Rockwall Slope Erosion in the Northwestern Himalaya

Rockwall slope erosion is an important component of alpine landscape evolution, yet the role of climate and tectonics in driving this erosion remains unclear. We define the distribution and magnitude of periglacial rockwall slope erosion across 12 catchments in Himachal Pradesh and Jammu and Kashmir in the Himalaya of northern India using cosmogenic 10Be concentrations in sediment from medial moraines. Beryllium‐10 concentrations range from 0.5 ± 0.04 × 104 to 260.0 ± 12.5 × 104 at/g SiO2, which yield erosion rates between 7.6 ± 1.0 and 0.02 ± 0.004 mm/a. Between ∼0.02 and ∼8 m of rockwall slope erosion would be possible in this setting across a single millennium, and >2 km when extrapolated for the Quaternary period. This erosion affects catchment sediment flux and glacier dynamics, and helps to establish the pace of topographic change at the headwaters of catchments. We combine rockwall erosion records from the Himalaya of Himachal Pradesh, Jammu and Kashmir, and Uttarakhand in India and Baltistan in Pakistan to create a regional erosion data set. Rockwall slope erosion rates progressively decrease with distance north from the Main Central Thrust and into the interior of the orogen. The distribution and magnitude of this erosion is most closely associated with records of Himalayan denudation and rock uplift, where the highest rates of change are recorded in the Greater Himalaya sequences. This suggests that tectonically driven uplift, rather than climate, is a first order control on patterns of rockwall slope erosion in the northwestern Himalaya. Precipitation and temperature would therefore come as secondary controls.

. Schematic diagram of glaciated catchment with primary debris transport pathways . The rockwall is defined for this study as the headwater bedrock slopes above the ELA of each glacier. This study focuses on erosion via periglacial processes only. The medial moraines and supraglacial debris are revealed below the ELA (gray shading). ELA, equilibrium-line altitude. rockwall slope erosion in the NW Himalaya. Finally, we determine to what extent rockwall slope erosion and its controls, in this high-altitude and high relief setting, can contribute to the longstanding debate over the significance of climate versus tectonics in driving both short-and long-term landscape change. Until now, much of the research that has contributed to this debate is based in either unglaciated or deglaciated environments. This study will provide a unique insight into how erosional processes modulated by climate and/or tectonics operate within glaciated catchment headwaters.

Regional Setting
The Himalayan-Tibetan orogen is the result of the continued continental collision and partial subduction between the Indian and Eurasian lithospheric plates (Searle et al., 1997). The Indus-Tsangpo Suture Zone (ITSZ) defines the collision zone between these plates in the NW Himalaya and contains remnants of the Neo-Tethys Ocean (Figure 2). The suture zone marks the northern boundary of the Tethyan Himalaya (Schlup et al., 2003;M. Searle, 1986;Steck et al., 1998). Between the early Miocene and Pleistocene, deformation driven crustal shortening initiated the development of a sequence of foreland propagating thrust systems that divide the lithotectonic units that lie south of the Tethyan Himalaya. The South-Tibetan Detachment (STD) and the Main Central Thrust (MCT) bound the Greater Himalaya Crystalline Core Zone to the north and south, respectively (Frank et al., 1973;Miller et al., 2001;Searle & Fryer, 1986;Vannay et al., 2004;Walker et al., 1999). This unit has been divided into two subunits: southern Greater Himalaya sequence (GHS-S) and northern Greater Himalaya sequence (GHS-N;DeCelles et al., 2001;Thiede & Ehlers, 2013). South of the Greater Himalaya and MCT lies the Lesser Himalaya sequence, which is bounded to the south by the Main Boundary Thrust (MBT). South of the MBT lies the Sub-Himalaya and Main Frontal Thrust (MFT; Upreti, 1999;Miller et al., 2000;Vannay et al., 2004). Continued crustal shortening and thrust and strike-slip faulting throughout the orogen means that the NW Himalaya remains tectonically active (Bojar et al., 2005;Hodges et al., 2004;Vannay et al., 2004), even though some regions in northern India, such as Ladakh, have undergone tectonic quiescence or dormancy since the early Miocene (Kristein et al., 2006(Kristein et al., , 2009. Hodges (2000), Yin and Harrison (2000), and Streule et al. (2010) provide further details of the Himalayan lithotectonic units and the timing of movement throughout the fault systems.
Two atmospheric systems primarily govern northwest Himalayan climate: the Indian summer monsoon that advects moisture from the Indian Ocean between late May and September, and the Northern Hemispheric mid-latitude westerlies, which bring moisture from the Mediterranean, Black and Caspian seas between December and March (Benn & Owen, 1998;Gadgil, 2003;Mölg et al., 2014;Lang & Barros, 2004;Wulf et al., 2010). A steep south-north precipitation gradient likely became established during the late Miocene, perpendicular to the strike of the mountain belt (∼8 Ma; Liu & Dong, 2013;Qiang et al., 2001), due to the high elevation topography of the Greater Himalaya inhibiting the northward migration of moisture to the interior of the orogen. Monsoon air masses are forced to ascend, condense and form clouds along the Himalayan front, which creates a rain shadow down the leeside of this orographic barrier (Bookhagen et al., 2005a(Bookhagen et al., , 2005bWulf et al., 2010). During times of increased monsoon strength, moisture is thought to penetrate farther into the interior of the orogen (Bookhagen et al., 2005a(Bookhagen et al., , 2005bFinkel et al., 2003;Wulf et al., 2010). The northern hemispheric mid-latitude westerlies operate at higher tropospheric levels than the Indian summer monsoon. The orographic capture of moisture transported by this atmospheric system is therefore focused in high elevation mountain ranges (>4,500 m asl) as winter snowfall (Lang & Barros, 2004;Weiers, 1995). Today, mean annual precipitation declines from ∼1,500 to 3,000 mm in the Lesser and Greater Himalaya ranges, to <150 mm in the interior of the Tethyan Himalaya and Tibetan Plateau (Bookhagen & Burbank, 2006, Figure 2).
The distribution and magnitude of precipitation has been shown to vary both temporally and spatially throughout the Himalayan-Tibetan orogen during the late Quaternary (Bookhagen et al., 2005a(Bookhagen et al., , 2005bBurbank et al., 2003). Fluctuations in monsoon strength driven by changes in orbital insolation, the migration of the intertropical convergence zone, convective localized monsoon storms, and sporadic heavy rainfall are thought to cause some of this variability (Finkel et al., 2003;Owen et al., 2008;Thomas et al., 2016). On the local to regional scale (10 2−4 km 2 ), topography and wind direction exert controls on the migration of moisture throughout the NW Himalaya (Bookhagen et al., 2005a, b), and create localized microclimates throughout individual mountain ranges (Benn & Owen, 1998;Bookhagen & Burbank, 2010;Wulf et al., 2010). Landscape change in the NW Himalaya is precipitation sensitive, where shifts in the availability and source of moisture is shown to initiate changes to sediment flux, hillslope processes (Bookhagen & Burbank, 2006;Bookhagen et al., 2005;Kumar et al., 2018;Sharma et al., 2017), and the timing of glaciation Saha et al., 2018).

Table 1
Details of the Investigated Catchments was reanalyzed. The Indian summer monsoon delivers two-thirds of the annual precipitation to Ladakh (87 mm/a; Table 1), whereas the mid-latitude westerlies provide the primary source of moisture to the Shigar region. Glaciers in the Ladakh region are small (1-10 km 2 ) cold-based subpolar glaciers, which are precipitation sensitive and sublimation dominated (Benn & Owen, 2002).
The arid/semiarid climatic setting of the Ladakh region is largely responsible for the preservation of very old landforms and sediment deposits (>400 ka; Hedrick et al., 2011;Orr et al., 2017Orr et al., , 2018Owen et al., 2006) and slow rates of landscape change (<0.07 ± 0.01 mm/a; Dietsch et al., 2015;Dortch et al., 2011a The Lahul-Spiti and Kullu district catchments are located in Pir Panjal and Greater Himalaya ranges of the Himachal Pradesh in northern India. Precipitation is primarily sourced from the Indian summer monsoon (950-1,020 mm/a; Table 1). Glaciers are large, temperate and melt dominated, and fed by precipitation from the summer monsoon and mid-latitude westerlies (Benn & Owen, 2002;Su & Shi, 2002).  catchment. The Indian summer monsoon also dominates annual precipitation in the Uttarkashi district of Uttarakhand, northern India, a region our study revisits and reanalyzes the rockwall erosion data of Orr et al. (2019).

Topographic Analyses
Geomorphic maps of the 12 investigated catchments were prepared in the field and then refined using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) global digital elevation models (GDEMs; 30-m-resolution), Landsat Enhanced Thematic Mapper Plus (ETM+) imagery and Google Earth imagery. Topographic parameters including catchment area, 3-km-radius relief, mean slope, hypsometry, and aspect were calculated using the Spatial Analyst Toolbox in ArcMap 10.1. These analyses were also conducted for Baltoro glacier system in the Shigar of Baltistan, Pakistan (Seong et al., 2009) and Bhagirathi glacier system in the Uttarkashi district of Uttarakhand, northern India  to enable comparisons between rockwall slope erosion and catchment parameters throughout the NW Himalaya.

Cosmogenic Nuclide Erosion Rate Theory for Rockwall Slopes
Production of cosmogenic 10 Be via neutron spallation occurs within the first few meters of the bedrock surface of the rockwall slopes and decreases approximately exponentially with depth. As material detaches from the rockwall slopes through erosion, new material moves into the zone of nuclide production (Balco et al., 2008;Lal, 1991;Uppala et al., 2005). The 10 Be inventory at the rockwall slope surface is the integrat-ORR ET AL.  ed cosmogenic nuclide production within quartz, during its exhumation through the zone of production Equation 1: N is the measured nuclide concentration at depth z, Po is the nuclide surface production rate, z* is the e-folding length scale for the falloff of production with depth from the surface (z * = Λ/ρ, where Λ is the spallogenic mean free path, and ρ is the target material density), E is the erosion rate, and λ is the radiogenic decay constant.
Rates of rockwall slope erosion can be inferred by measuring cosmogenic 10 Be concentrations within medial moraine sediment. In our study areas, medial moraines form within the glacier ablation zones as a result of englacial sediment melt out. This sediment is sourced and transferred from accumulation zone rockwall slopes to the glacier surface via rockfall processes and avalanching, before being transported englacially to the equilibrium line of the glacier and exhumed to the surface (Dunning et al., 2015;Goodsell et al., 2005;MacGregor et al., 2009;Matsuoka & Sakai, 1999;Mitchell & Montgomery, 2006; Figure 1). The 10 Be concentration of the medial moraine sediment reflects the averaged nuclide inventory of the source rockwall slopes. Due to the, stochastic nature of rockwall slope erosion, the 10 Be concentrations across the rockwalls are unlikely to be spatially uniform. The mean concentrations of these rockwall slopes are instead considered steady in time and linked to the mean erosion rate (Ward & Anderson, 2011).
ORR ET AL.  The medial moraine nuclide concentrations and rockwall slope production rates are used to calculate the rockwall slope erosion rate (E RS ; Equation 2; Balco et al., 2008;Granger et al., 1996;Lal, 1991). The longer the original sampled material has remained within the production zone of the rockwall slopes before being transferred to the glacier surface, the greater the 10 Be concentration measured in the medial moraine sediment, and therefore the slower the inferred rockwall slope erosion rate. This approach to quantifying rockwall slope erosion accounts only for the spallogenic component of incoming cosmic rays and assumes a negligible loss of 10 Be due to radioactive decay, and steady-state erosion and 10 Be production over time (Von Blanckenburg, 2005). Further details of this methodology and its assumptions are provided by Supporting information S1 and in Ward and Anderson (2011) and Sarr et al. (2019).

Cosmogenic Nuclide Analyses
Orr et al. (2019) argue that the rates of rockwall slope erosion in the upper Bhagirathi catchment in the Uttarkashi district of Uttarakhand are best represented by the rates derived from the centermost medial moraine of Gangotri glacier. This is because the 10 Be concentrations of the medial moraine fall within uncertainty of each other. Moreover, the concentrations from the other medial moraines are shown to be affected by input from lateral moraines and hillslopes along the ablation zone of the glacier. The study recommends that multiple samples should be taken from each medial moraine and/or glacier to constrain and evaluate any variability in rockwall slope erosion throughout the catchment headwaters. Two or fewer samples are only appropriate when the medial moraine is well preserved with steep relief ridges, has no interaction with ablation zone slopes and where other sampling locations do not fit these criteria. With these recommendations in mind, we carefully collected between one and five samples from stable and well-defined medial moraine ridges in each of the 12 catchments studied (Figures 3 and 4). Each sample location was ≥ 200 m 2 in area, to avoid sampling from a single source slope or rockfall event (see Supporting information S1). Approximately 3 kg of sediment with a grain size of <3 cm (clay-coarse gravels) was collected for each sample using bulk sediment sampling methods of Gale and Hoare (1991). Detrital samples of this grain size are shown to infer time-averaged erosion rates effectively, and for this study, are representative of the processes that contribute to rockwall slope denudation (Delunel et al., 2010;Lal, 1991;Puchol et al., 2014;Seong et al., 2009;Tofelde et al., 2018). Each sample was named using the initial term "G" for "glacier" followed by an abbreviated term for the catchment name. For glaciers with more than one sample, the samples were numbered in ascending order from the glacier snout. For example, the G Cht1 sample was located closest to the snout of Chhota Shigri in Lahul-Spiti, whereas G Cht5 was located farthest up-glacier.
Each sample was crushed and sieved in the Sedimentology Laboratories at the University of Cincinnati. A sample aliquot with equal mass input from each grain size fraction was created for each sample to avoid any one grain size from being overrepresented in the 10 Be analysis. The sample aliquot was crushed, sieved and the 250-500 μm fraction was retained for processing. The extraction of quartz and 10 Be isolation and purification was conducted at the Geochronology Laboratories at the University of Cincinnati, using the chemical procedures of Nishiizumi et al. (1989), Von Blanckenburg et al. (2004), and Wittmann et al. (2016). The 10 Be/ 9 Be was measured using accelerator mass spectrometry at the Purdue Rare Isotope Measurement (PRIME) Laboratory at Purdue University (Sharma et al., 2000). Native 9 Be was measured via ICP-OES for each sample upon the recommendations of Portenga et al. (2015). The total 9 Be, including native 9 Be, rather than just the 9 Be carrier, was then used to calculate the 10 Be concentrations for the data set.
This method for quantifying erosion assumes that sediment storage at the rockwall slopes of each catchment is limited and that the transport of sediment from the rockwall to the medial moraine is rapid (Von Blanckenburg, 2005). Ward and Anderson (2011) developed an analytical expression to quantify the accumulation of cosmogenic nuclides during the transport of sediment from the source rockwall to the medial moraine. They found that 10 Be accumulation during the burial, englacial transport, and exhumation of sediment to the glacier surface was negligible in landscapes with denudation rates ≤1 mm/a. This model was implemented in our study because some of the records of erosion local to our investigated catchments, particularly in Uttarakhand, exceed this threshold (0.13-5.37 mm/a; Lupker et al., 2013;Scherler et al., 2014;Vance et al., 2003). Moreover, the glaciers of this study share similar glacier geometries, surface velocities and debris cover characteristics as to those described in the study by Ward and Anderson (2011). The modeled 10 Be accumulation during this transport was then subtracted from the total 10 Be sample concentration for each sample, before deriving the rockwall slope erosion rates (Supporting information S1; Table S1).
Rockwall slope erosion rates were calculated using Equations 1 and 2, which are described in detail by Balco et al. (2008), Dortch et al. (2011a), Granger et al. (1996) and Lal (1991). A 2σ uncertainty ascribed to the AMS results was propagated through the erosion rate calculations. Beryllium-10 production rates were calculated and corrected for topographic shielding for each rockwall slope using a combination of Delunel et al. (2010) and Dortch et al. (2011a) codes in MATLAB R2017.a, an ASTER 30-m GDEM (16-m vertical resolution), a calibrated sea-level high-latitude 10 Be spallogenic production rate from Martin et al. (2017; http:// calibration.ice-d.org/) and a 10 Be half-life of 1.387 Ma (Chmeleff 2010;Korschinek 2010). Ward and Anderson (2011) and others have demonstrated that muonic production of 10 Be within amalgamated supraglacial debris sourced from the rockwall is negligible, and for the process timescales of rockwall slope erosion can be omitted from our calculation scheme (Akçar et al., 2014;Braucher et al., 2003;Sarr et al., 2019).
Widespread avalanching and minimal snow retention on the steep rockwall slopes within our study areas reduces our concern about the effects of snow shielding on the erosion data set. Scherler et al. (2014) estimated the impact of snow shielding on nuclide concentrations using remote sensing derived observations of snow cover duration and field-based measurements of annual daily snow depth in Uttarakhand, northern India. These data are unavailable for our entire study area. However, we have applied a 5.3% correction to our erosion rates; the mean correction value made by Scherler et al. (2014) for 10 catchments in Uttarakhand with similar topographic and climatic characteristics to our study area. This correction does not change any broad trends in the erosion data set (Table 4). However, due to the ambiguity attached to these correction estimates and the variability in mean annual precipitation across our study area, we prefer to use the uncorrected erosion rates herein.

Statistical Analysis of Erosion Data set
We calculated the Pearson Correlation Coefficient values (p) between the 10 Be rockwall slope erosion rates and climatic, topographic, tectonic and geologic parameters. A p value of <0.01 (at >99% confidence level) was applied. Each considered parameter has proven to influence rockwall and/or slope stability in other alpine regions (Krautblatter & Moore, 2014;McColl, 2012). In addition, we used Principle Component Analysis (PCA) to identify and evaluate the possible controls of rockwall slope erosion in the NW Himalaya (The R Core Team, 2018; Supporting informations S3 and S4). This approach has been successfully applied in other studies to identify and evaluate the nature and magnitude of the environmental and landscape response to changes in climate (Edwards & Richardson, 2004;Sagredo & Lowell, 2012;Seaby & Henderson, 2014). The topographic parameters included catchment and glacier area, mean slope of each catchment, rockwall and glacier, catchment 3-km-radius relief, mean catchment and snowline elevation, and glacier aspect. Climatic variables included mean annual precipitation (weather stations [as referenced in Table 1] and TRMM [1998][1999][2000][2001][2002][2003][2004][2005][2006][2007][2008][2009]) and temperature (weather stations and CRU2.0 [as referenced in Table 1]), mean rockwall slope temperature and minimum catchment temperature. Catchment specific temperatures were calculated using an adiabatic lapse rate of 7°C/km and methods outlined in Orr et al. (2019). Additional variables, such as sample grain size and mean apatite fission track (AFT) cooling age (as referenced in Table 5), were also included within these analyses. The latter enables us to identify correlations between modern erosion rates and regional exhumation histories on the million-year timescale. The Uttarkashi data set (Kirti, Bhagirathi) was not included in these analyses because the rockwall slope erosion rates characterize an extensive basin system with numerous tributary catchments, rather than a single catchment as the remaining data set does. This data set is examined in more detail in the discussion section below. P-values were also calculated between rockwall slope erosion rates and catchment parameters for Ladakh and Lahul-Spiti as discrete regions (Supporting information S4). The other studied districts were not subject this regional analysis because the data sets are restricted to only one or two catchments.
The ablation zones of the Lahul-Spiti and Kullu glaciers are partially to completely covered by debris, whereas in Ladakh, coverage is <30% of the glacier surfaces ( Figure 4; Table 3). Beryllium-10 sample concentrations for the Ladakh and Lahul-Spiti/Kullu catchments range from 6.0 ± 0.7 × 10 4 to 260.0 ± 12.5 × 10 4 at/g SiO 2 and 0.5 ± 0.04 × 10 4 to 30.6 ± 1.0 × 10 4 at/g SiO 2 , respectively ( Figure 3; Table 4). On average, ∼1% of the total 10 Be concentration of each sample was the result of 10 Be accumulation during transport from the source rockwall slopes to the medial moraine. For our study, this necessary correction to the final 10 Be concentrations had a negligible impact on the derived erosion rates (Table 4; Supporting information S1). Rates of rockwall slope erosion for the Ladakh region ranged between 0.02 ± 0.004 and 1.0 ± 0.2 mm/a, while rates in Lahul-Spiti/Kullu ranged from 0.2 ± 0.02 to 7.5 ± 1.0 mm/a (Figures 3 and 5; Table 4).
The catchment parameters with the most statistically significant relationship with rockwall slope erosion include mean rockwall slope, mean catchment and snowline elevation, mean annual precipitation, mean annual temperature, and mean AFT cooling age (Table 5). For the district-specific analysis, the same parameters are strongly correlated with rockwall slope erosion in Ladakh (p = <0.01; Supporting information S4). None of the parameters have a strong statistical correlation with the inferred erosion rates for the Lahul-Spiti district.

Discussion
ckwall erosion rates vary by up to two orders of magnitude throughout the NW Himalaya (0.02 ± 0.004-7.6 ± 1.0 mm/a; Figure 5). This is perhaps unsurprizing, considering the inherent complexities of periglacial-glacial environments, the application of cosmogenic nuclide analysis in these settings, and the range and variability in denudation recorded for this region (e.g., Scherler et al., 2014;Thiede & Ehlers, 2013;Vance et al., 2003). No relationship is apparent between 10 Be concentration and proximity of sample location to either a glacier margin or snout. Variability in 10 Be within the catchments is likely because the ORR ET AL.   medial moraine sediment is poorly mixed and/or has a nonproportional sediment supply from the rockwall that is dominated by stochastic rockfall events (Muzikar, 2008;Small et al., 1997;Ward & Anderson, 2011).
The strong variability in physical settings of the catchments prevent any meaningful interpretations or comparisons between specific erosion rates. Moreover, time-averaged nuclide-derived erosion rates come with large uncertainties when characterizing local areas (≤10 1 km 2 ), which has been shown to underestimate the true rates (Sadler & Jerolmack, 2014;Willenbring et al., 2013;Yanites et al., 2009). Instead, we focus on the broad trends of this rockwall slope erosion data set for the NW Himalaya. Rockwall slope erosion decreases with distance north from the MCT; up to two orders of magnitude difference in erosion exist between Ut-ORR ET AL.  tarakhand, Himachal Pradesh, Jammu and Kashmir, and Baltistan ( Figure 5). The Urgos catchment in northern Lahul-Spiti slightly deviates from this trend with erosion rates of 3.2 ± 0.5 and 7.6 ± 1.0 mm/a, which are equivalent to those records in Kullu and southern Lahul-Spiti. The elevated rates may be attributed to increased annual precipitation in Miyar, which exceeds much of Lahul-Spiti (snowfall: 120-400 cm/a; Patel et al., 2018) and allows for more rapid erosion. Alternatively, the low 10 Be concentrations could be due to the input of fresh debris from the large, steep relief lateral moraines along Urgos glacier (Figure 4e and 4f).
The applicable timescales of this time-averaged data set, although varied (∼0.1-24.6 ka), means that the erosion rates encompass recognized shifts in climate, sediment flux, glacier mass balance, and seismicity, which themselves operate across of timescales 10 1−6 years (Barnard et al., 2004;Finkel et al., 2003;Owen and Dortch, 2014;Scherler et al., 2015). Between ∼0.02 and ∼8 m of lateral rockwall slope erosion is possible for a single millennium in the NW Himalaya. When these rates are extrapolated for the whole Quaternary period, an estimated ∼2 km of rockwall retreat is accomplished in the NW Himalaya, which are similar estimates to the Sierra Nevada in the Western USA (Brocklehurst & Whipple, 2002). The magnitude of rockwall slope erosion evident in the NW Himalaya not only demonstrates the importance of slope erosion through periglacial processes, specifically frost cracking in high-altitude alpine settings, but also the significance that localized erosion has for understanding wider landscape change (Hales & Roering, 2005Moore et al., 2009;Sanders et al., 2012Sanders et al., , 2013Small & Anderson 1998). The rates of rockwall slope erosion reflect, in part, the pace of topographic change at the catchment headwaters.
The magnitude of erosion, particularly in the GHS-S, is sufficient to affect the strength of hillslope-glacier coupling, catchment sediment flux and contribute to topographic change such as the production of relief, the migration of catchment divides, and the reconfiguration of drainage basins (Heimsath & McGlynn, 2008;MacGregor et al., 2009 ibid;Naylor & Gabet, 2007;Oskin & Burbank, 2005). The rockwall slope erosion rates share a significant association with mean rockwall slope: the greater the mean rockwall slope, the more rapid the erosion (Figure 6a, Table 5). This points to important feedbacks between these variables, where the rockwall slope angle and erosion rate limit one another. A tentative relationship can also be recognized between relief and rockwall slope erosion; where catchments with the high-altitude peaks (>5,800 m asl), narrow ridgelines and high relief (>1.2 ± 0.2 km), record the highest rates of erosion. Part of this is because catchments with rockwall slope erosion rates >1 mm/a have mean rockwall slopes that exceed the 35° threshold, above which slopes are unable to retain regolith, snow or ice (Gruber & Haeberli, 2007;Nagai et al., 2013). This means that rockfall and avalanching is pervasive. More extensive glacier debris cover in these catchments compared to those with slower erosion demonstrate that coupling between rockwall and glacier is enhanced in catchments with steep accumulation areas, and that slope is important in moderating hillslope debris flux (Regmi & Watanabe, 2009;Scherler et al., 2011, Table 3). Other studies also recognize the importance of slope in landscape change, some of which argue that slope gradients can be used to infer rates of background denudation (Burbank et al., 2003;Finlayson et al., 2002;Ouimet et al., 2009;Portenga & Bierman, 2001;Scherler et al., 2011Scherler et al., , 2014. Rates of rockwall slope erosion in Uttarkashi and Ladakh districts are either equivalent to, or exceed by up to one order of magnitude, the local catchment-wide erosion and exhumation rates ( Figure 6). Quaternary exhumation rates range between ∼0.1 and 3 mm/a in the study areas (Thiede & Ehlers, 2013;Thiede et al., 2004). Catchment-wide rates for the Lahul-Spiti and Kullu districts are unavailable because much of the region remains glaciated . Orr et al. (2019) caution that comparing these erosion datasets can be problematic as they refer to landscape change through a variety ORR ET AL.

10.1029/2020JF005619
14 of 27 3.41 ± 1.08 × 10 −15 ; GGop1, GStk1, GPan1, GSht1, GBat1,2: 4.15 ± 3.9 × 10 −16 . c Accumulation of 10 Be during burial, englacial transport, and exhumation is calculated using methods detailed in Ward and Anderson (2011; see Supporting information S1). Accumulation is subtracted from the total concentration (  of erosional processes and across various spatial and temporal scales. Nevertheless, the order of magnitude difference in these rates shows that erosion at catchment headwaters in the NW Himalaya largely outpace the entire drainage basins (Naylor & Gabet, 2007;Oskin & Burbank, 2005), and that erosion can vary significantly across short distances downstream . Time-averaged rates for small areas such as catchment headwaters and rockwall slopes are sensitive to short-term local change, including single mass wasting events, and are therefore expected to record more rapid rates of erosion than a catchment-wide perspective (Yanites et al., 2009;Willenbring et al., 2013). The Karzok catchment in central Zanskar of Ladakh deviates from this trend as the rockwall slope erosion either equals or is slower than the catchment-wide erosion and exhumation rates ( Figure 6). The preservation and gradual reworking of landforms and sediment deposits that date to > 400 ka is likely affected by the low background denudation recorded in this region (Hedrick et al., 2011). A possible explanation is that sediment residence times exert a stronger control on the catchment-wide erosion signal in these ancient landscapes than the scale and various surface processes operating in the catchment area.

Controls of Slope Erosion
Considerable efforts have been made in recent years to define the parameters that control hillslope stability, and therefore determine the frequency and magnitude of mass wasting events (Ballantyne, 2002; ORR ET AL.

10.1029/2020JF005619
15 of 27  Fischer et al., 2006Fischer et al., , 2012Hales & Roering, 2005;Matsuoka, 2001;Regmi & Watanabe, 2009;Sanders et al., 2012Sanders et al., , 2013. The interactions between topography, climate, hydrology, geologic setting, and cryosphere dynamics have been shown to control rockfall activity. Of the catchment parameters that can be defined in the NW Himalaya, mean rockwall slope as already discussed, mean catchment and snowline elevation, mean annual precipitation, mean annual temperature, and mean AFT cooling ages show the strongest correlation with rockwall slope erosion rates (Figures 6 and 8; Table 5).
Catchments with the most rapid rockwall erosion have a greater proportion of the rockwall slope above the snowline than below, and larger glacier accumulation areas, than those with lower erosion rates. Aided by high gradient slopes that are set in part by erosion, field assessments, and satellite imagery suggest that snow and ice entrained debris is either removed from the rapidly eroding rockwalls via avalanching or is largely absent. Evidence of avalanching underlines the importance of snow processes and cover, whether set by climatic conditions or surface uplift, in the transfer of debris from the rockwall to the glacier system (Scherler et al., 2011. Estimated surface temperatures of the rockwalls are similar to those considered optimal for mechanical weathering processes (−8 to −3°C), for example, freeze-thaw, frost cracking, and frost wedging (Brozović et al., 1997;Hales & Roering, 2005;Hewitt, 2002;MacGregor et al., 2009;Matsuoka, 2001;Matsuoka & Sakai, 1999, Table 1). The medial moraine sediment characteristics are consistent with sediment from the supraglacial realm, which have detached from source slopes by periglacial weathering processes (Benn & Lehmkuhl, 2000;Benn & Owen, 2002;Hambrey et al., 2008;Lukas et al., 2012;Orr et al., 2019;Schroder et al., 2000, Table 3; Supporting information S5). Rates of periglacial erosion are likely further enhanced by seasonal and/or diurnal thermal variability in exposed bedrock surfaces of our investigated catchments, which is determined in part by the topographic steepness (Fischer et al., 2012;Gruber & Haeberli, 2007;Haeberli et al., 2017;Nagai et al., 2013). However, for high elevation catchments (>4,000 m asl) and/or rockwall slopes of our study area that lack an insulating layer of snow due to threshold slopes, bedrock surfaces can reach temperatures below −8°C, which inhibit further mass wasting (Ward & Anderson, 2011). This is tentatively reflected in the relationship between temperature and rockwall slope erosion; the catchments with lower regional temperatures record slower erosion rates (Figure 6c). The rockwall debris flux of each catchment is therefore likely influenced by the feedbacks between elevation, temperature, and slope.
A strong positive relationship between 10 Be-derived rockwall slope erosion and mean annual precipitation supports the view that the distribution and magnitude of Himalayan erosion and denudation is partly a function of orographically focused monsoon rainfall (Bookhagen & Burbank, 2006;Bookhagen et al., 2005a;Dey et al., 2016;Gabet et al., 2006;Theide et al., 2004;Wulf et al., 2010, Figures 6c and 7). The argument that precipitation provides a first-order control on the frequency and magnitude of mass wasting events in alpine settings is common (Dortch et al., 2009;Hovius et al., 2000;Iverson, 2000). Eppes and Keanini (2017) argue that the proficiency of mechanical weathering processes such as subcritical cracking is climate dependent, and specifically limited by moisture. Sources of moisture in alpine environments include snowfall, rainfall, and melt water. Although rockwall slope erosion is certainly influenced by the availability of moisture and is sensitive to the microclimatic conditions of each catchment, its distribution throughout the NW Himalaya cannot be fully explained by precipitation. A five fold decline in precipitation occurs between the first topographic high of the Lesser Himalaya (900 ± 400 m asl) and the interior ranges of the orogen (Bookhagen et al., 2005a(Bookhagen et al., , 2005bBookhagen & Burbank, 2006, Figure 7). If precipitation were the primary control ORR ET AL. Notes. Catchment and glacier characteristics data set detailed in Table 2 and Supporting information S3. Climatic data set detailed in Table 1 and Supporting information S3. Mean of AFT ages from Jain et al. (2000), Kristein et al. (2006Kristein et al. ( , 2009, Searle et al. (1999), Schlup et al. (2003), Sorkhabi et al. (1996, Thiede et al. (2004Thiede et al. ( , 2005Thiede et al. ( , 2006Thiede et al. ( , 2009  of rockwall slope erosion as we hypothesized, then we would expect to find that our maximum erosion rates coincide with maximum rainfall, and that a notable decline in these rates would be observed with distance north into the Greater Himalayan interior. However, our results show that this is not the case. Scherler et al. (2014) make a similar observation, where the highest catchment-wide rates in Uttarakhand are also located north of the precipitation maxima. To further, emphasize this point, there is an order of magnitude difference in the rockwall slope erosion rates between the GHS-N and the Tethyan Himalayan, yet a small decline in annual precipitation of < 300 mm.
Since the Late Miocene the steep orographic barrier of the Himalaya has restricted the northward advancement of moisture (Bookhagen et al., 2005a;Wulf et al., 2010), therefore preventing any subsequent major shift in the overall intensity or distribution of precipitation (Bookhagen & Burbank, 2010;Bookhagen et al., 2005a;Boos & Kuang, 2010;Thiede & Ehlers, 2013). The overall pattern in rockwall slope erosion throughout the NW Himalaya is therefore unlikely to be an artifact of a previous climatic regime, despite short-term fluctuations in monsoon strength during the Quaternary potentially affecting rockfall activity on the catchment scale (Demske et al., 2009;Fleitmann et al., 2003;Gupta et al., 2003;Thompson et al., 1997). One major concern in evaluating the role of climate in long-term landscape change is that the denudation records are averaged across million-year timescales and are therefore unable to account for the importance or variations in the Indian summer monsoon (Bookhagen et al., 2005a;Thiede & Ehlers, 2013). This study, is able to show that erosion records that reflect landscape change on timescales that would be sensitive to fluctuations in monsoon strength (10 2−5 years), that is, rockwall slope and catchment-wide erosion, are not unilaterally controlled by precipitation.
ORR ET AL.

10.1029/2020JF005619
17 of 27 The patterns in rockwall slope erosion rates are most closely associated with regional AFT cooling ages (Figures 6d and 7; Table 5). Much attention has been paid to understanding the patterns of cooling ages and exhumation rates in the Himalaya, and the feedbacks between tectonics and climate that are responsible for the distribution and intensity of Himalayan denudation across million-year timescales (Schelling & Arita, 1991;Srivastava & Mitra, 1994;Thiede & Ehlers, 2013). Many studies have argued that denudation is primarily governed by climate; orographic precipitation causes rapid erosion and exhumation along the Himalayan front and Lesser Himalaya (Biswas et al., 2007;Grujic et al., 2006;Kumar et al., 2018;Sharma et al., 2017;Thiede et al., 2004;Zeitler et al., 2001). However, young AFT ages (<10 Ma) and rapid rates of exhumation throughout the Lesser Himalaya and GHS-S instead reflect a close interaction between tectonics, denudation and monsoon-enhanced erosion, rather than just the latter (e.g., Thiede et al., 2004;Vannay et al., 2004;Wobus et al., 2003). Coupling between climate and tectonics becomes less evident farther into the Greater Himalayan interior; while the GHS-N becomes progressively more arid, the AFT ages remain <17 Ma and exhumation rates > 5 mm/a (Schlup et al., 2003;Thiede & Ehlers, 2013, Figure 7). The pattern in AFT ages and inferred exhumation histories for the NW Himalaya, like our rockwall slope erosion data set, cannot therefore be fully explained by precipitation. Instead, there is the argument that the patterns of Himalayan denudation are instead a function of tectonically controlled rock uplift; the result of crustal wedge deformation from the Indo-Eurasian collision and the flat-ramp-flat geometry of the Main Himalayan Thrust (e.g., Burbank et al., 2003;Bollinger et al., 2006;Godard et al., 2014;Herman et al., 2010;Robert et al., 2011). The lateral and vertical transport of rock over the ramp since the late Miocene has resulted in rapid and continuous exhumation, and the generation of steep topographic relief (Cattin & Avouac, 2000;Godard et al., 2004;Lavé & Avouac, 2000, 2001. Young AFT cooling ages and rapid rates of exhumation are therefore focused throughout the Lesser Himalaya and GHS-S (Figure 7). This is consistent with our patterns in rockwall slope erosion, therefore indicating that tectonically driven rock uplift throughout the NW Himalaya is likely to provide a major control on patterns of denudation since the late Paleogene, and also ORR ET AL. influence late Quaternary records of erosion (Scherler et al., 2011. The climatic parameters of precipitation and temperature are therefore likely secondary controls. Moreover, this confirms that similar casual relationships between rock uplift and erosion operate throughout the glaciated and non-glaciated regions of the NW Himalaya.
PCA indicate that ∼68% of the variance observed in rockwall slope erosion rates in the NW Himalaya can be explained by the six parameters discussed above (mean rockwall slope, mean catchment and snowline elevation, mean annual precipitation, mean annual temperature, and mean AFT age; Figure 8). To explain the remaining variance, other parameters must be considered. Rockwall lithology, rock strength and mass quality, and jointing and structure, for example, affect the thresholds for mass wasting and have been shown to govern hillslope debris flux and rates of erosion (Anderson, 1998;Augustinus, 1995;Fischer et al., 2010;Hales & Roering, 2005;Hallet et al., 1991;MacGregor et al., 2009). Rockfall activity in the investigated catchments is therefore likely affected by the erodibility of the rockwall and the periglacial processes acting upon it (Eppes & Keanini, 2017;Heimsath & McGlynn, 2008;Moon et al., 2017). The significance of this parameter in the patterns of rockwall slope erosion on the regional scale is however less clear. Previous work has argued that the difference in rock strength between the crystalline sequences of the Lesser and Greater Himalaya is negligible, and has little influence upon the denudation histories of the orogen (Burbank et al., 2003;Scherler et al., 2011Scherler et al., , 2014. Mass wasting events are closely associated with earthquakes and/or persistent microseismicity as documented by various studies throughout High Asia (e.g., Dortch et al., 2009;Hovius et al., 2000;Lupker et al., 2012;Menunier et al., 2008). For example, earthquakes in Uttarakhand such as the 1991 Uttarkashi (M 6.1; Bali et al., 2003;Valdiya, 1991) and 1999 Chamoli (M 6.6; Rajendran et al., 2000) events are found to trigger mass redistribution on a scale that affects short-term erosion rates (Bali et al., 2003;Scherler et al., 2014). The frequency of rockfall events and therefore rates of rockwall slope erosion in our catchments is therefore likely to be influenced in part by local tectonic activity.
A further candidate for rockwall slope erosion control is glaciation and glacial erosion; vertical incision and the debuttressing of slopes can lead to enhanced slope instability and failure (Fischer et al., 2010;Heimsath & McGlynn, 2008;MacGregor et al., 2009;Naylor & Gabet, 2007). Large, erosive temperate glaciers occupy catchments with rapid rates of rockwall slope erosion, while slower rates are from catchments with less erosive, subpolar glaciers . Past retreat and expansion of glacier ice may also have contributed to the evolution of the rockwalls; the downwasting of ice may encourage the unloading of slope debris, while a greater glacier volume may see an increase in glacial erosion processes acting upon the slope (Fischer et al., 2006(Fischer et al., , 2010(Fischer et al., , 2012Herman et al., 2010). For example, the Hamtah glacier in Lahul-Spiti has retreated ∼90 m in the last ∼200 years, during which time rockwall slope erosion rates have exceeded 3 mm/a (Tables 1 and 4; Figure 5; Saha et al., 2018). Rockwall slope erosion is therefore likely a critical component of the catchment headwater's response to shifts in glaciation, where the redistribution of stress from changing ice extents likely decreases slope stability (Gallach et al., 2018;McColl, 2012). These processes are part of a complex feedback; supraglacial debris cover above a critical thickness can act to insulate glacier ice, while a thinner layer can enhance melt by decreasing the albedo of the glacier surface (Ostrem, 1959). The erosion and delivery of debris from the rockwall slopes to the glacier can therefore affect surface melt rates, the mass balance of the glacier, and more broadly the glacier's sensitivity to environmental change (Anderson et al., 2011;Immerzeel et al., 2014;Gibson et al., 2017). To complete the feedback; glaciation and glacier dynamics regulate glacial erosion processes and local climatic conditions, which are both parameters that are shown to affect rockwall slope stability (Anderson et al., 2018;Heimsath & McGlynn, 2008;McColl, 2012).
Rather than a single control, we have confirmed the initial conclusions of Orr et al. (2019) by finding that rockwall slope erosion is instead more likely the result of longstanding feedbacks between climate, tec- tonics, topography, and surface processes which are specific to each catchment. The relative importance of these various parameters in driving rockwall slope erosion will likely vary across space and time. For example, the recognized relationship between rockwall erosion and slope for the NW Himalaya does not extend to the Lahul-Spiti district, if it were considered a discrete region. Only in Ladakh do the steepest catchment and rockwall slopes record the most rapid rates of erosion. This does not mean that rockwall erosion is unaffected by slope in Lahul-Spiti; however, it does suggest that other parameters are also necessary to explain the patterns of erosion. No parameters discussed show a strong correlation with rockwall slope erosion in Lahul-Spiti (Supporting information S4). Explanations for this might be that the erosion of rockwalls is influenced by a combination of parameters which together affect rockwall slope stability, or that this erosion is sensitive to undefined parameters such as glaciation and glacial processes (e.g., glacier type and dynamics, glacial erosion). Deciphering erosion controls may not be possible due to the inherent complexities of glaciated catchments in this district (e.g., rapid cycles of glacier retreat/advance, shifts in fluvial/meltwater discharge and erosion, variability in glacial/nonglacial sediment source-sink sedimentation; Adams et al., 2009;Bookhagen & Burbank, 2010;Bookhagen et al., 2006;Saha et al., 2018).
Controls of rockwall slope erosion may also be difficult to constrain across various temporal and spatial scales because for some catchments, once a threshold for a particular parameter has been met (e.g., moisture availability), rockwall slope erosion becomes predominantly limited by it. During a period of enhanced rainfall or monsoon along the Himalayan front for example (Bookhagen et al., 2005b;Clift et al., 2008), catchments with strongly contrasting geology and/or topography may display similar rockfall activity. In this case, the magnitude of precipitation is sufficient to govern rockwall slope stability and override any resistance to mass wasting (e.g., strong, nonerosive rock type or low gradient slopes). When averaged over time, these physically contrasting catchments will share a similar record of rockwall slope erosion. This may offer an explanation for why single high-magnitude events such as these, are viewed to be responsible for a significant proportion of the total landscape change in mountain environments (Craddock et al., 2007;Hasnain, 1996;Kirchner et al., 2001;Wulf et al., 2010). To tackle some of these outstanding questions, rockwall slope erosion controls should be evaluated for glaciated catchments with comprehensive geologic and climatic data and well constrained records of glacial history, topographic change, and mass wasting.
We suggest that, rockwall slope erosion is largely influenced by catchment-specific conditions that vary over temporal and spatial scales. However, our study is able to demonstrate that the broad spatial patterns in rockwall erosion follow long-term trends in denudation throughout the NW Himalaya, and is therefore broadly controlled by tectonically driven rock uplift. The climatic parameters of precipitation and temperature are therefore likely secondary controls. This suggests that, periglacial rockfall processes are part of the erosional response to structural change throughout the Himalayan-Tibetan orogen, and play a significant role within topographic change at catchment headwaters and the mass balance of the orogen. Overall, our study further adds to the growing body of evidence in the wider Himalaya and other regions such as the northern Bolivian Andes that denudation patterns do not always follow gradients in precipitation (Burbank et al., 2003;Gasparini & Whipple, 2014;Godard et al., 2014;Scherler et al., 2014).

Conclusion
Rates of rockwall slope erosion are defined for 12 catchments in northern India, NW Himalaya and range between 0.02 ± 0.004 and 7.6 ± 1.0 mm/a. Rockwall slope erosion largely outpaces local catchment-wide erosion and exhumation, and is sufficient to affect catchment sediment flux, glacier dynamics, and topographic change, such as the production of relief, the migration of catchment divides, and the reconfiguration of drainage basins.
Erosion rates become progressively lower with distance north from the MCT; up to two orders of magnitude difference in erosion rates are observed between the Uttarkashi, Kullu, Lahul-Spiti, and Ladakh and Shigar regions. Rather than a single control, rockwall slope erosion on a catchment-by-catchment basis is largely influenced by longstanding feedbacks between climate, tectonics, topography, and surface processes. The relative roles of these parameters are likely to vary over various spatial and temporal scales.
Our study demonstrates that, like records of denudation in the NW Himalaya, the broad trend in rockwall slope erosion cannot be fully explained by the distribution of precipitation. Instead, rockwall slope erosion can be considered part of the erosional response to tectonically driven uplift, the product of Indo-Eurasian convergence, and structural geology. The distribution and magnitude of erosion applicable to geomorphic (10 2−5 years) and geologic (10 6 years) timescales in the NW Himalaya therefore suggests that tectonics, rather than climate, provide a first-order control on landscape evolution. Our study also demonstrates the importance of lateral rockwall slope erosion via periglacial processes in helping set the pace of topographic change at catchment headwaters of high altitude and high relief mountain ranges, and the significance that localized erosion has for understanding wider landscape change.

Data Availability Statement
All supporting data can be accessed through the GFZ Data Services at https://dataservices.gfz-potsdam.de/ panmetaworks/review/26c9ad2dcd8617d7e18c8bb71b57b8e872a27eab88bcbf540b2d28db3d020d46/ Acknowledgments E. N. Orr thanks the University of Cincinnati for providing tuition and stipend to support this work as part of E. N. Orr's doctoral thesis and the processing of samples for 10 Be dating. E. N. Orr would like to thank PRIME Laboratories at Purdue University for a seed grant for AMS measurements (NSF-EAR0919759). E. N. Orr also extends thanks to National Geographic, the Geological Society of America and the Graduate Student Governance Association, University of Cincinnati for research grants to conduct fieldwork. E. N. Orr and S. Saha thank T. Dorje and Discover Ladakh for providing logistical support in the field. L A. Owen thanks the University of Cincinnati for providing support for fieldwork. The authors would like to thank the Associate Editor Harrison Gray, Brent Goehring, and four anonymous reviewers for their detailed, constructive comments and suggestions, which greatly improved the manuscript.