Tectonic controls of Holocene erosion in a glaciated orogen

. Recent work has highlighted a strong, worldwide, alpine glacial impact on orogen erosion rates over the last 2 Ma. While it may be assumed that glaciers increased erosion rates when active, the degree to which past glaciations influence Holocene erosion rates through the adjustment of topography is not known. In this study, we investigate the influence of long-term tectonic and post-glacial topographic controls on erosion in a glaciated orogen, the Olympic Mountains, USA. We 10 present 14 new 10 Be and 26 Al analyses which constrain Holocene erosion rates across the Olympic Mountains. Basin-averaged erosion rates scale with basin-averaged values of 5 km local relief, channel steepness, and hillslope angle throughout the range, similar to observations from non-glaciated orogens. These erosion rates are not related to mean annual precipitation or the marked change in Pleistocene alpine glacier size across the range, implying that glacier modification of topography and modern precipitation parameters do not exert strong controls on these rates. Rather, we find that despite spatial variations in 15 glacial modification of topography, patterns of recent erosion are similar to those from estimates of long-term tectonic rock uplift. This is consistent with a tectonic model where erosion and rock uplift patterns are controlled by the deformation of the Cascadia subduction zone.


Introduction
Before the onset of late Cenozoic cooling and glaciation, the topographic expression of mountain belts resulted from tectonic processes and the fluvial and hillslope processes which acted as the primary agents of erosion.High rock uplift rates in many of these ranges led to the buildup of topography and in some cases high relief, steep river channels and hillslopes, and commensurate high erosion rates (Willett, 1999).Because of the covariation between climate, topography, and rock uplift, erosion rates in fluvially dominated orogens have been shown to correlate with climatic and topographic metrics such as precipitation rate, relief, hillslope angle, and channel steepness via linear, non-linear, and threshold relationships (Ahnert, 1970;Montgomery and Brandon, 2002;Ouimet et al., 2009;DiBiase et al., 2010).The development of rugged mountain belts led to an increase in cooler, higher-elevation landscapes, which created the necessary conditions for alpine glaciers to form in the late Ceno-zoic.These glaciers possessed variable capacity to erode at the same rate as the rivers that existed before them and regional rock uplift rates.In many mountain ranges, glaciers appear to have accelerated erosion (Hallet et al., 1996;Shuster et al., 2005;Ehlers et al., 2006;Valla et al., 2011;Herman et al., 2013;Christeleit et al., 2017;Michel et al., 2018), while in other areas, glaciers do not appear to have changed erosion rates over the past few million years (Koppes and Montgomery, 2009;Thomson et al., 2010;Willenbring and von Blanckenburg, 2010).
As a result of Cenozoic climate change, the relationships between topographic metrics and observed Holocene (last ∼ 12 kyr) erosion rates in glaciated mountain ranges are more complex than in purely fluvial settings (Moon et al., 2011;Godard et al., 2012;Glotzbach et al., 2013).These poorly understood relationships are likely caused for two reasons: (1) glaciers reorganized previously fluvial channel networks and relief to create a landscape with their preferred geom-Published by Copernicus Publications on behalf of the European Geosciences Union.etry, radically changing the orogen topography (Whipple et al., 1999;MacGregor et al., 2000;Brocklehurst and Whipple, 2002, 2004, 2006Anderson et al., 2006;Adams and Ehlers, 2017), and (2) Holocene erosion rates may be dominated by transient signals as surface processes remove the topographic disequilibrium imposed by glacial erosion (Moon et al., 2011).
In light of the previous studies, what remains uncertain is how much (if any) signal of tectonic processes can be discerned from a heavily glaciated orogen and the degree to which common relationships between erosion and topographic metrics hold in post-glacial landscapes.Here we address this uncertainty and test whether Plio-Pleistocene glaciers have masked long-lived patterns of rock uplift as recorded by millennial-scale erosion rate estimates and modern topography.To do so, we have conducted a systematic study of basin-averaged erosion rates from 26  Al and  10 Be concentrations in modern river sediments from the Olympic Mountains, USA (Fig. 1).The Olympic Mountains are well suited for this study because the efficiency of Plio-Pleistocene glaciers was controlled by spatially variable glacial mass balance, and the orogen has been shown to contain a wide range of rock uplift rates.We use our new data, in addition to 10 Be concentrations from a previous study (Belmont et al., 2007), to investigate the spatial variations in erosion rates with respect to precipitation and characteristics of the modern topography including local relief, hillslope angle, and channel steepness.Further, we utilize 26 Al / 10 Be ratios and new modeling efforts to investigate the degree to which cosmogenic nuclide inventories can accurately constrain erosion rates in glaciated mountain ranges.

Background
The Olympic Mountains are part of a chain of mountain ranges that define the forearc high of the Cascadia subduction zone (Fig. 1a).This forearc high marks the topographic and structural apex of an accretionary wedge which formed between the North American plate and the subducting Juan de Fuca plate (Tabor and Cady, 1978).The core of the range is comprised of an essentially unmetamorphosed, homogenous assemblage of medium-to fine-grained greywacke interbedded with minor siltstone, mudstone, conglomerate, and basalt lenses.This group of rocks is referred to as the Olympic subduction complex and is located in the footwall of the Hurricane Ridge fault (Fig. 1).Pillow basalts, breccias, volcaniclastic rocks, and diabase make up the hanging wall of the fault, referred to as the Coast Range terrane.Sedimentological and bedrock cooling histories indicate accelerated rock uplift and unroofing of the range began around 17-12 Ma (Tabor and Cady, 1978;Brandon et al., 1998) 2001).These rates vary from ∼ 300 m Myr −1 at the fringes of the range to ∼ 800 m Myr −1 in regions close to the geographic center of the range, forming a concentric pattern.Previous interpretations of these erosion rates suggest that they are governed by rock uplift rates (Brandon et al., 1998;Batt et al., 2001;Pazzaglia and Brandon, 2001).In most orogenic wedges, the rock velocity field is governed by the sub-ducting plate geometry and convergence rate, and the pattern of accreted materials from the subducting plate (Willett, 1999).The Olympic Mountains are thought to be no different (Batt et al., 2001); however, the subduction zone dynamics are complicated by the significant arch in the subducting Juan de Fuca plate and the dome of accreted sedimentary units that make up the east-plunging Olympic anticline (Brandon and Calderwood, 1990).Indeed, it is likely that these nuanced characteristics of the subduction zone may be responsible for the observed concentric pattern in erosion-rock uplift rates (Brandon et al., 1998;Batt et al., 2001;Pazzaglia and Brandon, 2001;Bendick and Ehlers, 2014).
The Olympic Mountains have a general dome shape where the major drainages exhibit a radial pattern.The dome is asymmetric where the locus of highest topography lies to the northeast of the range divide (Fig. 1b).Plio-Pleistocene alpine glaciers carved large valleys in the core of the range (Porter, 1964;Montgomery and Greenberg, 2000;Montgomery, 2002;Adams and Ehlers, 2017).The largest glaciers which occupied the Hoh, Queets, and Quinault valleys all extended to the Pacific Ocean (Fig. 1b) (Thackray, 2001).Alpine glaciers were likely active in every valley of the Olympics (Porter, 1964); however, the size of the glaciers was highly variable, as the east-flowing glaciers would have been limited to the rugged core of the range by the Cordilleran ice sheet (see glacial deposits, Fig. 1b).This indicates that the west-flowing glaciers may have been nearly twice as long as those flowing east.Due to the W-SW prevailing wind direction and the effects of topography on precipitation patterns, mean annual precipitation values decrease from ∼ 6000 mm yr −1 in the southwest to less than 500 mm yr −1 in the northeast (Fig. 2a).This same precipitation gradient greatly influenced the Pleistocene equilibrium line altitude (ELA; the position where the ice flux in a glacier is at a maximum) and created an opposing pattern where the ELA increases at a rate of ∼ 25 m km −1 toward the northeast (Porter, 1964) (Fig. 1b), thus controlling the size and efficiency of alpine glaciers (Adams and Ehlers, 2017).The range was bordered to the north and east by the Cordilleran ice sheet (Fig. 1b) (Porter, 1964), which also likely restricted the size of alpine glaciers.This dichotomy in glacier size and erosional capacity is likely to have influenced the pattern in erosion rates during glacial times.A recent study by Michel et al. (2018) shows that bedrock cooling histories record a near doubling of exhumation rates around the time of the onset of glaciation (2 Ma).This effect is most pronounced on the west side of the range, where valley glaciers were larger.While the impact of Plio-Pleistocene glaciation on more recent erosion has not been previously quantified, the suggestion of significant glacial erosion would imply that Holocene erosion rates may not simply be a function of rock velocities as suggested by older erosion histories discussed above.

Topographic analysis
Our topographic analysis is based on the 10 m National Elevation Dataset provided by the United States Geological Survey (https://lta.cr.usgs.gov/NED,last access: 3 September 2014).Within this paper we calculate three topographic metrics which record relief at different spatial scales and controlled by different surface processes -hillslope angle, local relief, and channel steepness.Each metric also has strengths and weaknesses in quickly eroding glaciated ranges.There is good evidence that hillslope angle values can reach maximum values due to the limitations of the internal angle of friction of hillslope materials.In such high erosion areas, hillslope angle values become insensitive to changes in erosion rates (Schmidt and Montgomery, 1995).Local relief values may also be limited in glaciated ranges due to the buzzsaw effect, whereby efficiently eroding glaciers increase the area near their ELA and thus control mean elevations and restrict relief locally (Brozović et al., 1997;Meigs and Sauber, 2000).
Our local relief (R) map (Fig. 2b) was made by calculating the difference between the highest and lowest elevations within a 5 km diameter circular window.The local relief metric is designed to encapsulate the relief of hillslopes and channels.The size of this window captures the elevation difference between peaks and valley floors of medium-sized basins, but it is small enough to detect changes in the relief structure of large drainage basins.The hillslope gradient (S) map (Fig. 2c) was calculated by finding the steepest angle of descent across a 3 × 3 pixel (30 × 30 m) square window.
The channel steepness map (Fig. 2c) was created by adjusting channel gradients (S) (m m −1 ) by the non-linear change in downstream drainage area (A) (m 2 ) (Hack, 1957;Flint, 1974;Wobus et al., 2006): where k s is the channel steepness, and θ (dimensionless) is the channel concavity.Equation (1) normalizes slope values, for the concavity of the channel.For our calculations, we use θ = 0.45.A θ value of 0.45 has been shown to describe the concavity of fluvial systems in the Olympic Mountains (Adams and Ehlers, 2017).Since we utilize a single value of θ , we report normalized channel steepness index values (k sn ) (m 0.9 ).We used the Profile 51 tool (Wobus et al., 2006) to extract and analyze our river channels and calculated steepness values over 0.5 km reaches.To report a mean value for a basin we calculated the mean normalized channel steepness for all portions of a basin, which are governed by fluvial processes, which generally occurs at drainage areas > 1 km 2 .Normalized channel steepness index (k sn ) analysis, which quantifies channel relief, has been used successfully in a number of mixed fluvial and glacial landscapes as a fine- scale measure of the erosion potential of glacial-fluvial processes in a landscape (Montgomery, 2001;Brardinoni andHassan, 2006, 2007;Brocklehurst and Whipple, 2007;Robl et al., 2008;Hobley et al., 2010;Glotzbach et al., 2013;Adams and Ehlers, 2017).Many assumptions that are adopted in purely fluvial settings generally do not apply in mixed glacial-fluvial landscapes.For instance, in our study we do not require that the Olympic Mountains are in topographic steady state (where erosion and rock uplift at a point are balanced and therefore elevations remain constant over time), nor do we imply that our slope-area analysis relates directly to the processes of glacial incision, or that rock uplift rates need to be spatially uniform.We emphasize that the normalized channel steepness index provides a robust, geometric construct for understanding the importance of spatial changes in channel relief without demanding an understanding of all parameters within a specific incision law (fluvial or glacial).Furthermore, channel relief is likely to control the relief and hypsometry of landscapes, even in glaciated ranges (e.g., Adams and Ehlers, 2017).Unlike hillslope angle calculations, channel steepness values may be able to record changes in erosion-rock uplift rates in regions where hillslopes have reached a threshold (Ouimet et al., 2009).

Processing sediment samples and calculating erosion rates
Basin-averaged erosion rates were calculated from concentrations of cosmogenic nuclides ( 10 Be and 26 Al) in quartz sand from modern river basins throughout the Olympic Mountains (see Fig. S1 in the Supplement for sample location detail).Detrital cosmogenic techniques record the average erosion-denudation rate (we note that rates presented here incorporate both physical and chemical means of mass removal) integrated across the landscape upstream of the sample location (Brown et al., 1995;Bierman and Steig, 1996;Granger et al., 1996).Basins were selected to ensure a thorough sampling across precipitation, Pleistocene ELA, rock uplift, and topographic gradients.These basins are located within the Olympic subduction complex, where quartz is ubiquitous throughout the landscape.
Initial attempts to separate pure aliquots of quartz sand proved difficult due to the fine-grained nature of the lithologies found throughout the range.To reduce the need for aggressive hydrofluoric acid treatment, which would prematurely dissolve the quartz, we first disaggregated the 125-1000 µm sand fraction with a Selfrag, a high-voltage pulse disaggregater, at the University of Bern, Switzerland.From this stage on, samples were processed within the facilities at the University of Tübingen.After electronic disaggregation, sediments were re-sieved to 125-1000 µm and separated using a strong magnetic field and then cleaned in concentrated room temperature aqua regia for 24 h.Samples were further cleaned in boiling pyrophosphoric acid and then boiling sodium hydroxide at least 3 times.The quartz was then leached in 1 % hydrofluoric acid while in an ultrasonic bath for 1 week.A final leach was performed on the samples with concentrated hydrofluoric acid before spiking with beryllium.Samples were not spiked with aluminum.Beryllium and aluminum were separated, oxidized and loaded into cathodes for mass spectrometer analysis using established protocols (Von Blanckenburg et al., 1996).Native Al concentrations within samples were measured with an inductively coupled plasma optical emission spectrometer at the University of Tübingen.Beryllium and aluminum ratios were measured at the University of Cologne Centre for Accelerator Mass Spectrometry.
To calculate erosion rates, we followed the approach of Portenga and Bierman (2011), which simplifies each basin to a single point where the production rate is equal to the mean production rate of the entire basin, enabling the use of the CRONUS online calculator (Balco et al., 2008).Basin-averaged production rates were based on the elevation and latitude of each pixel in a basin using the scheme of Stone (2000).The effective elevation and latitude of each basin are the elevation and latitude values corresponding to this mean scaling factor (Table S1 in the Supplement).We calculated topographic shielding due to obstacles according to the equations of Dunne et al. (1999) and snow shielding from the equations of Gosse and Phillips (2001).Pixels under extant ice are assumed to be 100 % shielded.Our snow depth maps are based on satellite snow cover data that were calibrated by snow depth observations in the Olympic Mountains (see Supplement for more details).For CRONUS calculations, the following inputs were used: elevation flag = std, thickness = 1 cm, density = 2.7 g cm −3 , Be standard = 07KNSTD, and Al standard = KNSTD.We report erosion rates from the CRONUS calculator from the constant production rates determined by the constant production rate models of Lal (1991) and Stone (2000).To enable comparison between new and previous measurements, we recalculated erosion rates from seven sand samples within the Olympic Mountains previously reported by Belmont et al. (2007).

Isotopic equilibrium modeling
The application of detrital cosmogenic nuclide techniques as an estimator of basin-averaged erosion rates in post-glacial landscapes is not yet a common practice as there is a high potential for violation of the assumptions inherent to the calculation of erosion rates from nuclide concentrations.The assumptions that can be most problematic for a glacial terrain are the following: the eroding materials are in isotopic equilibrium, and modern river sediment is spatially and temporally representative of all sediment in the basin.To explore the nature of isotopic equilibrium we describe new modeling efforts in this section.Surface materials are in isotopic equilibrium when the production of cosmogenic nuclides is balanced by their removal through erosion and radioactive decay.In this state the concentration of nuclides in surface materials is steady over time.Since glacial ice intercepts and thus shields underlying material from cosmic radiation, previously and currently glaciated basins may violate the isotopic equilibrium assumption if ice was present recently, and erosion has not been able to remove the older shielding signal (Vance et al., 2003;Moon et al., 2011;Portenga et al., 2015).Therefore, interpreting cosmogenic nuclide concentrations as direct measurements of erosion in some glaciated landscapes can lead to overestimated rates (Gosse and Phillips, 2001;Vance et al., 2003;Portenga et al., 2015).
To test the hypothesis that our samples are in isotopic equilibrium we conducted a suite of numerical models to constrain the evolution of the concentration of cosmogenically produced nuclides at depth, starting at a time just after a period when ice completely shielded surface production.When the model starts, production occurs according to the equations of Anderson et al. (1996): where N is the cosmogenic nuclide concentration (atoms g −1 ), t is time (yr), N 0 is the inherited concentration of cosmogenic nuclides (atoms g −1 ), E is the erosion rate (cm yr −1 ), λ is the decay constant for 10 Be (1 yr −1 ), z is the depth below the surface (cm), P 0 is the 10 Be production rate at the surface (atoms g yr −1 ), is the attenuation length for cosmogenic nuclide production (g cm −1 ), and ρ is the material density (g cm −3 ).In our models the following values were used: λ = 4.99 × 10 −7 yr −1 , P 0 = 10 atoms g −1 yr −1 (though the results of this model are not sensitive to this value), = 160 g cm −2 , and ρ = 2.7 g cm −3 .Using a finite difference method, the model runs forward from the time since unshielding, and surface concentrations increase over time as production occurs, and deeply shielded materials are eroded from the top of the model.The concentration at the surface is compared to the steady-state value to assess the approach toward isotopic equilibrium.A range of www.earth-surf-dynam.net/6/595/2018/Earth Surf.Dynam., 6, 595-610, 2018 erosion rates which span the observed erosion rates in this study are tested.

Relationships between erosion rates and basin parameters
We performed non-linear least-square regressions on our new and existing erosion rate data.To provide a better sense of the distribution of topographic metrics within a basin, we provide box-and-whisker plots within our bivariate plots, though our regressions discussed in the following sections are based on mean statistics.We included the uncertainties in both variables by using a Monte Carlo sampling protocol.Goodnessof-fit values were determined by the mean square weighted deviation (MSWD).For well-fit data, MSWD values tend toward 1 within an uncertainty based on the degrees of freedom (based primarily on the number of samples).Elevated MSWD values are caused by the high degree of inter-sample variability and indicated at least one of the following is true: the two regressed variables are not highly correlated, a more complex function exists between the two variables, or that uncertainties are underestimated.
As a means to assess the relationship between erosion and hillslope processes we use a variation of the non-linear relationship proposed by Roering et al. (2001), which captures the effects of diffusive processes and landsliding: where E is the basin-averaged erosion rate (m Myr −1 ), K is a rate constant related to the diffusivity of the eroding material (m Myr −1 ), S is the basin-averaged hillslope gradient (m m −1 ), and S c is a critical slope at which soil flux approaches infinity (m m −1 ).We used a similar equation from Montgomery and Brandon (2002) to explore the relationship between erosion rates and 5 km local relief: where K is a different rate constant (m Myr −1 ), R is the basin-averaged local relief normalized by the diameter of the moving window (m m −1 ), and R c (critical relief) is a limit to the possible values of local relief normalized by the diameter of the moving window (m m −1 ).
Previous studies have suggested that channel steepness values can vary spatially according to a relationship with basin-averaged erosion rates through a stochastic threshold model of fluvial channel incision (DiBiase et al., 2010).Such a model generally produces a non-linear relationship.However, using a model based solely on fluvial incision in the Olympic Mountains would be misleading as the modern river channel likely still reflects the preferred geometry of Plio-Pleistocene glaciers (Adams and Ehlers, 2017).Instead, we implemented a least-squares power function regression to explore possible connections between erosion and normalized channel steepness, similar to other recent studies (Scherler et al., 2013): where C is the coefficient and p is the exponent.We used the same least-squares routine to analyze the relationship between erosion and precipitation (e.g., replace k sn with the mean annual precipitation in Eq. 5).

Topographic analysis
The topography of the Olympic Mountains is a mixture of high glacial cirque basins, wide and flat-floored valleys at low elevations, and steep landscapes in between.This juxtaposition of varied landscapes creates skewed and multimodal distributions of topographic metrics within drainage basins throughout the range (Tables 1 and 2, Fig. 2b, c and d).
While it is useful to report arithmetic means of basin statistics to simplify a landscape, it can often be difficult to constrain the significance of such means in the context of their relation to erosion rates.In complex landscapes not defined by uniform and steady surface processes, like the Olympic Mountains, normally distributed topographic metrics with good central tendency are unlikely, especially for metrics which capture fine-spatial-scale processes like those occurring at the scale of hillslopes and channel segments.To provide a better sense of these distributions, we have included simplified histograms next to our reported statistics in Tables 1  and 2. Because of this size limitation we are not able to calculate an accurate channel steepness value for one of the basins from a previous study, U-WC (Belmont et al., 2007).Basin-averaged hillslope angles are generally high, in most cases above 28 • , as is the standard deviation of hillslope angles within each basin (mean 2σ = 21 • ) (Table 2, Fig. 2c).Basin-averaged channel steepness values range between 23 and 181 m 0.9 and also have proportionally large standard deviations (mean 2σ = 89 m 0.9 ) (Table 2, Fig. 2d).Basin-averaged local relief values (calculated within a 5 km diameter window) range between 350 and 1443 m (Table 2, Fig. 2b).Relative to hillslope angle and channel steepness values, local relief values exhibit smaller variance within sampled basins (mean 2σ = 219 m).The lower-elevation basins on the western flank of the range, which evaded Last Glacial Maximum (LGM) alpine glaciers (Thackray, 2001;Belmont et al., 2007)  and the significant discrepancy in the size of alpine glaciers across the range divide, there is no difference in topographic metrics within the rugged core of the range across the divide.
There is a high degree of correlation between some basinaveraged precipitation values and basin-averaged elevation (Fig. 3a), as would be expected from the PRISM precipitation dataset, which includes an orographic precipitation model to do reanalysis simulations (Daly et al., 1994).This correlation is only strong on the western flank of the mountain where the topographic and precipitation gradients are smoothest.These same sub-group of basins also exhibit a strong correlation between basin-averaged hillslope angle and basin-averaged elevation (Fig. 3b).However, there is no correlation between elevation and precipitation or hillslope angle in the core of the range.There is good correlation between basin-averaged elevation and both basin-averaged local relief and channel steepness (Fig. 3c-d), across the range.

Cosmogenic basin-averaged erosion rates
While we present both 10 Be and 26 Al data (Table 3, see Tables S1 and S2 for complete nuclide analysis), we focus our analysis on erosion rates derived from 10 Be in this study (Fig. 2e) to provide a means of comparison to exist-ing data from the Olympic Mountains (Belmont et al., 2007).To a first order, basins located at elevations < 1000 m a.s.l.have been eroding at slow rates, all less than 240 m Myr −1 , whereas basins in the higher, rugged core of the range have higher erosion rates reaching over 1400 m Myr −1 (Table 3, Fig. 2e).We obtained the highest apparent erosion rates (> 1500 m Myr −1 ) from the flanks of Mount Olympus, whose drainages contain the largest extant glaciers in the Olympic Mountains (basins WA1519, WA1523, and WA1524).However, the low 10 Be concentrations (i.e., high apparent erosion rates) from Mount Olympus may be a signature of isotopic disequilibrium.Samples WA1519, WA1523, and WA1524 come from basins which likely contained thick ice the longest and still have small valley glaciers today.The 10 Be abundances for these three basins only range from 5 to 7 times the 10 Be blanks.These low abundances are likely caused by the shielding of rock and soil below glaciers.Such low measurements not only increase the internal uncertainty of the concentration calculation but also raise questions about the accuracy of the erosion rate calculation and interpretation.For this reason, we do not include these basins in our regression analysis.
Sample 26 Al / 10 Be ratios from the Olympic Mountains mostly vary between 8.5 ± 3.5 (2σ ) and 4.7 ± 1.6 (2σ ) (Ta-  ble 3, Fig. 4).Nearly all samples have 26 Al / 10 Be ratios that are statistically indistinguishable from the expected naturally occurring ratio (6.75, Balco et al., 2008) (Table 3, Fig. 4), suggesting that the sediments in our samples record a relatively simple erosion rate history over the integration time.
As such, there is no significant influence of reworking older sediments in our measurements.Furthermore, because our erosion rate calculations assume a natural production rate ratio of 6.75, and our measured ratios are mostly indistinguishable from this value, 10 Be and 26  statistically indistinguishable, though the 26 Al derived rates are much less precise (Table 3).Two samples from Mount Olympus basins, WA1519 and WA1523, have much lower ratios.
Snow shielding can reduce production rates and therefore reduce calculated erosion rates by up to 16 % in the core of the range, but only ∼ 3 % reduction is found in lower elevation basins on the western flank (Table S3).While it is difficult to assess our snow shielding estimates, we note the rela- tive effect on erosion rates is similar to those based on snowdepth measurements within other snowy orogens (Wittmann et al., 2007;Norton et al., 2010;Scherler et al., 2013).

Isotopic equilibrium modeling
As seen in Eq. ( 2) the likelihood of being in isotopic equilibrium for any cosmogenic radionuclide is mostly controlled by the time since deglaciation and the local erosion rate (assuming an inheritance of zero).Figure 5 illustrates that quickly eroding terrains more quickly remove ice-shielded materials; thus, these terrains can reach a new equilibrium state faster after the ice recedes.In fact, our model output indicates that at relatively low erosion rates (∼ 100 m Myr −1 ), terrains can achieve isotopic equilibrium in a few thousand years.These results suggest that the cosmogenic nuclide inventories from many glaciated landscapes on Earth could record accurate erosion rates (barring other complicating factors).

Relationships between erosion rates and basin parameters
Our best-fit curve (MSWD = 17) indicates the observed relationship between hillslope gradient and erosion is controlled by a critical slope value of 37 • and a rate constant of 250 m Myr −1 (Fig. 6b).These parameters fit our data considerably better than the previous boundary conditions suggested by Montgomery and Brandon (2002) for the Olympic Mountains (S c = 40 • , K = 500 m Myr −1 , MSWD = 54).
Our regressions also record a limiting local relief of 1820 m (K = 0.24 m Myr −1 , MSWD = 4.3) (Fig. 5a).These parameters are also different than those of Montgomery and Brandon (2002)  best-fit, nearly linear model (i.e., the exponent is 0.98) for the relationship between erosion and channel steepness (Fig. 5c).
The least-squares technique demonstrates that there is no strong linear or non-linear relationship between erosion and precipitation across the range (Fig. 5d).

Reliability of cosmogenic erosion rates in the glaciated Olympic Mountains
Our isotopic equilibrium model results show that even the slowest-eroding landscapes in the Olympic Mountains could achieve isotopic equilibrium within ∼ 3000 years (Fig. 5).Furthermore, the slowest-eroding basins from the western flank of the range did not contain valley glaciers during the LGM (Thackray, 2001); thus, these samples are even less likely to violate the isotopic equilibrium assumption.The most recently deglaciated portions of the range are in the rugged core, where erosion rates are also higher and where some landscapes can reach isotopic equilibrium in less than 1500 years (Fig. 5).
In landscapes where the cosmogenic nuclide inventories are a function of constant exposure or constant erosion, the ratio of 26 Al to 10 Be within sediments can be predicted based on the modeled (Lal, 1991;Balco et al., 2008) or measured (Corbett et al., 2017) ratios.Recent studies indicate that our samples should have natural 26 Al / 10 Be ratios of ∼ 6.75 (Balco et al., 2008), a value that is close to most of our measured ratios (Fig. 4).Therefore, we find it unlikely that sediment storage and reworking (e.g., from terraces or moraines) has violated our assumptions that modern sediments record a representative sample of all sediment in the basin.If anomalously low-concentration quartz was introduced into our samples via incision of older deposits (glacial or fluvial), we would expect to see depressed 26 Al / 10 Be ratios.Similar to previous work (Belmont et al., 2007) we have assumed that there is no risk to calculated erosion rates due to quartz infertility or proportional quartz sourcing from all parts of our basins in the Olympic Mountains.While there are some quartz-free lithologies in the range, these rocks are a minor occurrence the in Olympic subduction complex, and we have avoided sampling the Coast Range terrane completely.In locations where nested catchments are found, erosion rates are within error of each other, indicating a proportional sourcing of quartz from all parts of even the largest sampled basin (compare WA1526, DEN101, and DEN106).

Interpreting relationships between erosion and basin metrics
In landscapes with high fluvial and/or glacial erosion, soil production and hillslope transport may not be able to adjust to channel incision.In such a case, hillslope angles steepen and tend toward a threshold that is controlled by the strength of the material (Schmidt and Montgomery, 1995).Once hillslopes reach such a threshold, increases in erosion can only occur with a commensurate increase in hillslope failure (Burbank et al., 1996), and the forms of these hillslopes are no longer sensitive to changes in erosion.However, the gradients of channels in these steep landscapes are generally much lower than the internal angel of friction, and, as such, still have the capacity to adjust to increases in erosion rate.Therefore, it has been suggested that the morphology of channels is a more robust metric to detect erosion rate variations, as compared to the steepness of hillslopes (Ouimet et al., 2009;DiBiase et al., 2010).
Our data show that basin-averaged hillslope gradients cease to increase in basins eroding faster than ∼ 300 m Myr −1 (Fig. 6b).This limit has been observed in many other landscapes around the world (Montgomery and Brandon, 2002;Binnie et al., 2007;Ouimet et al., 2009;DiBiase et al., 2010).Basin-averaged hillslope angle values tend to reach a maximum around 34 • , as also shown by Montgomery (2001) using 100 km 2 grids across the Olympic range.The extent to which these threshold hillslope angles are indicative of rock uplift rates or glacial modification is not completely clear.While it is possible that the weak lithologies and fast erosion rates of the Olympic Mountains may be setting these threshold hillslopes, it has also been documented that hillslope angles have likely been increased throughout the range via glaciers widening valleys (Montgomery, 2002;Adams and Ehlers, 2017) or eroding headward and migrating ridge tops (Adams and Ehlers, 2017).
Similarly, basin-averaged local relief values do not exceed ∼ 1350 m despite increasing erosion rates (Fig. 6a).This apparent threshold relief may be due to the influence of the glacial buzzsaw effect, whereby efficiently eroding alpine glaciers have controlled the mean and range of elevations during the Plio-Pleistocene.

Orogenic processes governing erosion rates
With our new and previously published erosion rates, we have made several important observations in the previous sections that we elaborate on below.These observations are the following: 1.There is no relationship between precipitation and Holocene erosion rates across the range (Fig. 6d).
2. Basins with similar topographic characteristics have equivalent erosion rates, even across the range divide where glacial size changed drastically (compare black and grey samples in Fig. 6).
3. It is apparent from our regressions that there are nonlinear relationships between local relief, channel steepness and hillslope angle, and Holocene erosion rates (Fig. 6).
In tectonically active and previously glaciated mountain ranges there are three common orogenic processes that are most often suggested to dominate Holocene erosion rate patterns: climate gradients (Carretier et al., 2013;Olen et al., 2016), glacial modification of the landscape (Moon et al., 2011;Glotzbach et al., 2013), and patterns of tectonic rock uplift (Scherler et al., 2013;Godard et al., 2014;Adams et al., 2016).In the following we explore the relevance and applicability of these explanations to our dataset.First, we find it highly unlikely that a precipitation gradient similar to the modern has a significant control on recent erosion rates.There is no clear relationship between modern precipitation and erosion rates (Fig. 6d).Even in the neighboring glaciated Cascade Range, ∼ 70 km to the east of the Olympic Mountains, where the modern precipitation gradient is not as large, there is a strong linear relationship indicating erosion scales with precipitation over diverse timescales, thus making it an important condition for setting Holocene and older erosion rates (Reiners et al., 2003;Moon et al., 2011).
Second, our data do not indicate that destabilization of the landscape via glacial incision has played a primary role in setting the Holocene erosion pattern.Despite the significant gradients in the estimated Pleistocene ELA positions (Porter, 1964;Fig. 2e) and the change in glacier size, and the estimated higher Quaternary erosion rates (Michel et al., 2018) across the range divide, there is no statistical difference between the erosion rates across the range divide for basins of similar topographic characteristics (Figs. 2, 5a, b, c), and there is a very weak correlation between ELA and erosion rates (Fig. S2).However, it has been inferred that Plio-Pleistocene glaciers widened valleys in the Olympic Mountains (Montgomery, 2002;Adams and Ehlers, 2017), which led to the lengthening and steepening of hillslopes throughout the range.In the nearby Cascade Range, similar valley widening has led to hillslopes with higher likelihoods for failure (Moon et al., 2011).Unlike in the Olympic Mountains, findings from the Cascades indicate that the range was heavily influenced by glacial incision to an extent that the topographic form largely reflects relict glacial processes, and as a result, Holocene erosion rates are more likely to be correlated with precipitation in these landscapes further from equilibrium.Our analysis suggests that the changes in the landscape due to Plio-Pleistocene glaciation in the Olympic Mountains likely only steepened relatively small areas of hillslopes of landscapes relative to the already steep conditions imposed by high rock uplift and erosion rates.Similarly, glacial incision may have only influenced relatively small portions of the channel network and range relief, which might appear as threshold values of channel steepness and local relief, or simply to make the distributions of these parameters within a basin more complex.Therefore, the landscapes examined in the Olympic Mountains may have been only moderately perturbed by Plio-Pleistocene glacial incision, and they may still record a relatively close balance between recent erosion rates and rock uplift.The balance between post-glacial erosion rates and longer-lived rock uplift rates depends on whether post-glacial climate conditions (e.g., increase or decrease in precipitation) or topographic perturbations (e.g., hillslope steeping or channel shallowing) have changed the activity of extant surface processes.There are many examples of ranges where there have been significant changes to topography during glaciation (Montgomery, 2001;Brardinoni andHassan, 2006, 2007;Brocklehurst and Whipple, 2007;Robl et al., 2008;Hobley et al., 2010;Glotzbach et al., 2013), and erosion during and after glaciation (Reiners et al., 2003;Moon et al., 2011;Christeleit et al., 2017), and others where such changes are not clearly observed (Thomson et al., 2010).More generally, these changes were explored in a coupled ice dynamiclandscape evolution model testing the modification of topography and erosion rates due to alpine glaciation (Yanites and Ehlers, 2012).The results of these numerical models indicate that the degree of erosion change before and after glaciation is a function of regional temperature and the rock uplift rate.These two parameters control the glacier's ability to concentrate elevations at or near the ELA where ice erodes most efficiently.If too much or too little of the landscape lies above the ELA, then glacial erosion is not very efficient and little topographic perturbation occurs.In these landscapes, erosion rates may change during glacial periods, but interglacial erosion rates return to near rock uplift rates, as before.In the cases where glaciers were highly effective agents of erosion, relief (on hillslopes or channels) is reduced during glaciation (Whipple et al., 1999;Adams and Ehlers, 2017), and post-glacial erosion rates can be lower than pre-glacial, and vice versa.As such, there is a sweet spot within mountain range conditions where glaciers are more efficient than rivers and hillslopes; furthermore, even when glaciers are efficient it cannot be assumed how post-glacial rates might change.To put it another way, it should be assumed that landscapes will respond differently to alpine glaciation depending on climate and topographic conditions before, during, and after glaciation.
Figure 7 illustrates the similarity of the trends in other rock uplift rate proxies and the cosmogenic erosion rates presented here in a direction parallel to the convergence across the Olympic Mountain range.When our new Holocene erosion rate pattern is compared with older patterns of estimated rock uplift rates (Fig. 7), there are a few apparent mismatches.In some locations, our rates are higher or lower than rock uplift rates (as might be expected in post-glacial landscapes), but overall the pattern of increasing rates from the flanks to the core of the range is consistent between these datasets.Adams and Ehlers (2017) and Michel et al. (2018) proposed that a spatial pattern of rock uplift similar to the one described above was consistent with the observations of the bend in the subducting Juan de Fuca plate at the Cascadia subduction zone and the dome of accreted sediments in the core of the Olympic Mountains, which form the east-plunging Olympic anticline (Brandon and Calderwood, 1990).This pattern of focused rock uplift and erosion is also predicted for the geometry of the curved subducting Juan de Fuca plate (Crosson and Owens, 1987;Bendick and Ehlers, 2014).We suggest this long-lasting pattern is primarily con-trolled by tectonic forces, while the Plio-Pleistocene alpine glaciers of the Olympic Mountains have not radically altered the topography enough to drastically change the pattern of erosion.

Conclusions
Taken together, we suggest that the Holocene erosion rates (Fig. 2e), mean elevation, local relief (Fig. 2b), and channel steepness (Fig. 2d) observed in the Olympic Mountains most closely record a rock uplift pattern that increases from the low-relief flanks to the rugged core of the range (Fig. 7), similar to what has been shown in other datasets (Brandon et al., 1998;Batt et al., 2001;Pazzaglia and Brandon, 2001).Our interpretations are in line with previous authors who have highlighted the importance of subduction zone dynamics for setting the pace and pattern of erosion in the Olympic Mountains (Brandon and Vance, 1992;Brandon et al., 1998;Batt et al., 2001;Pazzaglia and Brandon, 2001;Stolar et al., 2007).This result may be unexpected given the glacial impact that has been previously documented throughout the range (Porter, 1964;Montgomery and Greenburg, 2000;Montgomery, 2002;Adams and Ehlers, 2017) and further described in this article.However, the Plio-Pleistocene glacier impact on small and range scales may have been limited in the Olympic Mountains because large portions of the range may have already been at maximum hillslope angle conditions (i.e., at the critical threshold) before glaciation (Montgomery, 2001), or a small proportion of the range was focused at the ELA during glacial periods.As such, post-glacial erosion rates exhibit the same spatial patterns and magnitudes as longer-term estimates.The alpine glaciers of the Olympic Mountains have left behind scenic, sculpted landscapes, but these landscapes may have not been as significantly altered as once thought, at least not enough to drastically change post-glacial erosion.

Figure 1 .
Figure 1.Topographic and geologic features of the Olympic Peninsula, Washington State, USA.(a) Simplified geology based on Brandon et al. (1998).The relative velocity of the Juan de Fuca plate toward the North American plate is ∼ 36 mm yr −1 with a bearing of ∼ 54 • (DeMets and Dixon, 1999).Red box denotes the extent of (b).HRF -Hurricane Ridge fault.Grey lines outline the coast of Washington State.Dashed black line is the cross-section line for Fig. 7. (b) Elevation map of the Olympic Mountains.Ice, including extant alpine glaciers, is marked in blue.The limit of the Puget Lobe of the Cordilleran ice sheet (Vashon glaciation) is marked with a black dashed line (Porter, 1964).Undifferentiated Quaternary alpine glacial till and alluvial deposits are marked in grey and yellow, respectively (Washington Division of Geology and Earth Resources, 2010).Contours of Pleistocene equilibrium line altitudes (ELAs) from Porter (1964) are denoted by white lines (values shown in meters above sea level).Black box denotes the extent of Fig. 2 panels.A red dot marks Mount Olympus.

Figure 2 .
Figure 2. Attribute and erosion maps of the Olympic Mountains.Solid outlines denote the boundaries of sampled basins.High rugged core outlined in black/white dashed lines.(a) Mean annual precipitation (MAP) map based on PRISM data (Daly et al., 1994).Open and closed circles mark new and previously published sample locations, respectively.(b) Local relief (5 km relief) map.(c) Hillslope angle map.(d) Normalized channel steepness (k sn ) map plotted for accumulation areas > 2 km 2 .(e) Basin-averaged erosion rate map.Range divide marked in magenta.Equilibrium line altitude (ELA) contours from Porter (1964) are in black.
, have the lowest topographic metric values of the sampled basins (eight basins: mean R = 544, mean S = 23, mean k sn = 43).The mean values from the eight glaciated west-side basins and the six glaciated eastside basins are effectively the same: mean R = 1296, mean S = 31, mean k sn = 151; and mean R = 1239, mean S = 31, and mean k sn = 143, respectively.Despite the rain shadow

Figure 3 .
Figure 3.Comparison of the mean basin elevation with other basin averaged metrics.(a) Mean annual precipitation.(b) Hillslope angle.(c) Local relief (using a 5 km diameter circle).(d) Channel steepness.

Figure 4 .
Figure 4. Erosion island plot for new Olympic Mountain samples.Each sample is represented by a 2σ error ellipse.Grey ellipses mark samples with poor 10 Be measurements.See text for discussion.

Figure 5 .
Figure5.Predicted evolution of landscapes toward isotopic evolution after deglaciation.Black and grey lines denote 10 % and 5 % contour intervals, respectively.Basin-averaged erosion rates from the Olympic Mountains are shown on the right.Basins marked in red were not glaciated during the Last Glacial Maximum (LGM).

Figure 7 .
Figure 7.Comparison between estimated erosion and relief across the Olympic Mountains.(a) Elevation and climate data across the Olympic Mountain range parallel to the direction of tectonic convergence (∼ 54 • ).Maximum and mean elevations are shown in thin and thick black lines, respectively.Mean annual precipitation data(Daly et al., 1994) are shown in blue.Equilibrium line altitude (ELA) data(Porter, 1964) are represented by a red trend line.(b) The blue circles show estimated rock uplift rates from apatite fission track data fromBrandon et al. (1998).Black circles are rock uplift rate estimates fromPazzaglia and Brandon (2001) based on river terrace incision.Basin-averaged erosion rates in this study are shown in red circles with bars denoting the width of the basins.See Fig.1afor cross-section location.

Table 1 .
Belmont et al. (2007)ristics.Mean equilibrium line altitude (ELA) based on estimates fromPorter (1964).2σ= 2 standard deviations on the mean.Curves represent simplified histograms with normalized counts.See labels below each column for minimum and maximum bin values.Basins in italics are fromBelmont et al. (2007).

Table 2 .
Belmont et al. (2007)sion processes.2σ= 2 standard deviations on the mean.Curves represent simplified histograms with normalized counts.See labels below each column for minimum and maximum bin values.Basins in italics are fromBelmont et al. (2007).Values exclude data from ice-covered regions.