prohibit foreland propagation of the frontal thrust and promote mechanisms leading to wedge thickening (i.e., out-ofsequence thrusting).Additionally, if eroded sediment is deposited along the frontal portion of the orogenic wedge, this should promote forward propagation of the thrust front, as this portion of the wedge could be critical or supercritical.This pattern of inhibited thrust propagation followed by forward propagation of the thrust front is observed in the Chitwan Basin of Nepal (DeCelles & Carrapa, 2021), where an MBT reentrant is coupled with a Main Frontal Thrust salient bounding the basin.This same structural pattern is observed in the Kangra reentrant, the eastern portion of which the Sutlej flows into (Figure 10).Anomalous erosion resulting from the capture of the Zhada Basin might have influenced wedge structure in the northwest Himalayas.However, anomalous erosion resulting from the breaching of the Lesser Himalaya within the Kullu-Rampur window in the late Miocene to early Pliocene might have had an influence as well (Colleps et al., 2019;Exnicios et al., 2022).
In alluvial rivers, increases in sediment flux can increase meander migration rates, which modifies floodplains by increasing sediment storage due to higher cutoff rates (Constantine et al., 2014) , 2006;Jerolmack & Paola, 2010;Romans et al., 2016).Furthermore, the Sutlej has undergone avulsions that have altered its course through channel switching and abandonment near where the Sutlej crosses the Himalayan front and enters the Indo-Gangetic plain (e.g., Roy et al., 2021;Singh et al., 2017).Increases in both discharge and sediment flux can increase avulsion frequency (e.g., Bufe et al., 2019;Jerolmack & Mohrig, 2007)

F
I G U R E 2 (a) Overview of the Sutlej River drainage and proposed capture point.LP-Leo Pargil gneiss dome; QR-Qusum Range; KF-Karakoram Fault; NM-Namru-Menci Basin; MK-Mount Kailash; GM-Gurla Mandhata gneiss dome; MP-Mukut Parbat.Black arrow indicates the location of the identified wind gap of (Saylor, DeCelles, Gehrels, et al., 2010) and paleoflow direction before basin enclosure.(b) Longitudinal profile of the Sutlej River with highlighted locations of the capture point (yellow) and prominent upstream knickpoints (orange) shown in (a), (c) and (d).(c) Relief calculated over a 1-km radius.(d) Surface slope.The low relief and shallow slopes of the Zhada Basin contrast with other portions of the Sutlej drainage that are higher in relief and have steeper slopes.[Color figure can be viewed at wileyonlinelibrary.com] throughout the model domain.The model begins by simulating pixel-by-pixel knickpoint propagation from the capture point.When a pixel is occupied by the knickpoint, Equation (3) is used to calculate dt before the knickpoint propagates to the next upstream pixel.The value of dt is added to the cumulative time (Σdt) of knickpoint migration, and the process repeats for the upstream pixel.The model stops once Σdt exceeds a user-supplied age of base level fall.The end location of the modelled knickpoint can be compared to the observed knickpoint location along the modelled stream.Before extracting model inputs from the Zhada drainage network, we confirmed the following for each knickpoint to ensure it was appropriate for use as model input: (1) the knickpoint was not located within 500 m of mapped contacts or faults, (2) the stream containing the knickpoint had no other potentially transient knickpoints along its flow path and (3) the knickpoint coincided with either a transition from upstream relict reaches with low normalized channel steepness . Since top-down capture requires sedimentary fill or surface water to reach the spill point elevation at the time of capture, it's straightforward to rule out top-down mechanisms as the primary driver of capture.Top-down integration of the Zhada Basin implies one of two criteria must have been satisfied at the time of capture: (1) the Zhada Basin became overfilled with sediment, forcing runoff across the spill point (basin overfilling); or (2) the water level of the paleolake within the Zhada Basin exceeded the spill point elevation for a time sufficient to incise, leading to transient incision downstream and upstream of the capture point (overspilling).If basin overfilling occurred, the upper surface of the Zhada Formation would have reached the pre-capture elevation of the capture point (i.e., spill point).To test this hypothesis, we compared the projected elevations of our reconstructed surfaces at the capture point to modern divide elevations to determine if the Zhada Formation was deposited to the estimated spill point elevation.If the projected elevation of the reconstructed surface at the capture point was greater than the adjacent modern drainage divide, we assumed basin overfilling occurred.Otherwise, overspilling may have occurred if the paleolake elevation at the capture point was greater than the estimated spill point elevation.Published paleodepth estimates were added to the reconstructed surface elevation at the spill point to assess the likelihood of overspill leading to capture (Gasse Knickpoints between 4350 and 4550 m exhibit a spatial distribution similar to what is expected for transient knickpoints generated by drainage capture.Prior to modelling, we removed 31 of the original 112 knickpoints within the cluster according to the conditions outlined in Section 3.2.The optimal value of m/n for the Zhada Basin drainage network using 'mnoptim' was 0.36.Using the DEM-derived drainage areas, upstream lengths, and m = 0.36, we optimized for K for each of the potential capture ages, resulting in best fitting K values ranging from 6.80 Â 10 À5 to 3.44 Â 10 À4 m 0.28 year À1

F
I G U R E 3 (a) Spatial distribution of detected knickpoints in the Zhada Basin drainage network with respect to the geomorphic surface.(b) Elevation distribution of detected knickpoints.Knickpoints cluster between elevations of 3800-4100, 4350-4550 and 5000-5500 m.Knickpoints at elevations between 4350 and 4550 m are highlighted and interpreted as transient knickpoints generated by capture-driven incision.(c) Example area showing knickpoints separating high-steepness, high-relief areas from low-steepness and low-relief areas corresponding to the current edge of the geomorphic surface.[Color figure can be viewed at wileyonlinelibrary.com]F I G U R E 4 Relationship between observed knickpoints and modeled knickpoints from a stream power-based celerity model.Results shown here are for the best-fitting erodibility (1.72 Â 10 À4 m [1-2m] year À1 ) with a capture age of 400 ka.(a) Longitudinal profiles of select tributaries in the Zhada Basin.Channel segments below transient knickpoints (red) are steeper than upstream segments (blue) and are actively incising through basin fill (yellow).(b) Observed and modeled knickpoints in map view.Average residual between modeled and observed knickpoints is 3.85 km.(c) Plot of knickpoint observed upstream distance vs. modeled upstream distance.Dashed line represents the 1:1 line.Similarity between observed and modeled distances suggests a transient response to capture can explain observed knickpoints.[Color figure can be viewed at wileyonlinelibrary.com]

F
I G U R E 5 (a) Hillshade of Zhada Basin with mapped outline of the Zhada formation from Kempf et al. (2009).Colored points correspond to relict surface nodes used in surface reconstructions.Dashed area also corresponds to the minimum extent area used in volumetric calculations using DEM differencing.Yellow area corresponds to the more highly dissected northern portion of the basin that was excluded from surface reconstructions.(b) Quadratic surface reconstruction within the extent of the Zhada Basin.Contours represent projected paleoelevations for the reconstruction.(c) Paleotopography reconstruction.Surface extent (colored areas) after removing portions where the modern elevations are greater than projected paleoelevations.Colored areas with paleoelevation contours also represent the estimated maximum extent of the Zhada Formation.(d) Plotted residuals for the surface reconstruction.(e) Incision map generated by differencing the paleotopography reconstruction in (c) and the modern topography.Incision increases towards the capture point and the extent of the incised region corresponds with observed transient knickpoints.[Color figure can be viewed at wileyonlinelibrary.com] this into consideration, the cubic reconstruction overfits the relict surface, and we interpret the quadratic surface to be a more representative geomorphic surface reconstruction.
to summed elevation differences within the mapped Zhada Formation, maximum volumes correspond to summed differences over the entire Zhada Basin.
Basins are colored according to their corresponding estimated denudation rate.Major geologic units and orogen-scale faults are shown, along with the proposed capture point of the Zhada Basin.Inset shows the location of samples RS-7 and RS-4.Mapped units and faults are from Webb et al. (2011).[Color figure can be viewed at wileyonlinelibrary.com]T A B L E 3 Results of cosmogenic radionuclide denudation rate analyses.
Production and denudation rates were calculated using the CAIRN open-source code(Mudd et al., 2016).Samples IN14-15, IN14-16 and IN14-17 analysed on ANTARES; all other samples analysed on SIRIUS.All uncertainties are reported at the 1 sigma level.Latitude and Longitude of sample locations based on the WGS 1984 datum.a Sample was not used in denudation interpolation for the orogenic Sutlej subbasin.F I G U R E 7 Interpolated denudation maps.Denudation rates for the Spiti and Zhada Basins correspond to values in Table 3. Interpolated rates for the orogenic Sutlej drainage area were derived using an inverse distance weighted (IDW) algorithm.[Color figure can be viewed at wileyonlinelibrary.com]T A B L E 4 Estimated average denudation rates for subbasins in the Sutlej drainage.average denudation rate has been spatially uniform and constant since capture.F I G U R E 8 Summary of capture age estimation using estimated removed volumes and denudation rate in response to capture for the Zhada Basin.Dotted lines represent the minimum and maximum volume estimates from surface differencing.Solid line corresponds to constant denudation at the estimated mean denudation rate for the Zhada Basin (0.13 mm/year) and dashed lines represent mean denudation rate, plus or minus uncertainty.Grey region corresponds to the predicted range of capture ages from comparing volume and denudation rate estimates.Previous capture age estimates from published studies are shown for comparison.[Color figure can be viewed at wileyonlinelibrary.com] 2.3 km 3 /ka; Ant on et al., 2019).Assuming the denudation rate of the Zhada Basin has been constant since capture and dividing the minimum and maximum volumetric estimates from surface differencing (1610 and 2410 km 3 , respectively) by this rate implies capture likely occurred around 735 ± 269 ka (Figure 8).
Zircon (U-Th)/He ages along the Sutlej River suggest removal of Greater Himalayan Crystalline and Tethyan Himalaya rocks in the Main Central Thrust hanging wall occurred by $7-10 Ma in response to the development of the Lesser Himalaya duplex(Colleps et al., 2019).Subsequent breaching of Lesser Himalayan rocks in the Kullu-Rampur window at $3-7 Ma would have exposed more erodible lithologies (e.g., greenschist facies metasedimentary rocks; Figure6;Colleps et al., 2019), generating an upstream propagating wave of enhanced incision, as demonstrated in the hard over soft models ofForte et al. (2016).Along the Sutlej, headward erosion initiated by breaching of the Lesser Himalaya may have been sustained by subsequent captures of upstream tributaries, including capture of the Spiti, suggesting all three of our proposed mechanisms may have operated in concert to drive Zhada Basin capture.5.3 | Downstream effects of captureBy altering drainage morphology, large-scale drainage captures can lead to downstream changes in incision or deposition rate(Yanites et al., 2013), which in turn can influence meander migration rate(Constantine et al., 2014) and the structural evolution of orogens(DeCelles & Carrapa, 2021).Drainage area transfer increases discharge and sediment supply in the capturing stream, which may lead to transient downstream incision or deposition depending on the impact on transport capacity (e.g.,Gilbert, 1877).Based on our field observations of limited stored sediment and steep canyon walls, we presume channel reaches along the Sutlej immediately downstream of the capture point have experienced an increase in incision rate since capture.The knickpoint coinciding with the inferred Zhada capture point suggests the river network is still undergoing transient incision in this region (Figure Capturedriven increases in sediment flux may have increased sediment storage on the Indo-Gangetic Plain, which, in theory, could have increased the buffering of environmental signals before reaching the Indus Fan F I G U R E 1 0 Red ellipses denote reaches downstream from suspected capture points in the Sutlej (a) and Arun (b) drainages where increased sediment supply may have increased channel bed cover, leading to channel steepening.Locations indicated in Figure 1.Main Frontal Thrust (MFT), Main Boundary Thrust (MBT) and Kangra Reentrant.[Color figure can be viewed at wileyonlinelibrary.com] (Clift , suggesting capture may have played a role in determining the current flow paths.While it's unclear what proportion of eroded sediment from the Zhada Basin has reached the Indus Fan, our findings suggest that capture coincided with a Pleistocene increase in sedimentation rate within the fan(Clift & Blusztajn, 2005).Our volume estimates help reconcile the Pleistocene sediment budget for the Sutlej-Indus River network by accounting for 2010 ± 400 km 3 of sediment supplied within the past 1 Ma.6 | CONCLUSIONSOur findings support the hypothesis that the Sutlej River captured the Zhada Basin and demonstrate that the landscape within the basin continues to adjust to base level.By reconstructing the paleotopography of the basin at the time of capture, we estimate that 2010 ± 400 km 3 of sediment was removed from the basin through capture-driven erosion.This has increased the amount of sediment supplied to the Sutlej network in the Himalayan orogen by $17%-29% on average since capture occurred at 735 ± 269 ka.Increases in sediment supply due to drainage capture can alter sediment travel time due to changes in intermediate storage, influencing whether and how these events are preserved in the stratigraphic record.Additionally, resolving erosion magnitudes and rates associated with capture is a crucial step towards reconstructing the Pleistocene sediment budget of the Indus sedimentary system.We identify several mechanisms that may have facilitated capture of the Zhada Basin, including: (1) preferential erosion of fault gouge within fault zones bounding the Leo Pargil dome, (2) headward erosion in response to prior capture of the Spiti River and (3) headward erosion generated by breaching of the Lesser Himalaya sequence beneath more resistant lithologies.This work provides a framework to assess the mechanisms linking arc-parallel extension, large-scale drainage capture, landscape evolution and orogenic wedge deformation.AUTHOR CONTRIBUTIONS Conceptualization was by BDP and KDM.KDM acquired the funding for this work and provided supervision.BDP, KDM, ATC, RHF, KMW and BJY developed the methodology for this work.BDP, KDM, BJY, AK and TM collected field data and ATC, RHF and KMW conducted laboratory analyses.BDP wrote the initial draft and all authors reviewed and edited subsequent drafts.
Selected published 10 Be-derived denudation rates.a Olen et al. (2020)Olen et al. (2020), except RS-7, which was provided by B. Bookhagen (unpublished data).b Sample was not used in denudation interpolation for the orogenic Sutlej subbasin.