A RECORD OF COUPLED HILLSLOPE AND CHANNEL RESPONSE TO PLEISTOCENE PERIGLACIAL EROSION IN A SANDSTONE HEADWATER VALLEY, CENTRAL PENNSYLVANIA

Abstract

Outside of the Last Glacial Maximum ice extent, landscapes in the central Valley and Ridge physiographic province of Appalachia preserve soils and thick colluvial deposits indicating extensive periglacial landscape modification. The preservation of periglacial landforms in the present interglacial suggests active hillslope sediment transport in cold climates followed by limited modification in the Holocene. However, the timing and extent of these processes are poorly constrained, and it is unclear whether, and how much, this signature is due to LGM or older periglaciations. Here, we pair geomorphic mapping with in situ cosmogenic ¹⁰Be and ²⁶Al measurements of surface material and buried clasts to estimate the residence time and depositional history of colluvium within Garner Run, a 1 km² sandstone headwater valley in central Appalachia containing relict Pleistocene periglacial features including solifluction lobes, boulder fields, and thick colluvial footslope deposits. ¹⁰Be concentrations of stream sediment and hillslope regolith indicate slow erosion rates (6.3 m ± 0.5 m m.y.^-1) over the past 38-140 kyr. From dating of buried valley-bottom deposits recovered from a 9 m drill core, we interpret two depositional pulses since ~290 ka, a record which spans at least three glacial terminations and implies limited removal of valley bottom deposits during interglacials. This age is consistent with independent calculations determined from debris volume estimates, total hillslope contributing area, and catchment erosion rate integrated over multiple climate cycles. Thus, we infer that erosion rates measured in upland basins in Central Appalachia reflect the integration both temperate and periglacial processes, and that in cold-warm transitions, erosion rates reflect only moderate departures from otherwise slow background rates. Furthermore, due to slow erosion rates, we show that sedimentary records in sandstone headwater valleys present opportunities for direct examination of climate-modulated hillslope processes.

 

      

TABLE OF CONTENTS

LIST OF FIGURES………………………………………………………………………………………………. v

LIST OF TABLES………………………………………………………………………………………………. vi

ACKNOWLEDGEMENTS…………………………………………………………………………………. vii

INTRODUCTION………………………………………………………………………………………………… 1

STUDY AREA…………………………………………………………………………………………………….. 2

METHODS………………………………………………………………………………………………………….. 7

Topographic Analysis……………………………………………………………………………………….. 7

Field Mapping of Regolith Texture……………………………………………………………………… 7

Grain Size Analysis of Surface Clasts on Hillslope and in Channels……………………….. 8

Cosmogenic Nuclide Analysis………………………………………………………………………….. 11

Determining Colluvial Fill History……………………………………………………………………. 13

RESULTS………………………………………………………………………………………………………….. 14

Topographic Analysis……………………………………………………………………………………… 14

Field Mapping of Regolith Texture……………………………………………………………………. 15

Grain Size Analysis of Surface Clasts and Hillslopes and in Channels…………………… 16

Cosmogenic Nuclide Analysis………………………………………………………………………….. 21

Constraining Colluvial Fill History……………………………………………………………………. 21

DISCUSSION……………………………………………………………………………………………………. 26

Controls on Spatial Variability in Regolith Texture and Morphology……………………. 26

Colluvial Fill History and Implications for Hillslope-Channel Coupling………………… 30

Implications for the Evolution of the Critical Zone……………………………………………… 33

CONCLUSIONS………………………………………………………………………………………………… 34

References…………………………………………………………………………………………………………. 36

Appendix A: Workflow of GPS-assisted soil mapping at Garner Run……………………….. 41

Appendix B. Soil pit descriptions…………………………………………………………………………. 45

Appendix C. Modeling Scenarios and Constants…………………………………………………….. 52

Appendix Tables………………………………………………………………………………………………… 55

Appendix Figures……………………………………………………………………………………………….. 58

INTRODUCTION

Quaternary climate fluctuations profoundly influenced the style and pace of erosion in glaciated landscapes (Hallet et al., 1996; Koppes and Montgomery, 2009), but the extent to which erosion rates in nearby periglacial landscapes were affected by concurrent changes in temperature and hydrology remains unclear. While a switch to colder temperatures is thought to stimulate bedrock lowering through increased frost cracking (Hales and Roering, 2007; Anderson et al., 2013; Rempel et al., 2017), downslope transport of regolith by ice-driven creep, solifluction, and permafrost-thaw mass wasting is more efficient during transitional warming periods (Taber, 1929; Matsuoka, 2001; Lewkowicz and Harris, 2005). Additionally, headwater channel networks may be periodically overwhelmed by periglacial hillslope sediment flux (Pizzuto, 1995; Simpson and Schlunegger, 2003), insulating hillslopes from regional base level change and thus complicating the overall landscape response to changing climate.

In landscapes where erosion rates are slow compared to the frequency of climate shifts, weathering profiles within regolith integrate the effects of multiple glacial-interglacial cycles (Yoo et al., 2011; Anderson et al., 2013). In central Appalachia, the periglacial debris mantling hillslopes and valley floors is commonly assumed to reflect a single cold climate cycle experiencing reequilibration in a temperate climate (Braun, 1989). However, the persistence of these relict landscape features alludes to a more complex history of sediment transport and coupled hillslopechannel responses. Periglacial landscapes like those in central Appalachia were common throughout middle latitudes during glacial periods (Vandenberghe et al., 2014). Thus, understanding how past environmental conditions control the composition and structure of the modern surface and shallow subsurface (Earth’s “critical zone”), as well as the steadiness of surface processes through time and space has wide applicability (Anderson et al., 2007; Brantley and Lebedeva, 2011).

In this study, we seek to understand the timescale, mechanisms, and extent of periglacial landscape modification by studying sediment stored in a headwater valley in central Pennsylvania. We quantify spatial variation in periglacial colluvium through detailed regolith mapping supplemented by analysis of lidar-derived high-resolution topography. We use the size, shape, and distribution of boulder colluvium on hillslopes and sediment in channels to infer sediment transport processes. To quantify erosion rates and patterns of hillslope sediment transport, we measure insitu cosmogenic nuclide concentrations in regolith and stream sediment. We also measure in-situ cosmogenic ²⁶Al and ¹⁰Be in a valley-bottom core to explore the timescale over which material derived from periglacial hillslope erosion is accumulated in the valley bottom. We then discuss implications of timescales of colluvium storage for the long-term evolution of headwater valleys and channels.

 

STUDY AREA

We focus on Garner Run, a 1 km² catchment underlain by the Silurian Tuscarora Formation within the Susquehanna Shale Hills Critical Zone Observatory (SSHCZO) in central Pennsylvania (Fig. 1a) (Brantley et al., 2016). The 150 m thick Tuscarora Formation is an orthoquartzite sandstone with minor thin interbedded olive-gray shale, and is overlain in some locations by an upper Castanea Member, an iron-cemented sandstone approximately 30 m thick (Flueckinger, 1969). Throughout central Pennsylvania, Paleozoic structures exhibit first order control on topography; the erosion-resistant Tuscarora Formation forms long linear ridges along plunging folds, which typically support the highest topography in the Valley and Ridge province (Fig. 1a).

 

Fig. 1: (A) The topography of the Valley and Ridge physiographic province is marked by tight folds of resistant lithologies and valleys of erodible units like shale. On resistant lithologies, hillslope angles are controlled by underlying bedrock dip on dip slopes (blue polygons). Draped over these hillslopes are landscape features that indicate mass movement – solifluction lobes (red outlines) are thought to indicate periglacial sediment transport, subtle cuestas which we interpret to be shadow bedding (black outlines) indicate influence of bedrock dip under soil mantle. Garner Run contains a steepened river reach known as a knickpoint as it flows over a resistant anticline. Study site and the area shown in cross-section (Figure 1B) are located in the black box. Point count locations referenced in Fig. 3 are shown as white dots along the channel. (B) Perspective view of slope-shade maps of the Garner Run subcatchment along with the adjacent Shavers Creek, a tributary to the Susquehanna River. Underlying geologic cross-section shows Garner Run contained within Tuscarora syncline; the Leading Ridge ridgeline is a local anticline, while Tussey Mountain is a breached anticline. We show colluvium mantling the bedrock in the valley at Garner Run, shadow bedding on Leading Ridge (also outlined in black) and slopeshade lidar imagery showing details of solifluction lobes in the valley (also outlined in red).

 

 

The Garner Run watershed lies within a synclinal valley bounded by two linear ridgelines—Tussey Mountain to the northwest, and Leading Ridge to the southeast—and the dip of the Tuscarora Formation parallels, or is slightly steeper than, hillslope topography (Flueckinger 1969). Further downvalley from the study catchment, the overlying Rose Hill Shale and the Keefer Sandstone crop out in the valley axis (Fig. 1b).

Overprinting the lithologic and structural control on topography in central Pennsylvania are patterns in bedrock river steepness and contrasting erosion rates above and below oversteepened river reaches known as knickpoints (Whipple and Tucker, 1999). Such disequilibria in river networks have been used to argue for a regional increase in the rate of base level fall that has propagated throughout the Susquehanna River basin since the Miocene (Miller et al., 2013). Downstream of the study catchment, Garner Run steepens at an elevation concurrent with the expected location of such a signal, but is also coincident with a lithologic and structural contrast across a Tuscarora anticline, where intact sandstone bedrock is exposed in the channel bed (Brantley et al., 2016), obscuring the degree to which this knickpoint reflects lithologic and/or base level change. At the original Shale Hills site, downstream of a knickpoint at a similar elevation, erosion rates are 15 m m.y.^-1 (West et al., 2013); thus we expect erosion rates at Garner Run to be lower (Brantley et al., 2016).

Garner Run is located ~75 km south of the Last Glacial Maximum (LGM) extent at 25 ka (Corbett et al., 2017b) and till deposits indicate that glacial advances remained north of the area in both the Illinoian (~130 ka) and early Pleistocene (~800 ka) glaciations (Ramage et al., 1998; Ciolkosz et al., 2008). In northeastern North America, a cold-climate vegetation assemblage (spruce, fir and pines) dominated until the end of the Younger Dryas ~ 11 ka, at which point the modern suite of Holocene vegetation (temperate deciduous and warm mixed forests) was established and persisted (Shuman, 2002; Williams et al., 2004). Palynological evidence indicates that tundra flora, perhaps in soils underlain by discontinuous or isolated permafrost, persisted in the central Valley and Ridge until ~16 ka, followed by a transitional period toward alpine communities until they were replaced by the modern hardwood community by 10 ka (Kneller and Peteet, 1999), though cold-climate vegetation communities may have endured longer at higher elevations (Kovar, 1965; Watts, 1979). Peak Holocene summer temperatures did not occur until 7 ka, corresponding to the weakening influence of the receding continental ice sheet and the thermal effects of summer insolation anomalies (Shuman and Marsicek, 2016). The modern climate of Garner Run is temperate (mean annual temperature of 10° C) with mean annual precipitation of 1.0 m (Thomas et al., 2013). Vegetation is characterized by deciduous trees with a few pines and hemlocks on ridgelines, though the area has been deforested a number of times since European settlement in the 18^th century  (Robinson, 1959).

Like other landscapes in the region, the Garner Run catchment bears evidence of extensive periglacial landscape modification as a consequence of colder temperatures in the Pleistocene. Boulder fields and landforms indicative of mass wasting processes are scattered across hillslopes at Garner Run (Fig. 2a). In the valley axis of Garner Run a broad, low-sloping bench with subtle lobate terraces hints at significant accumulation of colluvial deposits from adjacent hillslopes (Brantley et al. 2016). Throughout central Appalachia, colluvial deposits indicative of periglacial erosion blanket hillslopes and valley bottoms with blocky debris in resistant sandstone lithologies (Clark et al., 1992), influencing soils and hillslope morphology (Clark and Ciolkosz, 1988; Braun, 1989) as well as fluvial incision (Pizzuto, 1995; Reusser et al., 2004). The preservation of relict landforms today suggests vigorous periglacial landscape modification (Braun, 1989) and inhibited modification and evacuation of hillslope debris by modern landscape processes. Additionally, higher catchment-averaged erosion rates in the Susquehanna River Basin versus the Potomac River Basin to the south may reflect more intense periglacial erosion adjacent to the LGM ice margin (Portenga et al., 2017). However, both the timing of periglacial modification and extent of the landscape inheritance of Pleistocene cold-period conditions remain poorly constrained.

 

METHODS

Topographic Analysis

We used 1-m resolution lidar-derived topography from both the 2010 leaf-off SSHCZO lidar survey (OpenTopography, 2010) and the 2006 PAMAP lidar of Pennsylvania (PAMAP Program, 2006) to generate slopeshade maps for identifying periglacial landforms throughout the Garner Run study catchment and nearby landscape (Fig. 1a). We mapped lobate structures that we interpret as relict solifluction lobe crests, hummocky terrain and landslides that we interpret as relict permafrost thaw slumps, and faint cuestas that we interpret to reflect shadow bedding of the underlying sandstone bedrock. Additionally, we generated topographic cross-sections at Garner Run to estimate the orientation and thickness of the underlying folded bedrock and interpret the thickness of colluvial valley fill, which is additionally constrained by a 9 m core in the valley axis and shallow geophysical surveys (DiBiase et al., 2016).

Field Mapping of Regolith Texture

To characterize spatial variations in regolith surface texture, we mapped boulder density and canopy cover in the field using lidar-derived maps on a smartphone and sub-meter resolution Bluetooth GPS for positioning. We defined five mapping units based on bedrock exposure, soil and boulder cover, and tree canopy cover: (1) soil (<10% surface boulders), (2) boulders/soil (1067% surface boulders, embedded in soil), (3) boulder fields with tree canopy (>67% surface boulders not embedded in soil) (4) open boulder fields with no tree canopy, and (5) in place bedrock (Fig. 2a). These surface textures lie on a continuum; we distinguished between boulders/soil and boulder fields based on whether boulders wobbled underfoot when disturbed, a proxy for the amount of interstitial soil. We mapped at a resolution of 5 m, which was chosen to balance mapping speed and detail and generally reflects the minimum scale over which regolith texture varies. We also dug several shallow (<1.5 m) soil pits on the different mapping units to examine clasts at depth, as well as their relation to any soil horizons.

Grain Size Analysis of Surface Clasts on Hillslope and in Channels

We conducted Wolman (1954) point counts of the intermediate axis of 40-100 surface clasts of at least cobble size (≥ 6.4 cm) along a 1 m grid at 20 hillslope sites to characterize surface patterns in grain size distribution. Hillslope sites included all mapped units except for fully soil mantled areas with no coarse surface clasts. Both open and canopied boulder fields typically contained no finer-grained material, and thus point counts reflect the full grain size distribution of regolith. Point counts of coarse sediment on otherwise soil-mantled surfaces (“boulders/soil” mapping classification) are not directly comparable, but provide constraints on the maximum boulder size.

At 8 locations along the channel, we conducted similar point counts, but incorporated all grains, defining material with diameter < 2 mm as “fine”. Four channel point counts were spaced approximately every 100 m in the headwaters, two point counts were conducted 4 km downstream from the subcatchment outlet, and two point counts were conducted 6 km downstream from the subcatchment outlet, where Garner Run turns southeast and steepens across a knickpoint (Brantley et al., 2016).

 

 

Fig. 2: (A) Field photos showing heterogeneity in surface cover. (0) Tuscarora sandstone cropping out at ridgeline, with primary horizontal fracturing along bedding planes. (1) Blockfields with large, debris are devoid of soil and prevent canopy growth. (2) Slopes with blocky debris but also patchy soils and trees. (3) Boulders with interstitial soil, and frequent shrubs and other vegetation. (4) Soil mantled cover with few or no boulders. (B) Soil mapping results from the Garner Run subcatchment at 5m resolution. Contours highlight low slope of valley floor and relative steepness of adjacent hillslopes. Generally, more soil is located in the valley bottom, while nearer to ridgelines boulders are more prominent. Open blockfields are almost exclusively located on the south-facing slope, though rockiness tends to increase with distance down-valley on the north-facing Leading Ridge slope, where underlying bedrock dip steepens.

 

 

To aid in visualization of grain shape, sorting, and organization, we constructed structurefrom-motion photogrammetry models (Westoby et al., 2012) of three point-count sites using a digital single lens reflex camera with wide angle lens. We aligned 50-100 photographs of each site and constructed dense point clouds using Agisoft Photoscan, and scaled our models using 3-5 15 cm rulers scattered throughout the scene. Visualization of dense point clouds was performed using the EyeDome lighting shader in CloudCompare (http://danielgm.net/cc).

Cosmogenic Nuclide Analysis

To quantify the concentration of cosmogenic nuclides in regolith, we sampled material from a soil pit transect established at Garner Run as part of the larger SSHCZO investigation known as the GroundHOG (three along the north-facing slope of Leading Ridge and one on the south-facing slope of Tussey Mountain, Fig. 2b) (Brantley et al., 2016). These pits were dug to different depths – to 0.7 m on the Tussey Mountain Midslope (TMMS), to 0.65 m on the Leading Ridge Ridgetop (LRRT), to 1.4 m on the Leading Ridge Midslope (LRMS), and to 1.4 m on the Leading Ridge Valley Floor (LRVF) – based on the depth of refusal and thus integrates only a portion of the regolith overlying unweathered bedrock. We amalgamated 100 g of soil from each soil horizon identified for a given soil pit, and sieved amalgamated horizon material to the 250-

850 μm fraction sampled within the soil profile, creating a single sample for each soil pit.

We sampled surface boulders along three 30 m slope-normal transects adjacent to each of the three soil pits on the north-facing slope of Leading Ridge for both ¹⁰Be and ²⁶Al (Fig. 2b). Every two meters we removed the uppermost few centimeters of rock from the nearest boulder, sampling boulders representative of the typical boulder size on the slope (~1 m or less) for a total of 15 boulder chips per transect. Each chip was crushed and sieved individually, and 50 g of each clast was amalgamated into one sample per transect. We collected fluvial sediments ~15 m upstream and downstream from the north-facing soil transect, and sieved each sample in the field to the 250-850 μm fraction.

We sampled buried clasts recovered from a 9.1 m drill core in the valley axis of Garner Run, just upstream of the soil pit transect (Brantley et al., 2016; DiBiase et al., 2016). Material recovered during drilling consisted of discontinuous sandstone clasts dispersed throughout Festained sandy fill. Two sandstone clasts were recovered between 0-3.3 m but their depths were not recorded. Five clasts between 3.4 and 4.8 m were recovered as were four clasts between 4.8 and 6.4 m; while their precise depths are unknown, the order of their recovery was noted, as was the length of each cored section. Sand was present between depths of 6.4 and 9.1 m, with no clasts. We sampled the first clast recovered in the 3.4-4.8 m interval as well as the first and last clast recovered from the 4.8-6.4 m interval to analyze both ²⁶Al and ¹⁰Be to determine burial age and history (Granger, 2006). Because the drill casings were each 1.5 meters long, and cores were logged in the order they were recovered (rather than at discrete depths), there is some uncertainty in the depths of the individual clasts. The total length of cored clasts in the 3.4-4.8 interval was

0.38 m, and total length of cored clasts in the 4.8-6.4 interval was 0.45 m. We assigned a depth of 3.4, 4.8 and 6.4 to the top, middle and bottom clasts, respectively. However, the top and middle clasts may have been deeper and the bottom clast may have been shallower by as much as 0.5 m given the length of core recovered for each interval.

We purified quartz by heating ground samples in HCl and treating them with a series of leaches using dilute HF/HNO₃ mixtures at the University of Vermont (Kohl and Nishiizumi, 1992) and extracted ¹⁰Be and ²⁶Al following the methods of Corbett et al. (2016a). All Garner Run samples were analyzed for ¹⁰Be at Lawrence Livermore National Laboratory normalizing them relative to ICN standard 07KNSTD3110 with an assumed value of 2.85 × 10^-12 (Nishiizumi et al., 2007). Samples GR01-GR09 were analyzed in April 2016, and GR10-12 in July 2016. We corrected GR01-GR09 using an average of n=3 process blanks (6.43 ± 2.00 × 10^-16 atoms g^-1), and GR10-12 using an average of n=10 process blanks (1.35 ± 0.77 × 10^-15 atoms g^-1). We sent the drill core samples (n=3) and the surface boulder transect samples (n=3) to PRIME lab (n=6) for ²⁶Al analysis. Exposure ages were calculated using the CRONUS-Earth online calculator

(http://hess.ess.washington.edu/, wrapper script 2.2, main calculator 2.1, constants 2.2.1, see Balco et al. (2008)) based on the constant production rate model (Lal, 1991; Stone, 2000) calibrated to the northeastern United States production rate (Balco et al., 2009)(Table A1). Two samples, GR08 and GR10, were replicated for ¹⁰Be and ²⁶Al measurements at PRIME lab in May 2017.

Determining Colluvial Fill History

As a first order constraint on the age of colluvial fill in Garner Run, we divided estimates of fill volume constrained by surface topography, drill core observation, and shallow geophysics

(DiBiase et al., 2016) by the average hillslope lowering rates determined from ¹⁰Be concentrations in stream sediment. We assume no change in regolith storage on hillslopes, a colluvium density of 1500 kg m^-3, a bedrock density of 2700 kg m^-3, and we assume that the contributing area of sediment production is limited to the dip slopes that cover ~75% of the catchment area.

In order to evaluate potential fill histories consistent with cosmogenic data, we modeled the accumulation of cosmogenic ²⁶Al and ¹⁰Be in a 1-dimensional, 6.5 m-thick valley fill for comparison with buried samples from the Garner Run drill core. Specifically, we compared measured concentrations of ²⁶Al and ¹⁰Be ([²⁶Al] and [¹⁰Be]) with modeled concentrations from three scenarios: a single burial event, gradual burial, and pulsed burial. For each model scenario, we assume an inherited [²⁶Al] to [¹⁰Be] ratio of 6.5 and an inherited concentration of 4.0 x 10⁵ atoms g^{-1 10}Be (the average ratio and concentration of amalgamated boulder samples, respectively). We assume a surface production ratio of [²⁶Al] to [¹⁰Be] of 6.5. For simplicity, we assign depths of 3.4, 4.8 and 6.4 to the clasts for the following calculations but acknowledge a range in depths for each clast as discussed above. For details on modeling schemes, see the Appendix C.

Because we have some constraint on possible inherited [²⁶Al]/[¹⁰Be], we could test whether the assumption of an inheritance ratio of 6.5 is reasonable. We ran a Monte Carlo simulation in which an inheritance ratio was chosen from a normal distribution of 1,000 values with a mean of

6.5.

A RECORD OF COUPLED HILLSLOPE AND CHANNEL RESPONSE TO PLEISTOCENE PERIGLACIAL EROSION IN A SANDSTONE HEADWATER VALLEY, CENTRAL PENNSYLVANIA

Sharing is caring!