knickpoint evolution in a vertically bedded substrate ...€¦ · of individual waterfalls with up...

For permission to copy, contact [email protected]© 2007 Geological Society of America

ABSTRACT

We conducted fl ume experiments to inves-tigate the nature of knickpoint and bedrock channel longitudinal profi le evolution for a channel reach underlain by a vertically bedded substrate of alternating resistance. This research was motivated by the pres-ence, behavior, and apparent persistence of impressive knickpoints formed in steeply dipping or steeply foliated bedrock in Atlantic slope drainages of the Appalachian Mountains. The experiments were carried out in a 10-m-long wooden-box fl ume fi lled with a very fi ne– to medium-grained sand alluvial substrate and a single, 30-cm-wide ridge of vertically oriented, lacustrine silt-clay varves designed to simulate a bedrock channel reach. Numerous control experi-ments established a stable meandering chan-nel form and transport gradient, from which knickpoints were produced by a single, instantaneous base-level fall at the channel mouth. Knickpoint evolution in the alluvial material downstream of the bedrock reach was dominated by inclination of the knick-point face; this resulted in a rapid transition from a waterfall at the point of base-level fall to a broad, convex zone spanning the entire lower alluvial channel reach. A waterfall and plunge pool reformed at the contact between the lower alluvial reach and the simulated bedrock ridge where the knickpoint short-ened and steepened. The knickpoint then migrated upstream through the bedrock reach by a combination of parallel retreat and vertical channel incision. The knickpoint

evolution process resulted in the formation of upstream-dipping strath terraces, an uncommon landform possibly present in the Holtwood Gorge of the Susquehanna River. As base-level fall effects were transmitted to the alluvial channel above the bedrock reach, a complex channel response, accompanied by pulses of sediment, alternately buried and excavated the bedrock reach. These results illustrate the complex behavior associated with knickpoint evolution, the unsteadiness of the bedrock channel erosion process, and the signifi cant lag times that may exist in

natural fl uvial systems that are in the pro-cess of adjusting to base-level fall.

Keywords: Appalachian Mountains, knick-points, strath terraces, landscape evolution.

INTRODUCTION

Knickpoints and knick zones are steep reaches in river longitudinal profi les that typically form in bedrock channels in response to a base-level fall (Fig. 1) or where the channel encounters a relatively resistant substrate. Knickpoints are

GSA Bulletin; March/April 2007; v. 119; no. 3/4; p. 476–486; doi: 10.1130/B25965.1; 14 fi gures.

†Present address: Department of Earth Sciences, University of Southern California, Los Angeles, Cal-ifornia 90089, USA; e-mail: [email protected].

‡E-mail: [email protected].§Present address: Tetra Tech EM Inc., Albuquer-

que, New Mexico 87110, USA; e-mail: [email protected].

Knickpoint evolution in a vertically bedded substrate, upstream-dipping terraces, and Atlantic slope bedrock channels

Kurt L. Frankel†

Frank J. Pazzaglia‡

Jordan D. Vaughn§

Department of Earth and Environmental Sciences, Lehigh University, Bethlehem, Pennsylvania 18015, USA

A B

C Dchannel profile at time 0channel profile at time 1channel profile at time 2

τc << τoτc >> τo

τc < τo

τc > τo

τc << τo

τc < τo

Rotation Relaxation

Parallel Retreat Replacement

Figure 1. Schematic diagram illustrating the three end-member models of knickpoint evolu-tion, after Gardner (1983). The processes controlling knickpoint evolution are controlled by the critical shear stress of the channel bottom (τc) and the actual shear stress acting on the channel bottom (τo). (A–B) The process of inclination, which can occur through rotation (A) in a homogeneous nonresistant substrate or through relaxation (B) in a homogeneous highly resistant substrate. (C) The case of parallel retreat can occur when the substrate is composed of horizontally bedded alternating resistant and nonresistant layers. (D) Knickpoints evolve through the process of replacement in a homogeneous substrate of intermediate resistance.

476

Knickpoint evolution in a vertically bedded substrate

Geological Society of America Bulletin, March/April 2007 477

commonly convex or vertical reaches, in the extreme case of waterfalls, that interrupt the typically concave profi les of graded rivers where fl uvial erosive power is maximized by the steep channel gradients. The local steepening associ-ated with knickpoints concentrates fl uvial ero-sive energy, resulting in changes to the form of the stream; abrasion, plucking, pothole forma-tion, and cavitation lower the channel and cause upstream retreat and vertical down-wearing of the knickpoint face (Hancock et al., 1998). The temporal and spatial scales of retreat and inci-sion depend on the type of channel perturba-tion that generates the knickpoint, the overall rate of fl uvial incision, which tends to be highly correlated to the rate of rock uplift, the overall

resistance of the underlying rock type, and chan-nel bed erosion processes (Whipple, 2001).

Gardner (1983, and references therein) identi-fi ed three end-member scenarios for knickpoint behavior in a homogeneous, horizontally bedded substrate of variable resistance: (1) inclination, which occurs either by rotation of the knick-point about the mid-point or base of the origi-nally vertical knickpoint face (Figs. 1A and 1B); (2) parallel retreat, where the face of the knick-point migrates upstream while remaining verti-cal (Fig. 1C); and (3) replacement, in which the knickpoint wears down vertically just upstream of the lip while reclining in an upstream direc-tion from a fi xed position at its base (Fig. 1D). In all cases, the process controlling these three

knickpoint evolution pathways is a function of the critical shear stress (τ

c) on the channel fl oor

that is needed for erosion to take place, fi xed by the substrate, and the actual shear stress (τ

o)

acting on the exposed bedrock (Gardner, 1983; Schumm et al., 1987; Leopold et al., 1992). Rotation occurs when τ

c << τ

o in a homoge-

neous nonresistant substrate (Fig. 1A). Relax-ation will take place when τ

c >> τ

o upstream of

the knickpoint and τc < τ

o at the knickpoint face

in a highly resistant, homogeneous substrate (Fig. 1A). In the case of a horizontally layered substrate of resistant and nonresistant beds where τ

c > τ

o in the resistant layers and τ

c << τ

o in

the nonresistant layers, undercutting of the face occurs, which allows the knickpoint to migrate upstream through parallel retreat (Fig. 1C). This scenario is only favored in the case where a hard cap rock overlies a nonresistant layer, such as Niagara Falls (Tinkler et al., 1994). The fi nal case of replacement occurs when τ

c < τ

o in a

homogeneous, moderately resistant substrate (Fig. 1D).

While Gardner’s (1983) model describes the evolution of knickpoints in homogeneous or horizontally bedded substrates, it does not directly address the effects of vertically bedded or foliated bedrock. This study focuses on the effects of steep bedding on knickpoint forma-tion and evolution, a condition pertinent to the impressive knickpoints in Atlantic slope chan-nels (Reed, 1981) where steeply bedded or foli-ated substrates are common.

KNICKPOINTS AND THE APPALACHIAN LANDSCAPE

The Appalachian landscape is dissected by bedrock channels where the long-term rates of river incision and landscape erosion range from 5 to 50 m/m.y., with an average rate of ~20 m/m.y. (Fig. 2; Poag and Sevon, 1989; Sevon, 1989; Hulver, 1997; Springer et al., 1997; Conrad and Saunderson, 1999; Mills, 2000; Matmon et al., 2003; Ward et al., 2005). At the drainage-basin scale, south of the glacial boundary, Appalachian rivers are generally well adjusted to rock type and structure and have formed graded, concave-up longitudinal profi les within broad, long strike valleys in relatively soft rocks, parallel to fold axes, or along dominate joint orientations. Simi-larly, at shorter reach scales, as in the case of tight meander loops, channels are controlled by bed-ding, foliation, or related structures of the bed-rock (Braun, 1983). Nevertheless, Appalachian drainages also exhibit impressive examples of landscape disequilibrium in the form of locally steeper reaches, which can range from steep waterfalls in the middle of otherwise graded val-leys to the broad convexities of major rivers that

82° W

82° W

80° W

80° W

78° W

78° W

76° W

76° W

74° W

74° W

36° N36° N

38° N38° N

40° N40° N

42° N42° N

OH

WV

PA

VA

NC

NY

NJMD

DE

Roanoke

R.

James

Potomac R.

hannaS

usque

R.

M

R.

H

NewR.

Zone

Blue

Rid

ge

Fal

l

Ridge

and

Val

ley

Lake Erie

CB

Ohio

R.

100 km

Pie

dmon

t

Figure 2. Shaded 90-m-resolution digital elevation model (source: Shuttle Radar Topogra-phy Mission [SRTM] U.S. Geological Survey, http://seamless.usgs.gov/) of the Appalachian Mountains and the Atlantic passive margin of the eastern United States. The major water-sheds pertinent to the study are labeled. The box labeled M shows the location of the Maury River and Moores Creek in Figures 3 and 4, both of which are tributaries to the James River. The star labeled H indicates the location of the Susquehanna River and Holtwood Gorge area in Figures 3 and 5. NY—New York, PA—Pennsylvania, OH—Ohio, WV—West Virginia, VA—Virginia, MD—Maryland, NJ—New Jersey, DE—Delaware, NC—North Carolina, and CB—Chesapeake Bay.

Frankel et al.

478 Geological Society of America Bulletin, March/April 2007

fl ow perpendicular to the prevailing structural grain of the orogen across steeply dipping or foliated rocks (Fig. 3), such as the Susquehanna, Potomac, and James Rivers as they carve gorges across the Piedmont and Fall Zone (Fig. 2; Reed, 1981; Pazzaglia et al., 1998). An important goal of this study is to understand the distribution and form of Appalachian knickpoints by investigat-ing the knickpoint erosion processes acting on hard, but variably resistant substrates.

Our experiments were inspired by two Appa-lachian knickpoints that have contrasting size but are underlain by resistant substrates with similar bedding characteristics. The Ridge and Valley Province west of the Blue Ridge in south-western Virginia (Fig. 2) is underlain by steeply dipping Ordovician carbonates with local valley relief of ~30 m (Harbor et al., 2000). The South River is an incised strike-valley stream that drains this portion of the Great Valley south-westward into the Maury River, a main tributary of the James River (Fig. 2). Tributaries of the South River, such as Moores Creek, fall off the Great Valley upland as steep, 50-m-high knick-points that, at the reach scale, are constructed of individual waterfalls with up to 30 m of local relief (Figs. 3A and 4). Upstream of the water-falls, Moores Creek fl ows for ~0.5 km through a step-pool sequence before the channel gradient diminishes and the valley widens (Figs. 3A and 4). The only obvious structural control on the position of these knickpoints is that they com-monly coincide with more resistant, steeply bedded substrates. The longitudinal profi le form that we wish to draw attention to here is the broadly convex shape punctuated by shorter, nearly vertical segments (Fig. 3A).

Similarly, but on a much larger scale, the Susquehanna River at Holtwood Gorge, Penn-sylvania, descends ~20 m over ~10 km (gradient = 0.002) within a narrow, 200-m-deep bedrock gorge 40 km upstream from the Chesapeake Bay tidewater (Figs. 3B and 5). Here, the chan-nel is marked by elongate islands with smooth, rounded tops that defi ne a crude summit surface that merges with the channel upstream (Fig. 6). Holtwood Gorge is formed in the Ordovician Wissahickon Schist, a resistant, chlorite-mus-covite-garnet schist with a well-developed foliation that dips steeply west (upstream). This lithology underlies many of the rivers fl owing over the Pennsylvania Piedmont. As in the case for Moores Creek, there is no obvious structural control for the location of the Holtwood Gorge knickpoint, which is a very steep feature embed-ded in an overall broadly convex longitudinal profi le (Fig. 3B). Moores Creek and the Holt-wood Gorge of the Susquehanna River (Figs. 4 and 5) are two of many examples of knickpoints that have formed on steeply dipping substrates

in Appalachian drainage systems. In fact, many such cases exist along the eastern fl ank of the Appalachian Mountains where knickpoints are present as channels cross resistant substrates, as is the case for the Potomac and James Rivers (Reed, 1981).

Previous work by Miller (1991) is particularly pertinent to our approach because it addresses the issue of lithologic controls on knickpoint evolu-tion in response to base-level fall of the Ohio River. The bedrock in the study by Miller (1991) is relatively fl at lying; however, it contains many

vertical fractures that introduce an anisotropy that differs between the carbonates and siliciclastics. Miller et al. (1990) and Miller (1991) showed that knickpoint frequency and distribution can be controlled entirely by bedrock parameters; a single episode of base-level fall can result in multiple knickpoints rather than the traditional view that clusters of knickpoints in this land-scape have formed in response to climate-driven river incision (Gooding, 1971).

Our study aims to test the general con-clusions of Miller (1991) in investigating

distance (km)050100150200250300350400

elev

atio

n (m

)

0

50

100

150

200

250

300

distance (km)

elev

atio

n (m

)

3000123456789101112

350

400

450

500

550

600

ConowingoDam

HoltwoodDam

Safe HarborDam

Holtwood Gorgeknickpoint

Harrisburg

confluence ofNorth and West Branches

Susquehanna River

confluence with Juniata River

South River

Moores Creek

confluence

Moores Creekknickpoint

location ofFigure 4

B

A

Figure 3. (A) Channel longitudinal profi le of Moores Creek. The circle denotes the location of the photographs in Figure 4. (B) Channel longitudinal profi le of the Susquehanna River. The location of the photographs in Figure 5 is shown by the open circle. Both long profi les were extracted from the 30 m Shuttle Radar Topography Mission (SRTM) digital elevation model (http://seamless.usgs.gov/) using ArcInfo. A fi lled 30 m raster was analyzed with the hydrology tools to defi ne the drainage network, after which the trunk channel was isolated, converted to a coverage, and then used to extract the longitudinal profi le using the surface tools.



Appalachian knickpoint behavior. The goal is not to replicate any specifi c location in the Appalachians but rather to use conditions observed in this landscape as boundary char-acteristics for our model. We accomplish this using an experimental channel in a fl ume, not scaled specifi cally to natural systems but

designed in similar fashion to previous base-level fall analogue models (Begin et al., 1981; Gardner, 1983; Schumm et al., 1987) to mimic a steeply dipping, resistant lithology represen-tative of the fi eld conditions observed along both Moores Creek and the Susquehanna River, among other Appalachian rivers.

METHODS

The experiments in this study were motivated by the presence of spectacular knickpoints devel-oped in reaches of vertically dipping bedrock in rivers draining the Appalachians. As such, our experiments were designed so that we observed the spatial and temporal behavior of knickpoints as they migrated upstream through a section of channel underlain by alternating resistant and nonresistant, vertically dipping beds, similar to many Appalachian streams. We fi rst ran control experiments in a purely alluvial channel to deter-mine the most stable confi gurations of discharge and channel slope. After an acceptable combina-tion of these two variables was determined, we emplaced a reach of bedrock in the otherwise alluvial channel system. Again, a control run was completed so that the channel equilibrated to the addition of the bedrock reach. Once a stable, meandering pattern was established, base level was lowered and the channel was allowed to adjust to its new boundary conditions.

Experimental Apparatus

We used a fl ume that was 10 m long and 1 m wide (Fig. 7; http://www.lehigh.edu/~fjp3/fl ume.html). The front 0.8 m of the fl ume con-sisted of a pool of water, into which the mouth of

A

B

Figure 4. (A) Photograph looking upstream at a knickpoint in vertically bedded Ordovician carbonates along Moores Creek. This knickpoint is representative of many waterfalls in the Blue Ridge. Local relief at the knickpoint is ~30 m. Note person at the base of the waterfall for scale. (B) Photograph of step-pools formed in the exposed bedrock channel upstream of the knickpoint in A. Photograph was taken looking upstream. Step-pools with meter-scale relief are commonly observed in reaches ≥0.5 km upstream of major knickpoints in Appalachian drainages.

Figure 5. Photograph looking downstream through the vertically foliated, incised bedrock near Holtwood Gorge. The view shows an example of a typical reach of incised bedrock channel along the Susquehanna River. Note the strath terraces.

Frankel et al.


the stream fl owed, which allowed a delta to form in the process. A wooden plank at the front of the fl ume was raised or lowered in order to change base level by adjusting the maximum height of water in the pool (Fig. 7A). Water fl owed from a constant-head tank into a hose (Fig. 7A). From the hose, the water continued through a fl ow diffuser made of gravel and into the head of the channel (Fig. 7A). The channel consisted primarily of very fi ne– to medium-grained sand (grain size = 100–500 μm). A cohesive ridge of varved lacustrine sediments, collected from Pleistocene proglacial lake sediments in Tioga County, Pennsylvania, served as a proxy for the vertically bedded or foliated bedrock pres-ent both at Moores Creek and Holtwood Gorge (Fig. 7B). The varved sediments were set into the fl ume, stretching from side wall to side wall, forming a 30-cm-long reach such that the 1–2-mm-thick alternating beds of clay and silt dipped vertically and had a strike perpendicular to the stream-fl ow direction.

The alternating clay and silt varves were a suitable proxy for bedrock in our experiments because they are more resistant to fl uvial erosion than the sand underlying the rest of the channel system. However, the varves were not so resis-tant that the experiment had to run for a prohibi-tively long period of time in order for the stream to incise and the knickpoint to evolve through the simulated bedrock reach. The varves also made for a relatively simple way to reproduce heterogeneous, vertically dipping sedimentary bedrock of variable resistance at a reasonable experimental scale.

We measured channel longitudinal profi les throughout the stream stabilization and base-level fall experiments. The longitudinal profi les were determined by measuring the depth to the channel bottom from cross-pieces attached to the fl ume walls above the channel that remained at a constant height throughout the experiments (Fig. 7A). Measurement spacing varied from 0.5 to 2.4 m. A meter stick with an estimated precision of ±2 mm was used to make the mea-surements. Our goal here was not to model lon-gitudinal profi le gradients precisely, but rather to observe the large changes in channel incision and knickpoint evolution process. Therefore, given the total relief of 10 cm or more from the channel head to the delta, our measurement method provided the appropriate precision.

Control Runs

We could not adjust the gradient of our fl ume by tilting it during an experiment. Therefore, we performed a series of control runs to determine the optimum combination of discharge and channel slope so that a stable channel could be

established for the prevailing grain-size distribu-tion. We defi ned a stable channel as one with slowly sweeping meanders, nonslumping banks, and minimal sediment transport. We measured relative sediment transport by observing delta

growth or stability. Control runs began with a stable discharge of 0.0002 m3/s (0.2 L/s) and an average channel gradient of 0.003.

Control experiments involved channel devel-opment under a constant base level and discharge

Holtwood DamLake

Aldred

island summits

strath

HenneryIsland

Piney

Island

LowerBear

Island

UpperBear

Island

BigChestnut

Island

NW SE

distance downstream (m)

0 1000 2000 3000 4000 5000 6000el

evat

ion

(m)

30

40

50

60

70

flow direction

A

B

10 m

eter

s

1 meter

0.8 m

~25 cm

12

3

4 5

67

8

30 cm

Figure 6. Topographic profi le from 30 m Shuttle Radar Topography Mission (SRTM) digital elevation model (http://seamless.usgs.gov) of strath terraces superimposed on the longitudi-nal channel profi le of the Susquehanna River in the Holtwood Gorge area. Bedrock islands in the channel preserve an upstream-dipping suite of correlative strath terraces.

Figure 7. (A) Photograph of the experimental fl ume apparatus used in the study. See text for discussion. Black arrow indicates the direction of stream fl ow. 1—hose connected to constant-head tank, 2—location of the fl ow diffuser, 3—simulated bedrock reach, 4—base-level control pool, 5—wooden plank used to control base level, 6—channel bottom elevation measurement locations, 7—alluvial reach downstream of simulated bedrock, 8—alluvial channel reach upstream of simulated bedrock. (B) Close-up view of the simulated bedrock reach in the fl ume channel.



for ~30 h, at which point, base level was lowered by 6 cm, causing a new, stable, albeit incised, channel to develop. Following another ~30 h of constant discharge and base level, the fi nal channel had a full-length-fl ume channel slope of 0.017. The purpose behind these experiments was to determine how a purely alluvial channel would respond to a drop in base level for com-parison with a channel system including a bed-rock reach. We took this channel gradient to be the stable form that served as the starting point for the base-level fall experiments inclusive of the bedrock reach.

For the bedrock incision experiments, we buried the ridge of varved lacustrine sediments 1.0 m upstream of the channel mouth (before delta progradation), excavated a 25-cm-wide, 2-cm-deep, artifi cially cut valley to allow for a contiguous channel (Fig. 7B), and covered the simulated bedrock reach with a thin, ~1-mm-thick veneer of alluvium. We then directed water back into the channel, which evolved under stable base level and discharge conditions for ~30 h. We refer to this part of the experiment as the stabilization phase. Upon establishment of a stable channel that was fully integrated across the bedrock reach, base level was dropped 6 cm at the channel mouth, causing a knickpoint to form and setting up the conditions for observing fl uvial modifi cation of the bedrock channel.

RESULTS

Channel Stabilization

We noted three responses during the stabili-zation phase: (1) the bedrock reach deepened

overall by ~2 cm, (2) pulses of sediment from the upper alluvial channel made their way through the bedrock reach, resulting in channel incision and subsequent aggradation (Fig. 8), and (3) delta lobe switching resulted in minor base-level fl uctuations, none of which produced knickpoints that migrated upstream to the bed-rock reach. The combination of deepening and upstream sediment pulses produced a narrow inner channel fl anked by broad, fl at strath ter-races (Figs. 9A and 9B). Also accompanying the incision of the bedrock reach was a shal-low plunge pool that formed at the downstream interface between the bedrock and alluvial reaches (Fig. 9C). In addition, small episodes of alluvial aggradation and excavation occurred along the bedrock reach after the formation of an inner canyon (Figs. 9D and 9E).

Base-Level Fall with a Simulated Bedrock Reach

Following the establishment of a stable, meandering channel, base-level was instanta-neously lowered by 6 cm, and the system was left to respond for ~28 h (Fig. 10). This drop in base level was meant to serve as an analogue for the case of eustatic fall at the Appalachian Fall Zone (Fig. 2) in eastern North America during sea-level lowstands, which caused riv-ers to both lengthen their channels and incise across an emergent continental shelf (Schumm, 1993; Genau et al., 1994). Immediately upon base-level fall, a knickpoint formed at the chan-nel mouth because the slope of the delta front exposed by the base-level fall was steeper than the channel slope upstream. The knickpoint

propagated rapidly upstream through the lower alluvial channel reach by inclination, covering the 1 m distance between the delta mouth and the simulated bedrock in <1 min (Figs. 1 and 11A). Upon arrival at the bedrock ridge, the knickpoint immediately steepened, regained its vertical face, and formed a deep plunge pool at the downstream bedrock-alluvium interface (Fig. 11B).

The knickpoint then propagated through the bedrock reach by a combination of vertical inci-sion of the bedrock fl oor and parallel retreat of the knickpoint face. The parallel retreat process was unsteady in the sense that the more resistant clay layers impeded face retreat until undercut-ting in the plunge pool exposed the adjacent silt layer, causing mass failure and upstream migra-tion of the knickpoint face. At the same time, abrasion and plucking vertically lowered the channel fl oor upstream of the knickpoint face when the bedrock channel fl oor was exposed (Fig. 11C). The bedrock canyon became incised and narrowed from its initial width of 25 cm to 8.25 cm. The alluvial channel upstream of the bedrock reach was transformed into a series of pools and riffl es atop a broad channel convex-ity (Figs. 11D and 12). The development and subsequent excavation of the pools sent pulses of sediment downstream, causing episodic cut and fi ll cycles and resulting in several episodes of burial and re-excavation of the bedrock reach (as in Figs. 9D and 9E). While buried by allu-vium, no incision of the bedrock reach took place, and the channel was completely alluvial from head to mouth. During times of gorge excavation, bedrock channel incision resulted in gently upstream-dipping bedrock strath terraces

mouth headwaters

bedr

ock

reac

h

distance upstream (cm)

1000 200 300 400 500 600 700 800

chan

nel b

otto

m e

leva

tion

(cm

)

-12

-14

-16

-18

-20

-22

-24

0.0 hours3.25 hours9.25 hours14.25 hours25.0 hours32.67 hours

delta progradation

Figure 8. Longitudinal channel profi les at various time intervals during the stream stabilization phase of the experiment. The simulated bedrock reach is shown by the gray box. During the stabilization process, the bedrock reach was incised ~2 cm. Small pulses of sed-iment were alternately deposited in, and eroded from, the bedrock reach. In addition, small knick zones formed in response to delta lobe switching.

Frankel et al.


A

D

B

C

0 hours ~3.25 hours

~9.25 hours ~25 hoursFigure 9. Photographic time series of the stabilization phase of the experiment. Lens cap in photographs is 72 mm in diameter for scale. Black arrows on the side of the channel indicate fl ow direction. (A) Simulated bedrock reach at the beginning of channel stabilization. (B) A strath terrace forms (white arrow) as the channel incises vertically through the bedrock, forming an inner canyon. (C) Minor incision begins in the bedrock reach, and a small plunge pool forms (white arrow) as the channel begins to stabilize. (D) The canyon fi lls with a pulse of alluvium from upstream, burying the bedrock channel fl oor in the fashion of a complex response (e.g., Schumm and Rea, 1995).

mouth headwaters

chan

nel b

otto

m e

leva

tion

(cm

)


bedr

ock

reac

h

1000 200 300 400 500 600 700 800

-12

-14

-16

-18

-20

-22

-24

-26

-28

0.0 hours0.37 hours1.33 hours4.0 hours10.25 hours13.42 hours21.08 hours24.75 hours28.33 hours

delta progradation

Figure 10. Longitudinal channel profi les at various time intervals after base level was instantaneously lowered 6 cm. The location of the simulated bedrock reach is shown by the gray box. The longitudinal pro-fi les display the formation of a knickpoint and erosion of that knickpoint through a combination of replace-ment and parallel retreat. Additionally, a complex channel response consisting of a series of aggrada-tion/degradation events in the bedrock reach of the channel is observed in the profi les.



A ~ 0.01 hours

B ~0.37 hours

F >28.33 hours

C ~4 hours D ~13.42 hours

E ~24.75 hoursFigure 11. Photographic time series of the base-level fall phase of the experiment. Lens cap in photographs is 72 mm in diameter for scale. Arrows on sides of channels indicate fl ow direction. (A) Base level is instantaneously lowered 6 cm and a knick zone (white arrow) immedi-ately begins to propagate upstream through the lower alluvial reach by the process of inclination. (B) Channel incises through the bedrock canyon in response to upstream knickpoint migration subsequent to the base-level fall. (C) A knickpoint, plunge pool, and strath terrace form as the stream continues to incise through the simulated bedrock reach in response to the base-level fall. (D) Small pools and riffl es (bed forms indicated by white arrow) migrate upstream through the alluvial, upper reach of channel, similar to the processes occurring in Moores Creek in Figure 2B. (E) A pulse of sediment covers the bottom of the bedrock gorge. The sediment is deposited in the canyon when incision occurs upstream as a result of oversteepening of the broad channel convexity associated with the step-pools in D. (F) Fluvial features preserved in the bedrock reach at the conclusion of the experiment. Arrows point to strath terraces and upstream-dipping terraces formed in bedrock. Dashed white line highlights an upstream-dipping terrace that converges with the active channel bottom.

Frankel et al.


(Fig. 11E). The fi nal channel gradient with the simulated bedrock reach was 0.015, similar to the gradient of 0.017 attained during the control runs in the purely alluvial channel (Fig. 13).

DISCUSSION

Knickpoint Evolution

The knickpoint in our experiment followed an evolutionary pathway generally consistent with the process of replacement proposed by Gardner (1983). Replacement requires both vertical lowering of the bedrock channel bottom and upstream migration of the knickpoint face. Erosion in the plunge pool alternates between slow removal of the resistant clay layers and rapid undercutting of the soft silt strata. The undercutting process, in particular, effi ciently removes material from the knickpoint, thereby allowing the knickpoint to maintain a steep face. At the same time, abrasion and plucking on the exposed bedrock channel fl oor drives verti-cal incision, a process that works to reduce the height of the knickpoint face. It is possible that microrelief between the silt and clay laminae protruding up from the bedrock channel fl oor enhances the plucking process acting upstream of the knickpoint. Pieces of the simulated bed-rock were observed to break off into the stream fl ow, suggesting that the process of plucking enhanced vertical incision through the bedrock reach during our experiments.

Other erosional processes, such as wave action in the pool at the mouth of the fl ume and groundwater sapping, were negligible and did not generate any additional knickpoints throughout the base-level fall run. Delta aggra-dation outpaced wave erosion. Furthermore, no features consistent with erosion by groundwater sapping (e.g., head scarps, etc.) were observed during the experiment. The low permeability of the silt and clay layers in the simulated bedrock reach of our fl ume would have made it nearly impossible for such processes to occur in this portion of the channel.

The experiments resulted in a broadly convex longitudinal profi le punctuated by steep, nearly vertical knickpoints (Figs. 8 and 10). In this way, the experimental results reproduced a lon-gitudinal profi le common to many Atlantic slope drainages ranging in scale from Moores Creek to the Susquehanna River (Fig. 3). Given that both in the fl ume and in the Appalachian landscape, the steepest knickpoints were restricted to those regions characterized by resistant, vertically bedded substrates, we conclude that this partic-ular channel confi guration imparts a dominant control on knickpoint behavior and evolution.

mouth headwaters

bedr

ock

reac

h

chan

nel b

otto

m e

leva

tion

(cm

)


channel profile at 13.42 hourschannel profile at 24.75 hours

broad convexity

incision

sedimentpulse

0 100 200 300 400 500 600 700

-5

-10

-15

-20

-25

-30

plungepool

Figure 12. Channel longitudinal profi les at 13.42 and 24.75 h after base-level fall. The channel profi le measured at 13.42 h shows the broad convexity formed in the alluvial reach upstream of the simulated bedrock. The longitudinal profi le measured at 24.75 h illustrates incision of the stream through the convexity due to oversteepening of the channel. As the upper channel incises, pulses of sediment are delivered downstream, causing aggradation to occur through the bedrock gorge and in the plunge pool directly downstream of the bedrock reach.

final channel profile without bedrock reachfinal channel profile with bedrock reach

bedrock reach


1000 200 300 400 500 600 700 800

mouth headwaters

chan

nel b

otto

m e

leva

tion

(cm

)

-12

-14

-16

-18

-20

-22

-24

-26

-28

Figure 13. Final channel longitudinal profi les for the alluvial calibration channel without bed-rock and for the experimental channel with a simulated bedrock reach. The purely alluvial channel has a fi nal slope of 0.017, while the channel with a bedrock reach has a fi nal slope of 0.015, suggesting that regardless of substrate, channels will eventually grade to the same slope holding all other variables equal (i.e., base-level drop, discharge, grain size, etc.).



The observed replacement process differs spe-cifi cally from the Gardner (1983) example in that the knickpoint face remained more or less vertical, which led to the formation of a peculiar landform as the knickpoint migrated up-channel. Although we do not have rate data to directly support the following argument, we do base it on qualita-tive observations of the knickpoint behavior. We fi rst envision a case where vertical bed lowering occurs at a rate similar to, or faster than, the rate of parallel retreat. In this scenario, the knickpoint lip will lower in elevation as it moves upstream (Fig. 14). The result will be a strath terrace carved into the wall of the gorge that dips upstream and is time-transgressive; the downstream por-tion is higher and older than the upstream por-tion, which is zero age where it merges with the active channel (Fig. 11E). In contrast, we envi-sion a second case where downstream-dipping straths form when the parallel retreat component of knickpoint replacement does not maintain a vertical face and outpaces the vertical lowering of the bedrock channel fl oor upstream of the knickpoint (e.g., Gardner, 1983). For the case of parallel retreat resulting from a resistant cap rock overlying a more easily erodible substrate (Fig. 1C), where the knickpoint face remains ver-tical, upstream-dipping strath terraces would not be expected because the amount of vertical inci-sion would be negligible compared to the rate of upstream migration of the knickpoint face.

Our documented case of upstream-dip-ping straths complements the Gardner (1983)

classifi cation by adding a fourth scenario to the general behavior of knickpoints. The results pre-sented here suggest that it is the vertical bedding in the simulated bedrock reach that plays the key role in maintaining a steep knickpoint face and slowing the rate of knickpoint retreat such that it is commensurate with the rate of vertical channel incision. Our experiment differs from that of Gardner (1983) in that the substrate used was laterally heterogeneous owing to the verti-cal interbedding between the relatively resistant clay laminae and more easily erodible silt layers. This bedding geometry did not modify the shear stresses acting on the bedrock channel fl oor; however, it may have signifi cantly infl uenced the critical shear stress in the plunge pool at the base of the knickpoint. Although outside the scope of this study, future investigations employing a numerical model with a shear stress or stream power erosion rule would be well suited to fur-ther quantify knickpoint behavior similar to what was produced by our analogue model.

Upstream-dipping straths in the fl ume experi-ment are reminiscent of the upstream dip of island summits through Holtwood Gorge, which suggests that a replacement process similar to the one we observe in our model has shaped the formation of this major knickpoint on the Susquehanna River (Figs. 6 and 11E). There is a noteworthy lack of documentation of upstream-dipping straths in the literature, aside from those that have been tectonically tilted. Holtwood Gorge presents a special case where terraces are particularly well preserved, and the vertical and horizontal rates of knickpoint evolution are deli-cately adjusted to favor formation of upstream-dipping straths. The fact that similar landforms have not been described for other rivers argues that they are either very rare, or perhaps they have not been recognized. The latter possibility is intriguing because it demands that our results be tested by further experimentation and fi eld observations. In doing so, classic downstream-dipping strath correlations should be critically challenged given that we have demonstrated a process capable of generating an alternative ter-race geometry. The precise correlation of island tops through Holtwood Gorge remains open to interpretation; however, recent cosmogenic nuclide surface exposure ages of at least one upstream-dipping strath terrace, which merges with the modern channel, clearly show a trend that becomes younger in the upstream direction (Reusser et al., 2006).

Complex Channel Response

In both Appalachian drainages and in our fl ume experiments, step-pool–like features form upstream of the main bedrock knickpoint

(Figs. 4B and 11D). These small perturbations to the channel morphology over relatively long distances act in concert with the large knick-point to transmit the base-level fall signal from the mouth of the catchment to its headwaters. Alternatively, the features observed in our experiments could be bed forms that developed in response to the high sediment load upstream of the bedrock reach. Once formed, the step-pools form a broad convexity in the channel (Fig. 12). Eventually this portion of the chan-nel becomes oversteepened, and the channel is forced to incise (Fig. 12). The result of this incision is downstream transport of large sedi-ment pulses (Figs. 9D and 12). This episodic sediment delivery occurs rapidly at fi rst, and the frequency of these sediment pulses quickly decays as the channel adjusts to its new base level (e.g., Schumm and Rea, 1995). As a con-sequence, the bottom of the bedrock gorge being carved in our experiment was periodi-cally covered by sediment, which would sub-sequently be excavated as the channel contin-ued to respond to the base-level fall (Figs. 9D and 9E). During times of sediment cover in the bedrock canyon, there was no parallel retreat or vertical incision. Upstream knickpoint retreat could therefore only take place when the can-yon was devoid of sediment.

Our results complement recent studies provid-ing evidence for rapid, recent channel incision for both the Holtwood Gorge on the Susque-hanna River and Mather Gorge–Great Falls on the Potomac River (Reusser et al., 2004). Knickpoint erosion in the fl ume was unsteady both in response to the variable resistance of the vertically dipping strata as well as the complex sediment response that periodically covered the bedrock channel with alluvium. The overall inci-sion rates of Atlantic slope channels are slow and commensurate with the long-term rates of Appalachian unroofi ng (Poag and Sevon, 1989; Sevon, 1989; Hulver, 1997; Springer et al., 1997; Conrad and Saunderson, 1999; Mills, 2000; Matmon et al., 2003; Ward et al., 2005). How-ever, there are brief periods of time, such as the past 20 k.y., when the channel beds are exposed and hydrologic conditions favor rapid knickpoint erosion processes (Reusser et al., 2004).

CONCLUSIONS

Our base-level fall experiments highlight the processes by which knickpoints evolve and migrate upstream in a bedrock channel with a steeply dipping substrate of alternating resis-tant and nonresistant layers. Furthermore, our results suggest the addition of a fourth class to Gardner’s (1983) models of knickpoint evo-lution. Evolution of the channel following a

knickpoint lip

resistantnonresistant

Figure 14. Schematic diagram illustrat-ing how knickpoints evolve and migrate upstream through a vertically bedded sub-strate. The knickpoint evolves through a process of replacement (Gardner, 1983) where the rate of vertical incision upstream of the knickpoint face is equal to, or slightly faster than, the rate of parallel retreat. The thin dashed lines connect the lip of the knickpoint at three separate time intervals, illustrating how upstream-dipping terraces are formed. See text for further discussion.

Frankel et al.


base-level fall in our fl ume illustrates that both vertical incision upstream of the knickpoint as well as parallel retreat of the knickpoint face occur as a base-level fall signal is transmit-ted through the system. The knickpoint in our experiment evolved through a replacement process where the knickpoint face remained steep, but its upstream-retreat rate was simi-lar to the rate of vertical channel bed lower-ing above the knickpoint lip (Fig. 14). This behavior is attributable to the vertical orienta-tion of alternating resistant and nonresistant beds and is also responsible for the formation of upstream-dipping terraces (Fig. 14). The resulting longitudinal profi le form is broadly convex and punctuated by nearly vertical knickpoints. This fl ume-generated longitudi-nal profi le mimics those common for Atlantic slope drainages ranging in scale from Moores Creek (12 km long) to the Susquehanna River (400 km long).

Knickpoint evolution processes infl uence drainage-basin–scale erosional response to base-level fall. In the Appalachians, where base-level fall is driven by slow, steady iso-static rise of the orogen in response to unroof-ing and late Cenozoic eustatic fall (Pazzaglia and Gardner, 1994), steady exhumation is expected for those drainages where knickpoints evolve through a replacement process such as the one observed in our experiment. Hillslope and fl uvial gradients are likely to be uniform and erosion rates steady in these landscapes. In contrast, when a parallel retreat process domi-nates, erosion and landscape response times are rapid at, and just below, the knickpoint, where stream gradients and hillslopes are particularly steep. Upstream of the knickpoint, the land-scape erodes more slowly. Overall, catchments affected by this kind of knickpoint evolution experience unsteady and nonuniform erosion. Additionally, a base-level fall signal may be transmitted by the inclination of a knickpoint into a broad knick zone. In this case, the drain-age system will respond to the base-level fall over a large area, albeit more slowly because the signal is dampened and spread over more of the catchment area. As a result, the base-level fall will affect the landscape over longer periods of time, especially when landscape response times are relatively slow. Emerging data on long-term rates of erosion and exhuma-tion from the Appalachians, and other orogens, should be placed in the context of the processes of knickpoint evolution controlling the pace of landscape erosion as suggested by our experi-mental data and fi eld observations.

ACKNOWLEDGMENTS

This project was supported in part by National Sci-ence Foundation (NSF) grant EAR-9909393. We would like to thank the students of the spring 2000 Lehigh University advanced tectonic and fl uvial geomorphol-ogy course for building the initial fl ume facility, from which ours was modifi ed. Special thanks are due to Dan Zeroka, Gerard Lennon, and the Department of Civil and Environmental Engineering at Lehigh Uni-versity for allowing the use of the hydraulics laboratory facilities. Duane Braun gave us the idea to use varved lacustrine sediments to simulate bedrock in our fl ume and provided us with advice on the best location to col-lect the sediments used in our experiment. We thank Tom Gardner and Associate Editor Jim O’Connor for critical, thought-provoking reviews that helped clarify and improve the manuscript.

REFERENCES CITED

Begin, Z.B., Meyer, D.F., and Schumm, S.A., 1981, Devel-opment of longitudinal profi les of alluvial channels in response to base-level lowering: Earth Surface Pro-cesses and Landforms, v. 6, p. 49–68.

Braun, D.D., 1983, Lithologic control of bedrock meander dimensions in the Appalachian Valley and Ridge Prov-ince: Earth Surface Processes and Landforms, v. 8, p. 223–237.

Conrad, C.T., and Saunderson, H.C., 1999, Temporal and spatial variation in suspended sediment yields from eastern North America, in Smith, B.J., Whalley, W.B., and Warke, P.A., eds., Uplift, Erosion, and Stability: Perspectives on Long-Term Landscape Development: Geological Society of London Special Publications 162, p. 219–228.

Gardner, T.W., 1983, Experimental study of knickpoint and longitudinal profi le evolution in cohesive, homo-geneous material: Geological Society of America Bulletin, v. 94, p. 664–672, doi: 10.1130/0016-7606(1983)94<664:ESOKAL>2.0.CO;2.

Genau, R.B., Madsen, J.A., McGeary, S., and Wehmiller, J.F., 1994, Seismic-refl ection identifi cation of Susque-hanna River paleochannels on the Mid-Atlantic Coastal Plain: Quaternary Research, v. 42, p. 166–175, doi: 10.1006/qres.1994.1066.

Gooding, A.M., 1971, Post-glacial alluvial history in the Upper Whitewater Basin, southeastern Indiana, and possible regional relationships: American Journal of Science, v. 271, p. 389–401.

Hancock, G.S., Anderson, R.S., and Whipple, K.X., 1998, Beyond power: Bedrock river incision process and form, in Tinkler, K.J., and Wohl, E.E., eds., Rivers over Rock: Fluvial Processes in Bedrock Channels: Ameri-can Geophysical Union Geophysical Monograph 107, p. 35–60.

Harbor, D., Erickson, P., Rittenhouse, D., Carlson, M., Gard-ner, T., Merritts, D., Dorale, J., and Panuska, B., 2000, Bringing down Floyd: Rapid erosion of the upper James River basin in response to scarp retreat and basin capture, in Habor, D., ed., Landscape Evolution in the Upper James River Basin: Lexington, Virginia, 2000 Southeast Friends of the Pleistocene Field Trip Guidebook, p. 31–47.

Hulver, M.L., 1997, Post-orogenic evolution of the Appala-chian mountain system and its foreland [Ph.D. disserta-tion]: Chicago, University of Chicago, 1055 p.

Leopold, L.B., Wolman, M.G., and Miller, J.P., 1992, Fluvial Processes in Geomorphology: San Francisco, Free-man, 522 p.

Matmon, A., Bierman, P.R., Larsen, J., Southworth, S., Pavich, M., Finkel, R., and Caffee, M., 2003, Ero-sion of an ancient mountain range, the Great Smoky Mountains, North Carolina and Tennessee: American Journal of Science, v. 303, p. 817–855, doi: 10.2475/ajs.303.9.817.

Miller, J.R., 1991, The infl uence of bedrock geology on knickpoint development and channel-bed degradation along downcutting streams in south-central Indiana: The Journal of Geology, v. 99, p. 591–605.

Miller, J.R., Ritter, D.F., and Kochel, R.C., 1990, Morpho-metric assessment of lithologic controls on drainage basin evolution in the Crawford Upland, south- central Indiana: American Journal of Science, v. 290, p. 569–599.

Mills, H.H., 2000, Apparent increasing rates of stream inci-sion in the eastern United States during the late Ceno-zoic: Geology, v. 28, p. 955–957, doi: 10.1130/0091-7613(2000)028<0955:AIROSI>2.3.CO;2.

Pazzaglia, F.J., and Gardner, T.W., 1994, Late Cenozoic fl exural deformation of the middle U.S. Atlantic pas-sive margin: Journal of Geophysical Research, v. 99, p. 12,143–12,157, doi: 10.1029/93JB03130.

Pazzaglia, F.J., Gardner, T.W., and Merritts, D.J., 1998, Bed-rock fl uvial incision and longitudinal profi le develop-ment over geologic time scales determined by fl uvial terraces, in Tinkler, K.J., and Wohl, E.E., eds., Rivers over Rocks: Fluvial Processes in Bedrock Channels: American Geophysical Union Geophysical Mono-graph 107, p. 207–235.

Poag, C.W., and Sevon, W.D., 1989, A record of Appala-chian denudation in postrift Mesozoic and Cenozoic sedimentary deposits of the U.S. middle Atlantic conti-nental margin: Geomorphology, v. 2, p. 119–157, doi: 10.1016/0169-555X(89)90009-3.

Reed, J.C., 1981, Disequilibrium profi le of the Potomac River near Washington D.C.—A result of lowered base level or Quaternary tectonics along the Fall Line?: Geology, v. 9, p. 445–450, doi: 10.1130/0091-7613(1981)9<445:DPOTPR>2.0.CO;2.

Reusser, L.J., Bierman, P.R., Pavich, M.J., Zen, E., Larsen, J., and Finkel, R.C., 2004, Rapid late Pleistocene inci-sion of Atlantic passive-margin river gorges: Science, v. 305, p. 499–502, doi: 10.1126/science.1097780.

Reusser, L.J., Bierman, P., Pavich, M., Larsen, J., and Fin-kel, R., 2006, Late Pleistocene bedrock channel inci-sion along the U.S. Atlantic passive margin, Holtwood Gorge, Susquehanna River, Pennsylvania: American Journal of Science, v. 306, p. 69–102, doi: 10.2475/ajs.306.2.69.

Schumm, S.A., 1993, River response to base level change; implications for sequence stratigraphy: The Journal of Geology, v. 101, p. 279–294.

Schumm, S.A., and Rea, D.K., 1995, Sediment yield from disturbed earth systems: Geology, v. 23, p. 391–394, doi: 10.1130/0091-7613(1995)023<0391:SYFDES>2.3.CO;2.

Schumm, S.A., Mosley, M.P., and Weaver, W.E., 1987, Experimental fl uvial geomorphology: New York, John Wiley & Sons, Inc., 413 p.

Sevon, W.D., 1989, Erosion in the Juniata River drainage basin, Pennsylvania: Geomorphology, v. 2, p. 303–318, doi: 10.1016/0169-555X(89)90017-2.

Springer, G.S., Kite, J.S., and Schmidt, V.A., 1997, Cave sedi-mentation, genesis, and erosional history in the Cheat River Canyon, West Virginia: Geological Society of America Bulletin, v. 109, p. 524–532, doi: 10.1130/0016-7606(1997)109<0524:CSGAEH>2.3.CO;2.

Tinkler, K.J., Penjolly, J.W., Parkins, W.G., and Asselin, G., 1994, Post glacial recession of Niagara Falls in relax-ation to the Great Lakes: Quaternary Research, v. 42, p. 20–29, doi: 10.1006/qres.1994.1050.

Ward, D.J., Spotila, J.A., Hancock, G.S., and Galbraith, J.M., 2005, New constraints on the late Cenozoic incision history of the New River, Virginia: Geomorphology, v. 72, p. 54–72, doi: 10.1016/j.geomorph.2005.05.002.

Whipple, K.X., 2001, Fluvial landscape response time: How plausible is steady-state denudation?: American Journal of Science, v. 301, p. 313–325, doi: 10.2475/ajs.301.4-5.313.

MANUSCRIPT RECEIVED 10 JANUARY 2006REVISED MANUSCRIPT RECEIVED 11 OCTOBER 2006MANUSCRIPT ACCEPTED 19 OCTOBER 2006

Printed in the USA

knickpoint evolution in a vertically bedded substrate ...€¦ · of individual waterfalls with up...

Documents