Detection of paraglacial sediment supply using detrital 10Be in postglacial landscapes of southwest British Columbia
Funding information: The work was supported by a grant from the British Columbia Salmon Restoration Innovation Fund (BCSRIF) to J.V. and 7 others, and a Natural Science and Engineering Research Council of Canada (NSERC) Discovery Grant and Accelerator Supplement to J.V. E.H.D was supported by a Natural Environment Research Council (NERC) Independent Research Fellowship (NE/X017567/1).
Abstract
The legacy of glaciation persists for tens to hundreds of thousands of years in postglacial landscapes, where transient storage and release of paraglacial sediment masks signals of primary landscape denudation (i.e., bedrock incision). The timescales over which glacial legacies persist are difficult to quantify without detailed information on fluvial sediment load or landscape denudation. Here, we present 33 new detrital 10Be cosmogenic radionuclide analyses from the Fraser River basin in western Canada. We combine 10Be concentrations with paraglacial terrace distributions and present a sediment mixing model to assess the extent to which paraglacial terrace sediment contributes to fluvial sediment loads. Estimated basin-averaged denudation rates are ~0.23 mm yr−1 and largely invariable along the Fraser Canyon, a 375-km bedrock influenced reach, despite a doubling in drainage area and extensive paraglacial terrace distribution. Denudation rates are poorly correlated with landscape morphometry (slope, channel steepness) and climate. To reproduce patterns in 10Be concentrations along the canyon, terrace sediment contributions must be limited to ~15% of the total flux. We attribute the low terrace inputs to limited connectivity between the terrace fill and the Fraser River channel, which is now incised into bedrock. Longitudinally invariant specific sediment yields are likely a consequence of limited floodplain storage for sediment that is instead transported through the Fraser Canyon to the delta. Sediment derived from bedrock denudation exceeds sediment inputs from paraglacial terrace deposits in the lower Fraser River. Despite widespread prevalence of paraglacial terrace fill in the landscape, our results indicate that basin-scale paraglacial effects diminish once channels re-incise into bedrock and terrace fills become disconnected from lateral river channel erosion. These findings shed new light on the processes controlling the timescales of paraglacial effects on modern river sediment loads and wider postglacial landscape evolution.
1 INTRODUCTION
Modern sediment yields in previously-glaciated landscapes are thought to be dominated by reworked and incised paraglacial deposits (e.g., Church & Ryder, 1972; Fame et al., 2018), which produce downstream changes in specific sediment yield that do not reflect modern patterns of bedrock denudation (Church & Slaymaker, 1989). Sediment yield in some paraglacial landscapes is primarily influenced by direct glacially conditioned sediment release, such as debuttressing of bedrock valley walls (Ballantyne, 2002). However, in other paraglacial landscapes, the reworking of paraglacial sediment stored in upland tributary valleys is the main contributor to fluvial sediment loads (Ballantyne, 2002). Intermittent reworking of paraglacial sediment in upland tributaries can result in delayed peaks in paraglacial sediment release into lowland river systems (Harbor & Warburton, 1993), while changes in climate and base-level may modify patterns in paraglacial sediment reworking and contributions of paraglacial sediment to the modern fluvial system sediment load (Ballantyne, 2002). Understanding the timescales over which paraglacial sedimentation persists or is detectible within modern fluvial systems remains challenging.
Incision of Late Pleistocene paraglacial terrace fill by rivers in south-central British Columbia, Canada, is reflected by downstream increases in specific sediment yield (mass per unit area of upstream basin per time), implying rivers are still responding to sediment supply associated with the last glaciation which ended more than 10 kyr ago (Church & Slaymaker, 1989). However, changes in specific sediment yield are longitudinally variable with increasing drainage areas. As channels integrate larger upstream drainage areas (>104 km2), patterns of increasing specific sediment yield reverse and instead conform with the conventional model of decreasing specific yield with increasing area (e.g., Vanmaercke et al., 2011; Vanoni, 2006; Walling, 1983), because sediment generated by erosion of paraglacial terrace fill in headwater streams is redeposited downstream in main trunk channel valleys (e.g., Church, 2017). While paraglacial sedimentation (e.g., channel and floodplain aggradation) in smaller upland basins may have ended by ~6 ka (Church & Ryder, 1972), sedimentation in larger basins is thought to have continued over much longer timescales (Church & Slaymaker, 1989). Due to spatially limited gauging of the modern-day sediment load, it is difficult to constrain rates and patterns of sediment erosion, transport and deposition to explore how paraglacial sediment is redistributed in postglacial landscapes. Hence, it remains a challenge to determine whether sediment yields measured >10 kyr following deglaciation reflect bedrock denudation or whether remobilisation of paraglacial sediment still dominates.
The concentration of cosmogenic radionuclides (CRNs) such as 10Be in quartz-rich river sediment is commonly used to estimate upstream rates of landscape denudation, based on the assumption of a steady-state denudation rate that is averaged over the entire upstream basin (Bierman & Steig, 1996; Brown et al., 1995; Granger, Kirchner, & Finkel, 1996). Spatial patterns of CRN-derived basin-averaged denudation rates are routinely used to examine feedbacks between tectonics, climate and topography in mountainous landscapes (e.g., DiBiase et al., 2010; Portenga et al., 2015; Portenga & Bierman, 2011; Scherler, Bookhagen, & Strecker, 2014) over centennial to million-year timescales. Basin-averaged denudation rates can also be converted into longer-term sediment flux estimates when combined with upstream basin area (e.g., Kirchner et al., 2001; Lupker et al., 2012). Assessment of basin-averaged denudation rates relies on a number of assumptions, including that timescales associated with sediment transfer are vastly exceeded by timescales associated with denudation on hillslopes (von Blanckenburg, 2005). Glaciogenic sediments derived from recently deglaciated sub- or proglacial environments are often characterised by very low CRN concentrations (e.g., Stutenbecker et al., 2018) which when mixed into sediment loads of larger alpine fluvial systems, can yield basin-averaged concentrations that are not representative of actual headwater denudation rates. Predicting the initial acquisition of CRN in sediment generated in recently glaciated environments is difficult, as sediments may also contain inherited CRN concentrations from previous interglacial periods (e.g., Wittmann et al., 2007). These complex patterns of intermediate sediment storage within postglacial landscapes and CRN inheritance produce CRN concentrations in modern river sediments that may be decoupled from modern landscape denudation rates (e.g., Delunel et al., 2013; Jautzy et al., 2024; Wittmann et al., 2007).
There has been more success in using spatial patterns of 10Be in glaciated or previously-glaciated basins to characterise sources of glaciogenic sediment in modern river sediment loads (e.g., Delunel et al., 2013; Stutenbecker et al., 2018; Wittmann et al., 2007) and understand non-uniform patterns of erosion where evidence of glaciation and glacially-conditioned processes are preserved within landscapes (e.g., Fame et al., 2018; Norton et al., 2010). Sediment mixing and unmixing (i.e., inverse) models are commonly used to disentangle the relative contributions of distinct sources of sediment with known 10Be concentrations (e.g., Norton et al., 2010). For example, inversion of 10Be concentrations measured in river sands in a glaciated basin in central Nepal enabled quantification of the fluvial and glacial contributions to denudation (Godard et al., 2012). Standard assumptions in deriving basin-averaged denudation rates are often invalid in glaciated or postglacial landscapes, but attention to understanding factors that drive downstream evolution in CRN concentrations may allow contributions from different geomorphic processes to be disentangled (Godard et al., 2012).
The Fraser River basin drains nearly one quarter of British Columbia, Canada and discharges into the Pacific Ocean. It is commonly assumed that sediment currently transported by the Fraser River is sourced from Pleistocene paraglacial terraces, owing to a downstream increase in specific sediment yield and widespread occurrence of paraglacial terraces throughout the basin (Church & Slaymaker, 1989). Little is known about centennial to millennial scale denudation rates of the Fraser River basin, or whether signals of landscape denudation remain masked by reworked paraglacial terrace sediment supply more than 10,000 years after deglaciation. Models of paraglacial sedimentation consider changes in sediment yield following the onset of deglaciation and predict that sediment yield should eventually return to pre-glacial conditions that reflect rates of bedrock denudation (e.g., Ballantyne, 2002; Church & Slaymaker, 1989). The timescales over which paraglacial sedimentation cycles occur in the Fraser River basin have not been quantified, and it is unclear whether stores of paraglacial sediments must be fully exhausted from landscapes before reattaining sediment yields that are reflective of bedrock denudation rates. We explore this idea further, and hypothesise that where channels become sufficiently entrenched into paraglacial terrace material and then the underlying bedrock, channels should become laterally disconnected from terrace fills perched on adjacent valley sides. Under these conditions, downstream patterns in sediment yield should reflect changes in lateral connectivity between the modern channel and terrace fill, and signals of bedrock landscape denudation could emerge without fully removing paraglacial sediment from the landscape.
Here, we use detrital 10Be concentrations to both examine paraglacial sediment supply and interpret modern landscape denudation processes in the Fraser River basin. To do this, we present 33 new measurements of detrital 10Be concentrations in river channel and terrace deposit sediment from the Fraser River valley and explore spatial patterns in basin-averaged denudation rate estimates. We first assess potential drivers of denudation by testing whether patterns of basin-averaged denudation rates are related to landscape morphometrics and climate. We then explore downstream patterns in absolute CRN concentrations in samples collected along the main stem Fraser River and develop a CRN mixing model using mapped paraglacial terraces to quantify the potential impact of paraglacial terrace sediment supply on basin-averaged CRN concentrations. Finally, we use these findings to assess modern paraglacial sediment contributions to the lower Fraser River basin.
2 SETTING
The Fraser River is the largest river draining from Canada into the Pacific Ocean, with a contributing area of ~232,000 km2 (Rennie, Church, & Venditti, 2018). Much of the drainage area is occupied by the low-relief Nechako, Fraser and Thompson Plateaus (~1,000–1,600 m in elevation), which are bound by the high-relief Rocky Mountains to the east, and Coast Mountains in the west and south (Figure 1). The modern channel is incised by up to 500 m below the plateau surface within a broader 3–4 km wide valley (Eyles & Clague, 1991). The plateaus have subhumid to semiarid climates and lie in the rain shadow of the Coast Mountains, resulting in dry summers and notable snow accumulation (~391–924 mm daily mean snow water equivalent; Déry et al., 2014) in the winter. Synchronous melting of the snowpack on the plateaus results in seasonal high flows (the freshet) during late May to early July each year. Downstream of Marguerite, the Fraser River becomes deeply incised into Quaternary-aged alluvial terrace sediment (Figure 2) as it passes through a clear physiographic transition into the Coast Mountains. Where the channel is most deeply incised, exposures often reveal a complex sequence of basal Pleistocene tills and lacustrine sediments that are capped by Holocene sediments (e.g., Rennie, Church, & Venditti, 2018). These deposits are a mixture of locally derived sediments that vary with the spatial distribution of bedrock lithologies along the Fraser River and far-travelled granitic cobbles (e.g., Lian & Huntley, 1999; Lian & Hicock, 2001).
Downstream of Soda Creek, British Columbia, the Fraser River flows through the 375 km-long Fraser Canyon to Yale where the river alternates between being bedrock constrained (bedrock on both banks; 26% of its length), partially constrained (bedrock on one bank; 29% of its length), and long reaches where there is no bedrock on the banks (45% of its length) (Rennie, Church, & Venditti, 2018). The reaches where no bedrock is exposed are composed of glaciofluvial terraces and colluvial deposits with some alluvial material deposited in bars and at tributary junctions. The non-bedrock reaches are lined with boulders sourced from glaciofluvial terraces, talus and debris flow deposits washed clear of finer materials, but bedrock is probably not far below sediment deposits in the Fraser Canyon, based on the gradient and morphology of the river (Rennie, Church, & Venditti, 2018). At Yale, British Columbia, the river begins to transition from bedrock-dominated to an alluvial channel, and that transition is complete at Hope, British Columbia, where the channel is anabranching and gravel-bedded. At Mission, British Columbia, the river forms a single-threaded sand bed channel that flows into the Pacific Ocean at the Fraser River delta (Venditti & Church, 2014). Delta formation is thought to have initiated following the last glaciation at ~9 ka (Clague et al., 1983; Williams & Roberts, 1989).
The headwaters of the Fraser River are characterised by Upper Proterozoic and Cambrian clastic and carbonate sedimentary rocks deposited on and adjacent to the North American craton (Cui et al., 2017). As the channel enters a bedrock confined section at Soda Creek (Figure 1), the underlying bedrock is largely composed of Late Triassic to Early Tertiary volcanic and sedimentary rocks associated with multiple accretionary phases. The Coast Mountains, including the Coast and Lillooet Ranges, are a region of metamorphic and granitic rocks (Mid-Cretaceous to Tertiary in age) (Cui et al., 2017).
40Ar/39Ar dating of volcanic dams has placed an upper age constraint of 0.76 Ma on a significant drainage reversal that rerouted an ancestral Fraser River, which originally flowed north into the Arctic Ocean, to its current configuration (Andrews et al., 2012). The flow reversal was attributed to stream capture following accelerated uplift of the Coast Mountains that initiated ~4 Ma (Andrews et al., 2012). Thermochronological dating of rock exhumation in the Coast Mountains suggest that at least 2 km of uplift has occurred within the past 10 Ma, with the majority likely to have occurred within the past 2–4 Ma in response to glacial erosion and isostatic response of the landscape (Ehlers et al., 2006; Farley, Rusmore, & Bogue, 2001).
2.1 Glacial legacy
The Cordilleran Ice Sheet existed in southern and central British Columbia during the Wisconsin (or Fraser) Glaciation between 19 and 13.5 ka (Ryder, Fulton, & Clague, 1991). The timing of ice sheet expansion and recession was asymmetric, with ice build-up starting at ~29 ka and reaching a maximum extent between ~14–15 ka (e.g., Clague, 1981; Clague et al., 1989; Clague & Ward, 2011). The Fraser River basin was largely ice-free by ~11.5 ka (Ryder, Fulton, & Clague, 1991). Rapid deglaciation is argued to have resulted in a relatively short ~1 kyr phase of glacioisostatic uplift in southwest British Columbia, with adjustment largely complete by ~12 ka (Shugar et al., 2014). During deglaciation, ice-dammed lakes developed and a significant outburst flood occurred at ~10.5–12.5 ka from glacial Lake Fraser (Clague et al., 2021). Incised Pleistocene terrace fill was rapidly transported by floodwaters >300 km downstream and deposited on the proto-Fraser River Delta (Clague et al., 2021). Detailed mapping and OSL dating of terraces in the Big Bar region downstream of the failed ice-dam places an age constraint of ~11.3 ± 1.5 ka on paired terraces ~180 m above the modern river level, which was linked to sediment deposition and subsequent incision associated with the outburst flood (Gingerich, 2021). Since the outburst flood, the Fraser River has incised through ~180 m of ancestral floodplain and glaciofluvial sediments, reaching bedrock during the last ~4 kyr, with incision rates of up to 30 mm yr−1 during wetter middle Holocene conditions (Gingerich, 2021).
3 METHODS
3.1 Cosmogenic radionuclide sample collection
Our study reach encompasses 320 km of the Fraser River, a substantial part of the 375-km-long Fraser Canyon, from the Chilcotin River confluence to Yale, British Columbia. We collected 1–2 kg of sand from the surface of modern bars on the Fraser River main stem (11 samples) and tributary channels (10 samples) in the Fraser Canyon during medium- to low-flow conditions in 2021 and 2022. An additional 13 samples were collected from two vertical depth profiles from paraglacial terrace deposits at Iron Canyon and High Bar Canyon. We conducted repeat sampling at two main stems and one tributary site in 2021 and 2022 to assess temporal variability in 10Be concentrations. To ensure that sand was well-mixed, we avoided sampling bars with clear lateral sediment inputs from tributaries within ~0.5–1 km upstream. Depth profile samples were collected from the tops of paraglacial fills where exposed faces were stable and accessible (e.g., Figure 2c). Vertical faces were cleared of loose material, and 6–7 samples were collected along a profile extending ~2.5 m below the terrace surface. Plastic tubes of a known volume were pounded into the sediment and weighed to determine the bulk density of the paraglacial sediment.
3.1.1 Sample preparation and analysis
Sediment samples were dried before sieving to extract grain sizes in the 250–850 μm fraction prior to sample preparation and 10Be chemistry conducted at the University of Massachusetts Cosmogenic Nuclide Laboratory following Larsen et al. (2023) and Quirk et al. (2024). All samples underwent magnetic roller separation before non-quartz mineral phases were removed using a combination of etches in dilute hydrochloric acid, lithium heteropolytungstate heavy liquid separation, surfactant separation and repeated etching in 2% hydrofluoric acid following modified procedures from Kohl and Nishiizumi (1992). Phosphoric acid was used to isolate quartz in finer-grained and quartz-poor samples (Mifsud, Fujioka, & Fink, 2013), including the depth profile samples. We tested aliquots of each sample to determine quartz purity via ICP-OES measurement of Al, Be, Ca, K, Mg, Na, Fe and Ti. We added ~250 μg of 9Be carrier to each sample and a procedural blank in each batch of samples. The Be fraction of each sample was chemically isolated using ion chromatography following Ditchburn and Whitehead (1994) and Stone (2001), calcined by heating in a quartz crucible with a gas torch until BeO glowed orange for 1 minute, and packed with Nb powder into targets for 10Be/9Be measurement by accelerator mass spectrometry (AMS) at the Lawrence Livermore National Laboratory. The ratio of 10Be/9Be was measured against the AMS beryllium standard 07KNSTD with a known ratio of 2.85 × 10−12 (Nishiizumi et al., 2007). The minimum and maximum 10Be/9Be ratios of 10 procedural blanks run in the same batches that included the samples were 1.88 × 10−16 and 4.16 × 10−16, respectively.
3.1.2 Denudation rate calculations
Owing to variable geology in the Fraser River basin, parts of the basin underlain by non-quartz bearing lithologies were excluded from calculations, which accounted for up to 40% of the basin area (Table 1). Quartz-bearing lithologies were identified using British Columbia Digital Geology shapefiles (1:50,000 to 1:250,000 scale) (Cui et al., 2017). Where there was ambiguity as to whether lithologies were quartz-rich (e.g., simply described as ‘metavolcanic’), these were categorised as non-quartz bearing.
| Sample ID | Description | Lat. | Lon. | Basin area (km2)* | Median basin elevation (m) | Sample 10Be concentration (atoms g−1) | Calculated-denudation rate (mm yr−1) and analytical (internal) uncertainty |
|---|---|---|---|---|---|---|---|
| US-CHL-FR | Fraser main stem, upstream of Chilcotin confluence | 51.7851 | −122.3938 | 119,419 (61%) | 1,011 | 38,400 ± 900 | 0.22 ± 0.01 |
| DS-CHL-FR | Fraser main stem, downstream of Chilcotin confluence | 51.6542 | −122.3321 | 154,823 (61%) | 1,087 | 37,900 ± 1,300 | 0.24 ± 0.01 |
| DS-FRCH-FR | Fraser main stem, downstream of French Bar Canyon | 51.2056 | −122.1542 | 144,288 (60%) | 1,068 | 39,200 ± 900 | 0.22 ± 0.01 |
| US-THP-FR | Fraser main stem, upstream of Thompson confluence | 50.4145 | −121.6838 | 153,533 (61%) | 1,084 | 39,000 ± 900 | 0.23 ± 0.01 |
| LYT-FR | Fraser main stem, Lytton bar | 50.2503 | −121.5955 | 154,822 (61%) | 1,087 | 38,900 ± 900 | 0.23 ± 0.01 |
| LYT-FR-REPβ | Fraser main stem, Lytton bar | 50.2426 | −121.5877 | 154,822 (61%) | 1,087 | 38,100 ± 1,200 | 0.24 ± 0.01 |
| DS-THP-FR | Fraser main stem, downstream of Thompson confluence | 50.2135 | −121.5791 | 210,204 (64%) | 1,121 | 37,200 ± 1,100 | 0.24 ± 0.01 |
| SIS-FR | Fraser main stem, Siska canyon | 50.1441 | −121.5749 | 210,539 (64%) | 1,122 | 38,200 ± 1,000 | 0.24 ± 0.01 |
| BSTN-FR | Fraser main stem, Boston Bar | 49.8428 | −121.4417 | 212,670 (64%) | 1,124 | 39,100 ± 1,300 | 0.23 ± 0.01 |
| SPZ-FR | Fraser main stem, Spuzzum | 49.6723 | −121.4150 | 213,566 (64%) | 1,124 | 30,000 ± 900 | 0.30 ± 0.01 |
| SPZ-FR-REPβ | Fraser main stem, Spuzzum | 49.6730 | −121.4156 | 213,566 (64%) | 1,124 | 37,300 ± 1,000 | 0.24 ± 0.01 |
| DS-LFR-FR | Fraser main stem, downstream of Lady Franklin Rock | 49.5603 | −121.4286 | 214,008 (64%) | 1,124 | 33,300 ± 900 | 0.27 ± 0.01 |
| CHL | Fraser tributary, Chilcotin River | 51.7400 | −122.4074 | 19,358 (60%) | 1,283 | 98,200 ± 2,100 | 0.10 ± 0.01 |
| THP | Fraser tributary, Thompson River | 50.3490 | −121.3904 | 55,130 (71%) | 1,226 | 48,300 ± 2,300 | 0.19 ± 0.01 |
| NTHP | Thompson tributary, upstream of Lake Savona | 50.8181 | −120.3086 | 20,255 (74%) | 1,308 | 19,500 ± 600 | 0.51 ± 0.02 |
| STHP | Thompson tributary, upstream of Lake Savona | 50.6769 | −120.3150 | 17,115 (80%) | 1,165 | 27,200 ± 700 | 0.32 ± 0.01 |
| FRCH-T | Fraser tributary, French Bar Creek | 51.2322 | −122.1739 | 287 (92%) | 1733 | 109,900 ± 3,400 | 0.11 ± 0.01 |
| LHG-T | Fraser tributary, Little Hell's Gate | 49.7642 | −121.4433 | 36 (100%) | 1,241 | 28,100 ± 700 | 0.30 ± 0.01 |
| LHG-T-REPβ | Fraser tributary, Little Hell's Gate | 49.7642 | −121.4422 | 36 (100%) | 1,241 | 36,500 ± 1,000 | 0.23 ± 0.01 |
| E3-T | Fraser tributary, Black Canyon | 49.7582 | −121.4357 | 18 (100%) | 1,169 | 24,200 ± 800 | 0.33 ± 0.01 |
| E5-T | Fraser tributary, Yale | 49.5810 | −121.3968 | 66 (17%) | 1,079 | 24,000 ± 800 | 0.36 ± 0.01 |
| E1-T | Fraser tributary, Chapman | 49.7297 | −121.4372 | 5 (100%) | 1,240 | 15,300 ± 500 | 0.55 ± 0.02 |
- * The fraction that quartz-bearing lithologies represent in the total basin area is shown in parentheses.
- β Repeated sample – collected prior to the main 2022 sampling season (in 2021).
Basin-averaged denudation rates were generated using the rivers and python package (Stübner, Balco, & Schmeisser, 2024) and a 30 m Canadian Digital Elevation Model (CDEM, Natural Resources Canada, 2015) model of the Fraser River basin, which estimates CRN production rates while accounting for the hypsometry of the basin. The Digital Elevation Model was resampled to a spatial resolution of 90 m for computational efficiency, given the size of the basin (>200,000 km2). Riversand uses the online calculator of Balco et al. (2008) to calculate denudation rates. Topographic shielding was not included in these calculations because vertical attenuation lengths vary as a function of hillslope angle and skyline shielding. Vertical attenuation lengths, therefore, offset variable production rates that would be accounted for by topographic shielding (DiBiase, 2018). Given that only approximately 1% of the Fraser River basin is currently glaciated (Armstrong et al., 2005), we ignore glacier shielding in our calculations. We accounted for the non-uniform distribution of quartz-bearing lithologies within the basin by using a raster of quartz coverage that is an input in the riversand package to exclude quartz-free areas of the basin from denudation rate calculations. We produced this by assigning areas of ”quartz-free” lithology coverage to a value of zero, while “quartz-bearing” lithologies were assigned a value of 1. Basin-averaged denudation rates were then calculated using the riversand python package, assuming a rock density of 2.65 g cm−3. We report denudation rates based on the LSDn scaling model (Lifton et al., 2014). For LSDn scaling, the atmospheric model and primary 10Be production rate calibration dataset (Borchers et al., 2016) used by riversand result in a sea level and high latitude 10Be production rate of 4.16 atoms g−1 yr−1 (Martin et al., 2017). Integration timescales were calculated using the time needed to erode one mean attenuation path length (~60 cm; Lal, 1991) for each reported denudation rate.
3.2 Morphometry and climate
3.2.1 Basin and channel morphometry
3.2.2 Climate data
Mean Annual Precipitation (MAP) for the period 1991–2020 was calculated for the drainage area upstream from each CRN sample using ClimateBC v7.41, which is an 800 × 800 m gridded product based on Parameter-elevation Regressions on Independent Slopes Model (PRISM) monthly climate normal data (Wang et al., 2016). MAP was calculated as the average value of pixels clipped to each contributing area.
3.3 Paraglacial terrace mapping
Terrace areas were extracted using the method of Clubb et al. (2017) on a 3 m resolution LiDAR DEM (~0.15 m vertical accuracy) of the Fraser River between Soda Creek and Lady Franklin Rock provided by the Hakai Institute (Steelquist et al., in press). The DEM was divided into 11 sections of 25–60 km length to improve computational efficiency. The method automatically delineates terrace treads using thresholds of local surface gradient and elevation relative to the nearest river channel, allowing automatic identification of terrace treads over much larger spatial areas than could be efficiently mapped manually (Clubb et al., 2017). A swath width of 4,000 m and a maximum slope threshold of 10 degrees were used in the terrace extraction tool. A maximum relief threshold of 250–375 m was used based on visual inspection of each section of DEM and the corresponding relief of the highest detectable terrace. A minimum terrace height of 2 m was applied as this was equivalent to the maximum height of exposed gravel bars above the low-flow water line in the modern channel observed in the field. A surface fitting window radius of 20 m was used. Automatically extracted terraces were then cleaned using several methods. Sections of non-terrace features (e.g., houses, roads) or terraces with areas <0.001 km2 were manually removed. Each terrace feature was compared against aerial imagery to determine whether they were correctly classified. Visual checks were also made in areas where existing terrace maps were available (e.g., Lian & Hicock, 2001; Ryder & Church, 1986) and incorrectly identified terraces were manually removed.
To assign terrace relief values for individual terraces, a raster of terrace relief (terrace tread elevation relative to the local water surface elevation) was produced. To do this, a locally interpolated water surface was generated to remove noise in the 3 m LiDAR data. First, the water surface elevation was extracted from points every 100 m down the channel centerline on the 3 m LiDAR data. Point elevations were then interpolated using the Inverse Distance Weighting Interpolation tool in QGIS to 30 m pixel sizes (10 times the size of the LiDAR input data resolution) to minimise artificial noise between adjacent pixels. Terrace relief was then calculated by subtracting the elevation of the local interpolated water surface from the terrace elevation. Because identified terraces were characterised by high surface gradients and reworking (e.g., landslides, gullies), the elevation of an individual terrace was based on the 10th percentile of relief values across each terrace surface. A characteristic maximum terrace relief was then identified between Soda Creek and Lady Franklin Rock to calculate the volume of eroded terrace material. The initial terrace tread elevation (i.e., before the channel incised into the terrace fill) was calculated by adding the maximum postglacial terrace relief value to the water surface elevation using the smoothed LiDAR dataset. The interpolated LiDAR water surface DEM was then subtracted from the initial terrace surface to calculate the volume of eroded terrace sediment.
3.4 Mixing model
We developed a mixing model that predicts the amalgamated sediment flux and average 10Be concentration in the main stem of the Fraser River at 500 m downstream increments. The effect of paraglacial sediment on main stem 10Be concentrations was determined by estimating the mass of sediment supplied from terraces and its average 10Be concentration, and calculating the degree to which the paraglacial sediment alters sediment in the main stem channel, which is characterised by a known sediment flux and average 10Be concentration. CRN samples US-CHIL-FR and DS-LFR-FR are from the upstream and downstream limits of the study reach, respectively (Figure 1), and provide reference 10Be concentrations. Decadal averaged sediment fluxes from gauging stations at Marguerite and Hope allow us to constrain the total amount of sediment entering the study reach (Church, Kellerhals, & Day, 1989). The total increase in suspended sediment load between Marguerite and Hope from gauged records is ~5,847,300 t yr−1 (Church, Kellerhals, & Day, 1989). The change in drainage area between the two stations is ~98,000 km2, which corresponds to ~60 t of sediment added to the river per km2 of drainage area. There are two major tributaries that deliver sediment to the study reach, the Chilcotin River and Thompson River drainages (16% and 36% of the total drainage area, respectively) which may impact Fraser River 10Be concentrations and sediment loads, but the presence of lakes in the Thompson River drainage limits the downstream delivery of sediment. As shown below, samples collected from the main stem Fraser River upstream and downstream of both the Chilcotin and Thompson River confluences (US-CHL-FR/DS-CHL-FR and US-THP-FR/DS-THP-FR) suggest negligible impact of these tributaries on 10Be concentrations in the Fraser River main stem, so we do not consider the influx of sediment from these sub-basins explicitly in the model, as the volumetric load is unconstrained.
4 RESULTS
4.1 Basin-averaged denudation rates, morphometry and climate patterns
Basin-averaged denudation rates vary from 0.22 to 0.30 mm yr−1 for samples collected in the Fraser River main stem (Figure 3a), with integration timescales of ~2,000–2,700 years. Denudation rates derived from samples in the Thompson (THP) and Chilcotin River (CHL) tributary basins immediately upstream of their confluences with the Fraser River are lower at 0.19 and 0.10 mm yr−1, respectively (Table 1). The sample from French Bar Creek (FRCH-T), a relatively small tributary with a stream that is deeply incised into paraglacial terrace fill, also yields a comparatively low denudation rate of 0.11 mm yr−1. Tributaries further south (at Little Hell's Gate, Black Canyon and Yale; LHG-T, E3-T, E5-T) that incise into bedrock have comparatively higher denudation rates of 0.23–0.36 mm yr−1, relative to the Fraser River main stem. The range of integration timescales of tributary inputs ranges from ~1,200–6,000 years. Sample E1-T yielded the highest calculated denudation rate of 0.55 mm yr−1 but was removed from further analysis. The E1-T sample was collected downstream of a large debris flow that created Tikwalus Rapid during an atmospheric river in 2021, which predates the 2022 sample collection. The large volume of low 10Be concentration sediment released by the debris flow likely overwhelmed the non-event sediment load from the small basin area (~5 km2), and likely yielded an unrealistically high denudation rate. Samples NTHP and STHP are sub-basins of the Thompson River upstream from Lake Savona, which is a major sediment trap in the Thompson River basin and has similarly higher denudation rates of 0.51 and 0.32 mm yr−1, respectively, relative to the Thompson River sample (THP) at the Thompson-Fraser confluence. Denudation rates determined from the repeat samples at Lytton (LYT-FR and LYT-FR-REP) were comparable at 0.23 ± 0.01 and 0.24 ± 0.01 mm yr−1, respectively, but the repeat samples differed at Spuzzum (SPZ-FR and SPZ-FR-REP) by 24% (0.30 ± 0.01 and 0.24 ± 0.01 mm yr−1, respectively) and in the Little Hell's Gate tributary (LHG-T and LHG-T-REP) by 30% (0.30 ± 0.02 and 0.23 ± 0.01 mm yr−1, respectively). Observations by Baird (2024) indicated that areas downstream of Lytton were more heavily impacted by the 2021 atmospheric river, which may account for the lowering of 10Be concentrations between repeat sampling years if low-concentration debris flow sediment was not fully mixed or overwhelmed sampling locations.
Relations between basin average denudation rates and drainage area, median basin slope, and MAP are generally weak (Figure 4a). Using samples along the Fraser River main stem only, there is a modest downstream increase in basin-averaged MAP (Table 2) as increasing proportions of basin area draining the Coast Mountains are incorporated (Figure 3b). Regressions between drainage area, MAP, basin slope, and denudation rates for samples along the Fraser River main stem show substantial scatter, and the proportion of the variability in denudation rate explained by drainage area, MAP, basin slope, are not high (r2 = 0.31 to 0.35). Including the tributaries in the regressions yields a statistically significant relation between denudation rates and MAP, but not with drainage area, basin slope, or at the 90% confidence interval (p-value ≤0.1). Smaller tributaries in the southern part of the study area are generally characterised by higher rainfall (and denudation rates), and higher values of (Figure 3c). Along the Fraser River main stem, is 5 m0.9 between Soda Creek and the Thompson River confluence and increases to 6 m0.9 downstream of the confluence (Table 2).
| Sample ID | Description | Corrected drainage area (km2) | Basin median slope (degrees) | Basin median (m0.9) | Basin mean annual precipitation (mm) |
|---|---|---|---|---|---|
| US-CHL-FR | Fraser main stem | 71,704 | 6.15 | 15 | 746 |
| DS-CHL-FR | Fraser main stem | 83,208 | 5.76 | 22 | 722 |
| DS-FRCH-FR | Fraser main stem | 85,338 | 5.76 | 26 | 717 |
| US-THP-FR | Fraser main stem | 91,888 | 6.27 | 36 | 722 |
| LYT-FR | Fraser main stem | 93,044 | 6.32 | 38 | 723 |
| DS-THP-FR | Fraser main stem | 132,246 | 7.47 | 61 | 758 |
| SIS--FR | Fraser main stem | 132,285 | 7.47 | 60 | 758 |
| BSTN-FR | Fraser main stem | 134,053 | 7.55 | 62 | 763 |
| SPZ-FR | Fraser main stem | 134,720 | 7.63 | 63 | 766 |
| DS-LFR-FR | Fraser main stem | 135,026 | 7.63 | 63 | 768 |
| CHL | Tributary | 11,452 | 4.14 | 61 | 574 |
| FRCH-T | Tributary | 266 | 13.39 | 250 | 817 |
| THP | Tributary | 38,862 | 11.07 | 154 | 866 |
| NTHP | Tributary | 14,792 | 12.58 | 144 | 1,075 |
| STHP | Tributary | 13,709 | 14.11 | 163 | 1,006 |
| LHG-T | Tributary | 37 | 29.66 | 797 | 1863 |
| E3-T | Tributary | 11 | 28.78 | 1,105 | 1743 |
| E5-T | Tributary | 19 | 27.52 | 339 | 2033 |
4.2 Spatial patterns in 10Be concentrations
Measured 10Be concentrations along the Fraser River main stem vary little through the study reach, with concentrations ranging from 30,000 ± 900 atoms g−1 at Spuzzum (SPZ-FR) to 39,200 ± 900 atoms g−1 at French Bar Canyon (DS-FRCH-FR). 10Be concentrations were higher on the larger Chilcotin River and Thompson River tributaries (98,200 ± 2,100 and 48,300 ± 2,300 atoms g−1, respectively) and the delivery of this sediment appears to have little influence on concentrations of the main stem Fraser River samples taken upstream and downstream of the tributary confluences (Figure 5) where the main stem 10Be concentration remains relatively consistent.
Lateral sources of sediment in the upstream reaches (e.g., terraces and tributaries) are characterised by comparatively higher 10Be concentrations than the Fraser River main stem. 10Be concentrations from the two terrace depth profiles at High Bar and Iron Canyon (Figure 6) were relatively homogenous as a function of depth, suggesting significant vertical mixing and erosion of the terrace surface since deposition. The 10Be concentrations in the depth profile varied between 46,500 ± 1,800 and 71,100 ± 2,200 atoms g−1 across both sites (Table 3), which is slightly lower than the concentration recorded at French Bar Creek (FRCH-T) where the channel is deeply incised into terrace fill (109,900 ± 3,400 atoms g−1).
| Sample name | Surface elevation (m) | Shielding factor | Deptha (cm) | Sample 10Be concentration (atoms g−1)* |
|---|---|---|---|---|
| Iron Canyon | ||||
| EC221006-IC-41 | 385.93 | 0.997 | 41 | 49,400 ± 1,200 |
| EC221006-IC-61 | 385.93 | 0.997 | 61 | 53,200 ± 1,300 |
| EC221006-IC-81 | 385.93 | 0.997 | 81 | 50,000 ± 1,100 |
| EC221006-IC-92 | 385.93 | 0.997 | 92 | 46,500 ± 1,800 |
| EC221006-IC-150 | 385.93 | 0.997 | 150 | 49,400 ± 1,200 |
| EC221006-IC-198 | 385.93 | 0.997 | 198 | 52,500 ± 1,300 |
| EC221006-IC-259 | 385.93 | 0.997 | 259 | 62,600 ± 1,400 |
| High Bar | ||||
| EC220930-HB-36 | 499.85 | 0.990 | 36 | 57,900 ± 1,900 |
| EC220930-HB-54 | 499.85 | 0.990 | 54 | 58,400 ± 1,700 |
| EC220930-HB-77 | 499.85 | 0.990 | 77 | 58,700 ± 1,800 |
| EC220930-HB-104 | 499.85 | 0.990 | 104 | 71,100 ± 2,200 |
| EC220930-HB-154 | 499.85 | 0.990 | 154 | 44,500 ± 1,700 |
| EC220930-HB-221 | 499.85 | 0.990 | 221 | 59,800 ± 1,500 |
- a Reported depth is the centre of ~2 cm thick sampling interval.
- * AMS analytical (internal) uncertainties reported.
4.3 Paraglacial terrace mixing
4.3.1 Terrace distribution
We identified more than 140 km2 of terrace surface between Soda Creek and Lady Franklin Rock (Figure 5), 100 km2 of which were terraces with areas >0.001 km2. However, based on visual inspection of satellite imagery, there were many terrace surfaces that were not correctly classified (false negatives) using the objective method of Clubb et al. (2017) meaning our terrace coverage may be underestimated. The underestimation likely arises from the relatively steep and irregular surfaces of some of the terraces, and by post-depositional gullying, alluvial fan deposition and mass movement (e.g., small landslides, earthflows and slumping), that often heavily dissect the terrace treads. Downstream from Lillooet, terrace surfaces become increasingly vegetated and appear to supply less sediment to the modern channel. Downstream of Lytton, terrace surfaces are heavily vegetated and given their relatively low relief in areas such as Boston Bar, it is more likely that these terraces are former late-Holocene floodplain surfaces rather than older Holocene paraglacial terrace fill.
Most terrace treads are 90–230 m above the modern water surface (Figure 7). The greatest proportions of terrace areas are located 120–140 m above the water surface, with secondary peaks at 90–100 m, 150–160 m and 180–190 m. These elevations are consistent with mapped terraces in the Big Bar region, where several postglacial terrace surfaces were identified at elevations between ~160 and 200 m above present water levels (Gingerich, 2021). Terrace sediment density at High Bar and Iron Canyon was calculated as 1,386 and 1,432 kg m−3, respectively. The total volume of eroded terrace fill (assuming a maximum terrace elevation of 250 m) was 88.5 km3, which assuming a terrace fill density of 1,400 kg m−3, corresponds to ~1.24 × 1011 t of sediment (or ~7,600 years of the modern annual sediment load at Hope).
4.3.2 Mixing model terrace contributions
The mixing model reproduced the observed change in 10Be concentration between the most upstream and downstream samples on the Fraser River only when terrace contributions are less than ~15% of the sediment mass carried by the main stem (Figure 8). To reproduce the observed change in main stem 10Be concentration with 0% terrace contribution, a background sediment flux with an average 10Be concentration of ~24,000 atoms g−1 would be required (Figure 8b). While low in comparison to the observed Fraser River main stem 10Be concentrations, ~24,000 atoms g−1 is comparable to concentrations observed in the tributary basins that join the Fraser River at Black Canyon and Yale and in the Thompson River basin upstream of Savona Lake (STHP and NTHP). For a 25% terrace contribution, a background sediment flux with an average 10Be concentration of 8,500–17,000 atoms g−1 would be required to reproduce main stem concentrations. This background flux concentration is comparable to the lowest 10Be concentrations observed in the tributaries, but would also require the highest measured terrace 10Be concentrations to be used in the model to reproduce main stem concentrations (Figure 8b). With increasing contributions of terrace material, the background sediment flux 10Be concentration required to reproduce the observed change in 10Be concentration along the Fraser River main stem becomes unrealistically low, and unachievable with terrace contributions of more than ~30% (Figure 8b).
5 DISCUSSION
5.1 What controls the downstream evolution of 10Be concentrations?
Despite a near doubling of drainage area and a clear physiographic transition within the study region, basin-averaged denudation rates calculated from 10Be concentrations in the main stem of the Fraser River remain relatively uniform, suggesting that high 10Be concentration inputs from tributaries and terrace material are buffered. Despite large increases in drainage area at confluences with the Chilcotin River and Thompson River basins, which are also characterised by considerably higher 10Be concentrations, samples in the Fraser River main stem appear unaffected, suggesting that tributary sediment inputs are small in comparison to the sediment load in the main stem. We note that 10Be concentrations for repeat samples vary by up to 24%, which is surprising given the size of the basin (e.g., Niemi et al., 2005), although notable temporal variability has been reported for repeat samples in >10,000 km2 basins in the Himalaya (e.g., Dingle et al., 2018; Lupker et al., 2012) and there may be incomplete mixing of debris flow sediment generated by the 2021 atmospheric river event (Baird, 2024). The samples taken at Little Hells Gate are tributary samples with a small upstream basin (37 km2), so hillslope inputs may not be fully mixed (e.g., Binnie et al., 2006) or may be overwhelmed by low-concentration material from the debris flows in 2021. The Spuzzum samples were taken on the Fraser River main stem ~1 km upstream of where Spuzzum Creek joins the Fraser River main stem, and there are no obvious lateral sediment sources for several kilometres upstream, so it is unclear what has driven temporal variability at this site. There is a slight decrease in 10Be concentration downstream from Boston Bar (5–23% reduction across the three downstream samples) that may reflect inputs from tributaries with higher denudation rates (and lower 10Be concentrations), which closely mirrors patterns in ksn and MAP which are both elevated in the Coast Mountains (Figure 3b,c). There are a few paraglacial terraces in this downstream reach of the Fraser River, which may also prevent these lower concentration inputs from being offset by high concentration terrace sediment.
Paraglacial terrace fill characterises much of the landscape upstream of Soda Creek, where our study reach begins, so it is difficult to discern whether the 10Be concentration we observe in our most upstream sampling location (US-CHL-FR) reflects terrace storage and mixing, or modern landscape denudation processes. The ~0.22 mm yr−1 denudation rate averages over ~2,700 years. A ~ 2,700 year timescale corresponds to the timeframe within which the modern Fraser River started incising into bedrock (e.g., Gingerich, 2021). As such, we can assume that the Fraser River has been incising into bedrock rather than paraglacial sediment during our integration timescales and reported denudation rates are reflective of the modern channel configuration.
The denudation rates we calculate are also consistent with longer-term rates of exhumation, and incision rates averaged over longer timescales in the region. For example, minimum incision rates based on dated Neogene volcanic lavas in the Dog Creek region yield incision estimates of 0.20 mm yr−1, averaged over the last 0.76 Ma following a major drainage reversal of the Fraser River at this time (Andrews et al., 2012). Longer-term exhumation rates from thermochronology were of the order of 0.2–0.4 mm yr−1 from 10 to 4 Ma, and then accelerated by 70% during the past 4 Ma due to glacial erosion in the Coast Mountains (Farley, Rusmore, & Bogue, 2001). Exhumation rate estimates of 0.4–0.8 mm yr−1 based on apatite fission track data from the Coast Mountains have also been interpreted as a possible range of preglacial denudation rates across the region (Ehlers et al., 2006). Our CRN-derived denudation rates are within the range of these pre-Pleistocene estimates, despite a large contrast in integration timescales with other estimates. One explanation of this could be a bias introduced by the small fraction of high-CRN concentration paraglacial sediment still present in the modern river sediment load, resulting in a CRN-derived denudation rate estimate that is lower than would otherwise be expected when integrated over shorter millennial timescales.
5.2 Representation of paraglacial terrace sediment in the mixing model
Representation of terrace material in the mixing model could have biased our results. The two depth profiles were collected within a few meters of the modern terrace surface, and these concentrations were used to characterise the modelled terrace sediment inputs. Terrace sediment input is likely driven by lateral erosion and undercutting closer to the channel, which could input lower 10Be concentration sediment that has been more effectively shielded by the overlying terrace fill. However, the measured 10Be concentrations from the Chilcotin River (CHL) and French Bar Creek (FRCH-T) samples are notably higher than the Fraser River main stem samples, and the sediment load in these basins that drain interior plateaus are dominated by terrace fill (e.g., Church & Slaymaker, 1989). Field observations also indicate that material from the terrace tread is readily supplied to the main channel via gullying and shallow landslides (Figure 2). Dissection of the terrace riser is also likely to release high-concentration sediment into the channel. Sediment lower down in the terrace fills is likely to contain inherited 10Be from previous exposure and complex denudation histories. As such, it seems unlikely that the terrace fill 10Be concentrations are underestimated in the model.
We acknowledge that a large proportion of the modern sediment load in the Fraser River is represented by grain sizes finer than those collected for detrital CRN sampling. The suspended sand load of the Fraser River at Agassiz represents only one-third of the total suspended load (McLean, Church, & Tassone, 1999), where the suspended load constitutes more than 95% of the total sediment load. Silt-sized particles make up a significant portion of the sediment load. Similarly, the grain size of sediment capping Quaternary terrace fill is often reported as being in the silt grain size range (e.g., Clague et al., 2021; Lian & Hicock, 2001). Our two depth profiles were extracted from the upper terrace surface within a well-sorted massive fine-grained sand body with minor granules, yet sufficient material in the 250–850 μm range was extracted for CRN analysis. The widespread presence of sand in the sampled terrace fills suggests that these grain sizes should have experienced the same 10Be accumulation history as silt-sized particles, and that terrace fill contributions should be well-represented within the sand fraction of the modern Fraser River sediment load.
5.3 How strong is the paraglacial sediment signal?
We estimate that at least ~1.24 × 1011 tonnes of sediment have been excavated from the terrace fill between Soda Creek and Lady Franklin Rock. The sediment volume is approximately 7,500 times the modern annual sediment load of the Fraser River, and does not account for intermediate phases of incision and aggradation. We assume that the majority of terrace downcutting would have been focused during cooler and wetter mid-Holocene periods ~7–4 ka, with the Fraser River incising into bedrock during the last 4 kyr (Gingerich, 2021). Downstream of Soda Creek, where the Fraser River is flowing over and incising into bedrock, our analysis suggests that sediment input from paraglacial terrace fill is now limited. Terrace fill is largely perched above the main channel, despite widespread prevalence of paraglacial terraces, limiting lateral inputs of terrace material to either smaller tributaries, which have yet to incise into bedrock or through discrete episodes of mass wasting which deliver paraglacial sediment directly into the channel (e.g., Clague & Evans, 2003).
In terms of spatial patterns of specific sediment yield, Church and Slaymaker (1989) argued that with increasing basin area above ~104 km2, specific sediment yield of the Fraser River starts to decline in response to sequestration of reworked paraglacial sediment on floodplain surfaces. However, the confined canyon morphology downstream of Soda Creek likely limits long-term storage of remobilised sediment within this part of the landscape, which is assumed to be postglacial in age (Rennie, Church, & Venditti, 2018). Downstream changes in specific sediment yield along this reach should not necessarily represent remobilised paraglacial sediment going back into temporary storage (Church & Slaymaker, 1989), as it seems likely that the Fraser River would directly transport sediment to its lowland delta. Any downstream reduction in specific sediment yield below Soda Creek and Hope is more likely to reflect a change in lateral connectivity between terrace fill and the main stem channel rather than significant floodplain sedimentation. Finally, the increase in basin-averaged denudation rates downstream of Boston Bar and overall steepening of the landscape may indicate rejuvenation and dominance of primary denudation of bedrock in this downstream part of the basin, although this requires higher-resolution CRN sampling to confirm.
Our findings suggest that sediment supply from paraglacial sediment deposits is sensitive to the configuration of the modern fluvial network. Signals of primary bedrock denudation can be detectable in postglacial landscapes where channels are sufficiently incised (entrenched) into bedrock underlying paraglacial terrace sediments, which limits connectivity between terraces and river channels. These signals can emerge over much shorter timescales than those associated with the complete evacuation of paraglacial sediment deposits from postglacial landscapes. The evacuation of paraglacial terrace material will likely depend on the frequency of mass wasting along valley margins, such as the Chilcotin River landslide in July 2024 (Government of British Columbia, 2024). The effects of landslides delivering paraglacial sediment into the channel are unlikely to be detected using CRNs downstream of Soda Creek because of the relatively small magnitude of such events relative to the total sediment load of the Fraser River. We also further demonstrate the wider utility of CRNs such as 10Be in understanding sources of sediment in postglacial landscapes (e.g., Delunel et al., 2013; Norton et al., 2010).
6 CONCLUSION
The legacy of glaciation can persist in postglacial landscapes for thousands of years following deglaciation. In the Fraser River, the glacial legacy is thought to be expressed through distinct downstream patterns of specific sediment yield that reflect erosion and downstream transport of Pleistocene-age paraglacial fill terrace sediment. Combined cosmogenic radionuclide concentration measurements with mapped paraglacial terrace distributions within a sediment mixing model framework suggest paraglacial terrace sediment constitutes <15% of the modern Fraser River sediment load downstream of Soda Creek. The low contribution of paraglacial sediment is due to diminished connectivity between paraglacial terrace fill and the main stem Fraser River channel, which is now incised into bedrock. Changes in specific sediment yield downstream from Soda Creek are likely a consequence of limited valley storage for remobilised Quaternary sediments. Finally, we suggest that denudation of the bedrock landscape, rather than paraglacial terrace sediment, dominates late-Holocene landscape denudation in the lower Fraser River Basin.
AUTHOR CONTRIBUTIONS
E.H.D., I.L. & J.V. conceptualised the study. All authors were involved in collecting and/or preparing cosmogenic radionuclide samples. E.S. and J.C. performed the topographic analyses, and E.H.D. developed the mixing model. E.H.D. wrote the initial manuscript with contributions from all authors.
ACKNOWLEDGEMENTS
The authors are grateful to Derya Whaley-Kalaora and Matteo Saletti for field assistance, as well as Sage Moore for laboratory assistance, Alan Hidy for conducting AMS measurements and Edward Anderson for assistance with climate and topographic data analysis. We thank Eyal Marder for assistance in denudation rate calculations and David Small for helpful discussions. We thank two anonymous reviewers for thoughtful comments that improved the initial manuscript. We also thank the Kanaka Bar Indian Band and the Lytton First Nation for giving us permission to work on their traditional territories.
CONFLICT OF INTEREST STATEMENT
The authors have no conflicts of interest.
Open Research
DATA AVAILABILITY STATEMENT
All data used in this manuscript are available within the manuscript and Supporting Information, and at a Federated Research Data Repository (DOI:10.20383/103.01166). LiDAR data are held by the Hakai Institute, and were collected by the Airborne Coastal Observatory with partners at the University of Northern British Columbia and Simon Fraser University. Data are discoverable through Hakai's metadata catalogue.
