Journal of Geophysical Research: Earth Surface

Glacial erosion, evolution of river long profiles, and the organization of process domains in mountain drainage basins of coastal British Columbia



[1] In glaciated British Columbia, Canada, Quaternary climate changes are responsible for profound spatial reorganization of Earth surface processes. These changes have left a landscape characterized by topographic anisotropy associated with a hierarchy of glacial troughs. The evolution of glaciated landscapes is examined by analyzing the structure of geomorphic process domains and channel long profiles. To identify process domains we use channel surveys and GIS analysis to construct slope-area transects of the channel network. This analysis reveals generalized process-form disequilibrium with a mismatch between topographic signatures and currently active geomorphic process domains. At the landscape scale of “source” colluvial channels (contributing area <1 km2), the glacial/paraglacial signature commonly overrides that produced by contemporary debris flows. Along the axis of former ice flows, relict glacial cirques introduce a “hanging” fluvial domain at contributing areas as small as 8 × 10−2 km2 and produce complex channel long profiles similar to those observed for rivers responding to tectonic forcing. Slope-area relations typical of unglaciated equilibrium environments do not apply here. The concept of process domains appears to hold, however, some major glacially forced modifications in the alluvial-colluvial transition are observed and the definition of a depositional colluvial subdomain is proposed. Comparison between field- and GIS-measured slopes reveals that GIS-associated error is not uniform between process domains, and that GIS-based plots do not successfully discriminate field-based process domains. The combination of glacial and post-glacial fingerprints and the effects of ongoing Earth surface processes generate a complex landscape whose glacial signatures may persist until the onset of the next ice age.

1. Introduction

[2] Drainage basins are complex systems, owing to the variety of Earth surface processes at work, the complexity of their interactions at different spatial and temporal scales, and the confounding influences of tectonic activity and past episodes of climate change. They comprise unchanneled terrain (hillslopes) dominated by diffusive surface erosion and/or mass wasting, and stream channels where debris flows and fluvial processes prevail [e.g., Horton, 1945; Jackli, 1957].

[3] Process domains are defined as regions within which one or a collection of Earth surface processes prevails for the detachment and transport of mass. Plots of the logarithms of local slope gradient (S) vs. contributing drainage area (A) [e.g., Montgomery and Foufoula-Georgiou, 1993; Ijjasz-Vasquez and Bras, 1995; Tucker and Bras, 1998] can be used to delineate process domains (Figure 1). Slope and area represent first-order approximations to the physical conditions at which processes are active. In environments assumed to be at steady state, where topographic change is merely the product of contemporary Earth surface processes, slope and area have been used to identify process-specific topographic signatures. Steady state conditions require that the rate of local erosional lowering (E), limited by sediment transport capacity Qc (E = Qc = kSn Am), keeps pace with local rock uplift rate (U) while all other factors such as tectonic and climatic forcing are invariant, so that elevation does not change through time. In an idealized soil-mantled (transport-limited) equilibrium landscape, when local erosion and local uplift are balanced, local slope can be expressed as a power function of drainage area (Figure 1),

equation image

where K, m, and n are constants. The coefficient K expresses lithologic, climatic, and discharge variability; m and n are functions of active dominant processes [e.g., Montgomery, 2001].

Figure 1.

Schematic representation of process domains (dashed boundaries) and topographic signatures (solid lines) in slope-area space for an unglaciated mountain landscapes at steady state [after Montgomery and Foufoula-Georgiou, 1993].

[4] Domains for unglaciated equilibrium landscapes are: (1) hillslope, (2) colluvial (or debris flow), and (3) fluvial. The boundaries of these domains are marked by kinks in slope-area plots and by transitional geomorphic features in the field: (1) channel heads, that separate unchanneled and channeled portions of the topography, and (2) debris flow fans, which form at the transition between colluvial channels (in the sense of Montgomery and Buffington [1997]) and fluvial reaches, where contemporary fluvial terraces form [e.g., Montgomery and Foufoula-Georgiou, 1993; Sklar and Dietrich, 1998; Stock and Dietrich, 2003] (Figure 1).

[5] Research in long-term landscape evolution has made substantial use of slope-area plots, particularly for modeling fluvial erosion in bedrock channels [e.g., Seidl and Dietrich, 1992; Sklar and Dietrich, 1998; Whipple, 2001] as well as for assessing relevant controls exerted by lithology [Duvall et al., 2004], tectonic activity [e.g., Whipple et al., 1999; Snyder et al., 2000; Kirby and Whipple, 2001; Schoenbohm et al., 2004], and vegetation [e.g., Istanbulluoglu and Bras, 2005]. In such models of fluvial erosion, slope can be empirically expressed as a function of drainage area,

equation image

where ks is the channel steepness index, and θ is the channel concavity index. It has been suggested that ks is dependent and θ is independent of tectonic uplift rates [e.g., Sklar and Dietrich, 1998; Whipple and Tucker, 1999]. Note that in a detachment-limited (bedrock-covered) landscape equation (1) becomes S = (U/K)1/nAm/n, where m/n corresponds to θ only when the study area is at steady state and local tectonic uplift, climate, and lithology are uniform in space.

[6] Process domains, their boundaries (i.e., channel heads and debris flow fans), and channel long profiles are dynamic entities that respond to variations in hydro-climatic regime [e.g., Ryder, 1971; Dunne, 1991; Montgomery, 1999], yet limited quantitative research has been conducted on the topic [e.g., Tucker and Slingerland, 1997; Whipple et al., 1999]. Recent studies of transient conditions due to the regional Quaternary pattern of uplift and climate change in unglaciated mountain systems [e.g., Schoenbohm et al., 2004] report slope-area relations with multiple kinks and discontinuities that are in clear contrast with systems in steady state. However, unglaciated terrains, in or out of equilibrium, do not represent the full range of conditions that one can observe in mountain environments. In fact, a number of orogens are currently glacierized or have been glaciated in the Quaternary but, because of their intrinsic complexity, these environments have received less attention and therefore merit closer study [e.g., Hooke, 1991; Whipple et al., 1999; MacGregor et al., 2000; Montgomery, 2002; Tomkin and Braun, 2002]. Glaciated environments are transient and contain “fingerprints” of glacial erosion that are progressively erased by post-glacial processes.

[7] The objectives of this paper are to evaluate the effects of glacial history on the organization of currently active geomorphic process domains and topographic signatures, to determine how these compare to previous conceptual models proposed for unglaciated analogues; and on the structure of channel long profiles. In addition, we aim to evaluate whether or not our GIS-based measurements of slope are sufficient and suitable to address research questions about landscape evolution and watershed management.

[8] To pursue these goals we determine slope-area relationships for five basins in coastal British Columbia using equation (2) and we infer values of ks and θ through linear regression of the model on field and GIS measurements of their topographic relief. To address the question of GIS-data suitability, we use field-measured slopes to evaluate the error associated with GIS-measured slopes and assess to what degree this affects the discrimination of process domains.

2. Study Area

[9] The physical environment of coastal British Columbia is the result of a series of geo-climatic processes, which in chronological order include diastrophism, glaciations, and paraglacial sedimentary dynamics [Ryder, 1981; Muhs et al., 1987]. Accordingly, the landscape contains elements of three phases of evolution. First-order landforms resulted from the orogeny that took place during the Tertiary period. These features have undergone repeated episodes of profound glacially induced modifications in the Quaternary period leading to second order landforms. Finally, third-order features developed as local modifications of the glacially derived landforms, through accelerated mass wasting processes on valley sides and downstream accumulation of fluvial sediment on valley floors, during deglaciation at the end of the Pleistocene [Ryder, 1981] about 14,000 years ago [Kovanen and Easterbrook, 2002].

[10] The Quaternary legacy still dominates the morphology and affects the sediment dynamics of the landscape today [e.g., Church and Ryder, 1972; Church and Slaymaker, 1989]. In terms of morphology, ice flows have introduced a strong topographic anisotropy and created a variety of features at different landscape scales including glacial cirques, hanging valleys, valley steps, and oversteepened valley walls. As a result, geomorphic processes have acquired a characteristic spatial organization: Fluvial processes preferentially dominate along the longitudinal axis of relict glacial troughs; on the valley walls, transverse to this axis colluvial processes dominate [e.g., Ryder, 1981].

[11] All study basins are located on homogeneous intrusive lithologies, largely granodiorite and quartz diorite [Roddick, 1965; Muller et al., 1974; Monger, 1989]. The study basins are Lembke Creek, Hesketh Creek and East Cap Creek, tributaries of the Capilano River in the Pacific Coast Ranges; Elliott Creek in the Northern Insular Ranges; and Radium Creek in the Northern Cascades (Figure 2). These mountain drainage basins, for which a conspicuous amount of background geomorphic information is available, represent a considerable part of the physiographic variability of southwestern British Columbia [Holland, 1964].

Figure 2.

Map of the Capilano River basin indicating the location of the ground-surveyed streams. The inset shows locations of Elliott Creek and Radium Creek.

3. Data Collection

[12] The methodologies we adopted include air photo interpretation (API), GIS-based topographic analyses, and field surveys. API yielded a preliminary delineation of the downstream boundary of the “debris-flow domain” via identification of debris-flow fans, fluvial terraces, and evaluation of geomorphic coupling from hillslope sediment inputs. The term coupling is used in geomorphology [e.g., Brunsden and Thornes, 1979; Caine and Swanson, 1989] to indicate the degree of connectivity between hillslope and fluvial processes. Specifically, coupled systems exhibit direct colluvial-alluvial interaction, as opposed to decoupled (or buffered) systems, where colluvial sediment inputs do not reach the channel network. Evaluation of degree of coupling is fundamental to drainage basin sediment dynamics as it controls (1) sediment and process-disturbance cascades, and (2) in what proportion hillslope denudation rate contributes to drainage basin sediment storage and fluvial sediment yield respectively [e.g., Roberts and Church, 1986; Reid and Dunne, 1996].

[13] In consideration of the topographic anisotropy introduced by glaciers and ice sheets, we conducted slope-area transects along relict glacial troughs (Figure 2, white lines) and along their adjoining valley walls (Figure 2, black lines). For the former component we seek to document the effects of valley modification by glacial activity; for the latter we aim to verify the existence of debris flow-induced topographic signatures, particularly to assess the state of landscape recovery from glacial disturbance. Collectively, the longitudinal and transverse transects present a picture of glaciated terrain structures and process domain arrangements.

[14] GIS-based analysis was used to derive and compare area-slope plots of two kinds for main-trunk streams: (1) automatically extracted from a 25-m gridded DEM and (2) hand-digitized along the channel network from 1:20,000 digital topographic maps. In the DEM-based analysis standard procedures were used [e.g., Montgomery and Foufoula-Georgiou, 1993; Ijjasz-Vasquez and Bras, 1995], slope gradient was computed using the steepest descent algorithm and contributing area by the D8 single-flow accumulation algorithm [O'Callaghan and Mark, 1984; Garbrecht and Martz, 1997]. For the contour-based map analysis, we replicated the procedure adopted by Stock and Dietrich [2003]. This second procedure was employed to eliminate algorithm-related artifacts we observed in the drainage network extracted from DEM grids. The D8 flow algorithm is certainly not the most appropriate solution for modeling flow-divergent surfaces (hillslopes) where instead multiple-flow algorithms mimic real hydrologic conditions more closely. To test this limitation, D8-derived contributing areas were compared with values obtained by the multiple outflow algorithm D-inf [Tarboton, 1997] at a number of GPS-recorded locations (n = 88) with drainage area ranging between 3.15 × 10−3 km2 and 1.7 × 102 km2. Results reveal that the discrepancy between D-inf and D8 lies within the uncertainty associated with DEM resolution.

[15] Field surveys involved mapping as well as measurement and description of morphologic features (e.g., channel heads, debris-flow fans, landslide tracks, and strath terraces) that were identified or misinterpreted during API. Depending on site accessibility, slope gradients were measured by means of a hand-held clinometer or laser theodolite. In order to ensure appropriate comparison with GIS-measured slopes, field measurements were taken at a length scale equal to the local channel width (up to a maximum of 25 m). Field measurements are considered the true values from which we calculate the error associated with the API and GIS slope estimates (see section 4.5). In order to overlay and compare the different types of measurements, positioning information was gathered via GPS devices and orthorectified air photos.

4. Results

[16] We ordered our presentation of results by system size (drainage area), starting from streams located on valley walls, progressing to incipient hanging valleys, and ending with fully developed glacial troughs. As will become clear, system complexity increases with system size. In slope-area plots this is matched by increasing complexity of the graphs. For the simplest cases a single power law is adequate. For the most complex cases the slope-area relationships are multisegmented. We then consider how well geomorphic process domains are resolved in the slope-area space by plotting all field transects together, and finally analyze the error associated with GIS-measured slope gradients.

4.1. Valley Walls and Cirque Walls

[17] Slope-area transects of unglaciated debris-flow (colluvial) dominated channels are typically fit by a single power law (e.g., Stock and Dietrich [2003] and Figure 1). For debris-flow dominated channels in formerly glaciated landscapes the situation is more complex. Some of the study slope-area transects are fit by a single power law relation (Figures 3a, 3b, and 3g), while others, even though debris-flow dominated for their entire length, require a double power law fitting (Figures 3c–3f and Table 1). One could attribute the slope break of the latter plots to progressive loss of erosive power as the debris flow travels downstream and fluvial processes become more important. However, the kink observed in most slope-area plots of the study debris-flow channels casts doubt on this interpretation. Specifically, the kink occurs within a broad range of slope gradients, between 0.48 and 1.01, values that are undoubtedly too high (1) to justify a debris flow loss of power, and (2) to allow formation of fluvial terraces (diagnostic landforms of dominant fluvial activity). For reference, Stock and Dietrich [2003] reported kinks occurring at slopes between 0.04 and 0.40 for unglaciated granitic basins and interpreted trend differences, such as the presence and/or location of the kink and variation in the shape of the curvature, in terms of system size (drainage basin area). In particular, they recommended using sufficiently large basins to define fluvial trends, hence deviations due to debris flow scouring. In our work, single versus double power law trends cannot be explained as an effect of system size because basins of comparable drainage area display different trends. For example, single power law relations yield excellent fits for Rapid Creek, Hesketh01 and Lembke01, whereas double power law relations are required for Hesketh100, Slide Creek, and Lembke03 (Table 1).

Table 1. Field Data for the Transverse Tarnsects
Stream ChannelBasin Area, km2Surficial MaterialsaKinkθks
Slope, m/mArea, km2
  • a

    M, morainal till; C, colluvium; R, bedrock; GF, glaciofluvial; F, fluvial; . components are approximately in equal proportions; /, the left hand side component is more extensive; /CM, till partially covered by colluvium.

  • b

    Truncated fan.

Slide lower R.C0.30n.a.0.7272 × 103
Lembke020.062R0.50–0.770.036–0.039− 0.1220.2
Lembke02 lower  C.M0.16n.a.0.49592.7
Lembke03 lower C0.23n.a.1.7802 × 108
Hesketh02 0.100Rn.a.n.a.0.42355.4
Hesketh100 lower GF/M0.23n.a.1.8895 × 109
East Cap1000.068R/C0.72–0.850.019–0.029− 0.2480.1
East Cap100 lower R0.42n.a.1.0524 × 104
East Cap1200.057R0.74–1.010.010–0.022− 0.1790.2
East Cap120 lower C0.27n.a.1.0713.5 × 104
Rapid lower C/R0.27n.a.2.3511 × 1013
Sisters02 mid C/R0.31–0.401.203–1.461− 1.6792 × 10−11
Sisters02 lower /CMn.a.n.a.3.8341 × 1023
Sisters Pass1.010C.R0.320.092–0.1340.596325.9
Sisters Pass mid M/C0.520.291–0.403−0.8659 × 10−6
Sisters Pass lower Fn.a.n.a.1.3342 × 107

[18] A careful look at the geomorphology of the study basins suggests that the shape of the slope-area relations is imposed by glacially induced local topography rather than controlled by differential erosive power associated with dominance of colluvial and fluvial processes. In this context, morphometric parameters, namely valley bottom width, valley wall profile curvature and steepness, and the way these are connected through the presence or absence of paraglacial fans (see Ryder [1971] for details) and cones seem to play a prominent role. Applying this thinking, we can explain the slope-area plots for all the debris-flow dominated channels. Strachan01, Strachan02 (Figure 3a), Lembke04, and Sisters01 Creeks (Figure 3b) flow down steep, till-blanketed, planar valley walls, which form non-buffered V-shaped valleys: this explains why they exhibit single power law relations and possess low concavity indices (0.089 < θ < 0.301, Table 1). In contrast, the higher concavity indices for Lembke01, Hesketh01 and Hesketh02 Creeks (0.380 < θ < 0.546) are due to more concave valley walls which form buffered U-shaped hanging valleys.

Figure 3.

Slope-area plots (a–g) along debris-flow dominated channels and (h) at a glacially induced saddle. Numbers indicate concavity indices defined between the dashed lines using field-measured slopes (see Table 1).

[19] Turning to systems that display a double power law relation, the kinks in the Slide Creek graph (Figure 3c) and Hesketh100 graph (Figure 3d) correspond to the apices of paraglacial fans. For Lembke02, Lembke03, EastCap100 (Figure 3e), and EastCap120 (Figure 3f) the kinks correspond to a transition from the bedrock/colluvium-dominated upper slopes to the till-blanketed base slopes. Rapid Creek (Figure 3g) is an extreme example of the sharp transition from a rocky planar valley wall (θ = 0.288) to a paraglacial fan (θ = 2.351) that protrudes into the Capilano River valley. Most of the plot fits a single power law, but to fit the right limb we require the largest θ value of any in our study.

4.2. Glacially Induced Saddle

[20] The glacially induced saddle of Sisters Pass (Figure 3h) illustrates the breakdown of the foregoing power law characterization. Sisters Pass and Sisters02 Creeks (Figure 2, dashed lines) illustrate an intermediate case between valley walls and glacial troughs. Both stream channels originate from a glacially carved saddle (Sisters Pass), then flow into an incipient hanging valley before intersecting a major glacial trough (Lembke Creek and Capilano River). Ice flow through Sisters Pass has created a higher degree of plan view curvature on both sides of the saddle compared to that of adjacent streams (Hesketh100 and Rapid Creek). Sisters Pass and Sisters02 Creeks possess morphologies that are characteristic of intermediate stages of valley development by glacial activity; hanging valleys are not fully developed and channel main stems are not completely decoupled from hillslope sediment inputs nor longitudinally scoured by debris flows. In slope-area terms, this translates into plots having two distinct kinks of opposite sign, which bound a portion of stream channel characterized by negative concavity index (−1.679 < θ < −0.865, Table 1; positive slope-area relation). For unglaciated basins such a trend has been defined at very small drainage areas as diagnostic of unchanneled topography, where hillslope diffusive styles of sediment transport are dominant (Figure 1) [e.g., Montgomery and Foufoula-Georgiou, 1993; Tucker and Bras, 1998]. For our study of glaciated basins, transient topography clearly alters the slope-area relations that are accepted for unglaciated environments at steady state.

4.3. Glacial Troughs

[21] The terrain structure along glacial troughs presents much more composite configurations than observed in valley walls and as such it requires a more detailed description. Hesketh Creek originates from an unchanneled saddle (H in Figure 4a) followed by a poorly defined colluvial channel (C1, θ = 0.817) which degrades into a spoon-shaped hanging valley (HF1, θ = n.a., not available, owing to the presence of a kink in the slope-area relation within the domain boundaries) characterized by a poorly defined channel. The hanging valley terminates abruptly into a steep and incised colluvial channel (C2, θ = 0.885) dominated by large boulders commonly interlocked in massive jammed structures. Here material is transferred preferentially via debris flows into the principal decoupled hanging valley (HF2, θ = 3.892), which presents typical riffle-pool morphology. Downstream, the channel reacquires a colluvial and confined character (C3, θ = n.a) with step-pool and boulder-cascade morphologies, sediment being supplied by near-bank failures. Gentler slope gradients (F, θ = 4.607) and riffle-pool morphology reappear in the distal reaches where Hesketh Creek intersects the Capilano glacial trough (Figure 2).

Figure 4.

Slope-area plots along the direction of formerly active ice flow. B, bedrock canyon; C, colluvial domain, typically having chaotic, cascade or step-pool morphology; F and HF, fluvial and hanging fluvial domains, commonly with rapids or riffle-pool morphology; and H, hillslope domain. Numbers indicate field-based concavity indices (see Table 2). (a) Hesketh Creek is the prototype for the description of the other basins (see text for explanation). (b) Elliott Creek (C1 = 0, HF1 = 0.623, C2 = −1.479, HF2 = n.a., C3 = n.a., F = 41.429) exhibits the same domain sequence as Hesketh Creek. (c) Lembke Creek has a simple structure with a hanging valley (HF = n.a.) that separates two colluvial channels (C1 = 0.277, C2 = 0.303), and no distal fluvial domain. (d) East Cap Creek originates as a colluvial channel (C1 = 0.136) that flows into a hanging valley (HF1 = n.a.), hence through a relict meltwater channel (bedrock canyon, B = 1.368). Downstream, the primary hanging valley is replaced by a glacial trough (C2 = 1.368 and F = 0.086). (e) Radium Creek exhibits the same domain sequence as Elliott Creek (C1 = 0.346, HF1 = n.a., C2 = 4.732, HF2 = 0.835, C3 = n.a., F = n.a.).

[22] This sequence of “building blocks” (hillslope-colluvial-fluvial-colluvial-fluvial-colluvial-fluvial) applies with some variations to the structure of all other basins studied (see Figures 4b–4e). Owing to the formation of cirques and hanging valleys, transient situations such as those observed in Sisters Pass and Sisters02 Creeks are here developed to a further stage, with increased relative relief, longer valley walls, and wider valley floors. Longitudinal profiles exhibit a stepped topography which, in slope-area plots, yields multiple sequences of kinks or discontinuities, extreme positive limbs (θ ≪ 0), lower minima for hanging valleys and, in some cases extremely high negative slope coefficients for confined colluvial channels (θ > 3.9). During glaciation transient topography likely evolved to a stable “glacial equilibrium state” that produced fully developed glacial troughs that today are associated with drastically decreased channel gradient variability (e.g., zone F, Figure 4d). Along relict glacial troughs, the contemporary riverine valley floor, which we term distal fluvial subdomain, is permanently decoupled from hillslope sediment inputs and differs from the hanging fluvial subdomain, typically associated with relict glacial cirques. Similarly to what is seen for transverse transects, the presence of kinks within single process domains (e.g., zone C2 in Figure 4a, and zones HF2 and C3 in Figure 4b, Table 2) requires the use of double power law fitting.

Table 2. Field Data for the Longitudinal Transects
Stream ChannelBasin Area, km2Process DomainLandscape Macro-FormDominant Channel MorphologyDomain Upstream Limitθks
Slope, m/mArea, km2
  • a

    Presence of a kink within the domain that prevents from using a single power law fit.

  • b

    Not shown in Figure 4.

  • c

    GIS-measured data (digitized).

Hesketh5.5C1cirque wallchaotic0.5940.0160.8174 × 103
HF1hanging valleyrapids0.1140.154n.a.n.a.
C2valley stepcascades; step pools0.6490.2310.8854 × 104
HF2hanging valleyriffle pools; rapids0.1900.6043.8929 × 1022
C3valley stepcascades; step
Fglacial troughriffle pools; rapids0.0695.3004.6076 × 1029
Elliott10.4C1open saddlestep pools0.3640.012−0.0050.2
HF1hanging valleyrapids; riffle pools0.0950.8050.623589.9
C2valley stepstep pools; bedrock0.2212.027−1.4796 × 10−11
HF2hanging valleyriffle
C3valley stepcascades; step
Fglacial troughstep pools; rapids0.13810.16241.4295 × 10289
Lembke22.3C1cirque wallchaotic; bedrock0.7830.0030.2738.7
HFhanging valleyriffle pools;
C2valley step/troughstep pools; cascades0.1491.0830.2333.6
East Cap40.7C1cirque wallchaotic0.6700.0040.1361.9
HFhanging valleyrapids;
Brelict meltwater channelbedrock0.3640.3881.3683 × 107
C2glacial troughchaotic; step pools0.1261.0221.3683 × 107
Fglacial troughrapids; riffle pools0.0601.9890.0860.237
EastCap 110b1.3C1colluvial fanchaotic; step pools0.5200.1030.7624 × 103
HFhanging valleyriffle
C2valley stepstep pools; cascades0.1291.043−3.4114× 10−22
Fglacial troughrapids0.0761.1679.6915 × 1057
Radiumc10.1C1cirque walln.a.c0.8570.0110.34625.59
HF1hanging valley
C2valley step 0.5631.3074.7324 × 1028
HF2hanging valley 0.1562.0860.8352 × 103
C3glacial trough
Fglacial trough

4.4. Process Domains

[23] By plotting together all field-based transects of our study basins (Figure 5a) we seek to illustrate how well geomorphic process domains are resolved in slope-area space, and therefore gain a better understanding of the causal linkages connecting processes active in different morphological units of the glaciated landscape. Specifically, this should allow testing of Montgomery and Foufoula-Georgiou's [1993] idealized conceptual model (Figure 1), originally proposed for unglaciated environments on the basis of DEM-derived data. Results are encouraging: hillslope, colluvial and fluvial domains exhibit good separation. The two alluvial subdomains (hanging and distal) overlap substantially, owing to the topographic anisotropy created by formerly active ice flows. Accordingly, the landscape scale (drainage area) at which slope gradients characteristic of purely alluvial reaches occur is controlled by the length of cirque walls for hanging fluvial channels and that of valley walls for distal fluvial channels, which appear to be of comparable size. However, the position of the colluvial domain in the slope-area space deviates significantly from what has been proposed for unglaciated environments [Montgomery and Foufoula-Georgiou, 1993; Sklar and Dietrich, 1998]. In particular, the transition from colluvial to alluvial channels occurs at a systematically decreasing slope beyond drainage areas larger than 1 km2 (Figure 5a). Interestingly, 1 km2 is the characteristic spatial scale of valley step and glacial trough initiation (see contributing area values of C2 zones in Figure 4 and Table 2), which might imply some sort of dependence on valley morphometry. Examination of the spatial distribution and of the magnitude-frequency relations of sediment sources in the Capilano River basin [Brardinoni et al., 2003; Brardinoni and Church, 2004] confirm that the typically glacial, trellis-like structure of the channel network and stepped nature of the topography affect debris flow trajectories and run-out distances. Consequently, we decided to classify our study colluvial channels into two subcategories, depending on how these are spatially connected to debris flows: source and sink colluvial channels (Figure 6). Source colluvial channels are very steep, first- and second-order streams, typically located on valley walls and cirque walls. Distinctively, they are longitudinally scoured by debris flows and may receive colluvial lateral inputs (e.g., debris slides and avalanches) from open-slope locations. Sink colluvial channels are moderately steep second- and higher-order streams, typically flowing along valley steps and glacial troughs whose floor is not wide enough to prevent lateral colluvial inputs from impacting the stream main stem. They receive exclusively lateral colluvial inputs (as opposed to colluvial inputs from upstream), including debris flows at tributary junctions.

Figure 5.

(a) Process domains based on field-measured slopes plotted in slope-area space. (b) Process domains based on DEM-measured slopes. Solid lines are the schematic boundaries of process domains for unglaciated drainage basins sketched by Montgomery and Foufoula-Georgiou [1993]. Dashed line is the “glacially forced” colluvial-alluvial transition detected in this study.

Figure 6.

Contour map of Hesketh Creek and Lembke Creek showing the spatial configuration of sediment sources, colluvial channels and alluvial channels. Open circles indicate initiation points of sediment sources (debris slides and debris flows) identified via API within a 30-year time window. Sediment sources were mapped during extensive field surveys conducted to establish the reliability of API-based landslide inventories (for details, see Brardinoni et al. [2003]).

[24] This source-sink distinction reveals that the inflection point of the colluvial-fluvial transition corresponds to a shift in colluvial subcategory. The source colluvial-fluvial transition occurs at a roughly constant slope of 0.2 for all drainage areas, a value typical of debris flow fans in glaciated British Columbia [e.g., Hungr et al., 1984; Van Dine, 1985; Fannin and Rollerson, 1993]. The sink colluvial-fluvial transition decreases with increasing drainage area and is associated with channels flowing along coupled glacial troughs. The declining trend of the transition (its gradient, onset, and length) appears to be forced by the glacially induced topography. Along relict glacial troughs, as contributing area increases so does relative relief and length of valley walls. As a result, increasingly large debris flows intersect stream channels which possess gradients otherwise typical of alluvial environments. This spatial configuration of processes explains the deposition of a colluvial blanket at channel gradients as low as 0.06 and a drainage area at least as large as 20 km2. We expect the colluvial-alluvial boundary to stop declining (and potentially reverse its trend) at a scale where glacial troughs are wide enough to decouple lateral colluvial inputs and/or valley walls become considerably less prone to the initiation of shallow rapid failures.

4.5. Field- Versus GIS-Measured Data

[25] Typically, GIS-based calculation of local slope gradient for the purpose of analyzing channel long profiles and process domains has relied upon 90 m [e.g., Schoenbohm et al., 2004], 30 m [e.g., Wolinsky and Pratson, 2005], or 10 m [e.g., Montgomery, 2001] DEMs, rarely on high-resolution (meter-scale) topography (e.g., some catchments in the work of Stock and Dietrich [2003]). Our extensive field surveys allow us to use field-measured slopes to (1) evaluate the error associated with GIS-measured slopes and (2) assess the reliability of an “automated” DEM-based delineation of geomorphic process domains/signatures.

[26] Comparison of field-measured slopes with those calculated using GIS data (contour-based and gridded DEM) exhibits a high degree of scatter (Figures 7a and 7b). GIS-measured slopes nearly all plot within ±100% error; exceptions are field slopes smaller than 0.27, which in places are overestimated by more than 100%. Furthermore, slope discrepancies seem to vary between process domains. Respectively, GIS slopes underestimate field slopes on actively eroding, steep, highly dissected terrain (i.e., source colluvial channels) and overestimate field slopes on dominantly depositional environments characterized by a less rugged topography (i.e., hanging valleys and sink colluvial channels). These trends are visually confirmed by the box plots of Figure 8. Process domains may be ranked in terms of increasing root-mean square (RMS) residual error of GIS-measured slopes relatively to field-measured slopes as follows (Table 3): distal fluvial < sink colluvial < hanging fluvial < source colluvial.

Figure 7.

Scatterplots comparing (a) field slopes versus DEM slopes and (b) field slopes versus digitized slopes.

Figure 8.

Box plots of field slopes, digitized slopes, and DEM slopes across channel types.

Table 3. Error Analysis of GIS-Measured Slopes Performed Against Field-Measured Slopes
 RMS Residual ErrorStatistical Significance
  • a

    Here ns denotes not significant at the 0.05 level.

Bedrock canyon0.1050.074nsansnsns
Colluvial (source)0.1140.127nsnsnsns
Colluvial (sink)0.0560.0560.0020.0090.0010.017
Distal fluvial0.0170.012nsnsnsns
Hanging fluvial0.1140.1130.0910.0680.0590.085

[27] In addition, we evaluated treatment effect (field, DEM, and digitized) within process domains. Accordingly, one-way ANOVA followed by Bonferroni post-hoc tests shows that along sink colluvial channels the mean of field-measured slopes is significantly smaller than those calculated with GIS (Table 3). Since slopes are not normally distributed within process domains we also performed Kruskal-Wallis tests. These tests yield virtually identical results to those of the Bonferroni procedure.

[28] Figure 5b shows that DEM data are markedly inferior for delineating process domains in the slope-area space. DEM-based domains are not well resolved. Specifically, while sink and source colluvial stay well separated, the source colluvial lower boundary is shifted up to a gradient value greater than 0.3. More importantly, fluvial and hanging fluvial reaches tend to plot within the sink-colluvial domain.

[29] Overall, we think that a semi-automated discrimination of process domains and topographic signatures is more appropriate than a fully automated one. This is because: (1) inflection points in slope-area relations do not always correspond to a transition of process dominance, (2) bias (underestimation or overestimation) is not uniform between different process domains, (3) some complex features located at relatively small drainage areas (i.e., HF in East Cap Creek) may not be captured by a 25-m DEM, and (4) process domains are not well discriminated in the slope-area space (Figure 5b). The semiautomated assessment procedure would entail coupling DEM-based slope-area plots to some API analysis.

5. Discussion

[30] Debris-flow dominated channels, which typically initiate as steep, rectilinear streams, degrade to gentler gradients toward the base of valley walls, where till-mantled slopes and paraglacial fans and cones provide sedimentary linkages between valley walls and valley bottoms. In process-related terms these have been zones of transition between colluvial- and fluvial-dominated environments in the early Holocene epoch. Today, active debris flow fans in the area are typically located at the terminus of paraglacial fans.

[31] In slope-area terms, the presence of paraglacial fans imposes a kink in the relation, which at this landscape scale is otherwise controlled by the valley wall and bottom bedrock geometries carved by past ice flows. We conclude that in debris-flow dominated channels of coastal British Columbia the glacial/paraglacial signature commonly overrides that produced by contemporary debris flows. Operationally, this result complicates the identification of the debris flow transition zone (the kink does not always correspond to a change in process dominance) from area-slope plots, hence from automated GIS analysis (see section 4.5).

[32] Along the direction of formerly active ice flows, hanging valleys that originated from relict cirques are the most striking features. Their peculiar morphology imposes low channel gradients and isolates sediment inputs delivered from bordering cirque walls. As a result, in steep glaciated mountain environments relict glacial cirques enclose distinctive hanging fluvial domains. In contrast to steep unglaciated drainage basins, where the degree of coupling progressively weakens downstream until colluvial reaches grade to fluvial analogues, decoupled hanging valleys separate strongly coupled channel reaches that together produce stepped long profiles. Therefore the glacially imposed variations in valley long profiles impart substantial variability to process domain sequencing in individual watersheds (Figure 4).

[33] In terms of slope-area relations this translates into fragmented, “saw-tooth” patterns with hanging valleys and steep colluvial channels corresponding to slope relative minima and maxima, respectively (Figure 9b). The well-established slope-area relations for unglaciated environments do not apply in glaciated environments. In these environments, slope-area relations have the following features: (1) segments with positive slope-area relations occur at the transition between the downstream end of hanging valleys (or saddles) and the inception of colluvial reaches, and (2) high negative concavity indices are associated with colluvial channels. Positive slope-area relations are also observed in unglaciated orogens [e.g., Snyder et al., 2000] and have been explained as a transient response to tectonic forcing [e.g., Schoenbohm et al., 2004].

Figure 9.

Schematic representation of (a) longitudinal profile and (b) correspondent area-based process domains and topographic signatures (solid lines) in a glaciated study basin of coastal British Columbia. Degree of coupling and relevant glacial macro-forms are reported in brackets. Arrows indicate apex of paraglacial fans. The presence of a relict glacial cirque imposes a “hanging fluvial” domain between colluvial domains, which translates into a multifragmented slope-area sequence. Note that kinks (reversal and inflection points) do not necessarily correspond to change in process dominance.

[34] In strictly fluvial geomorphology terms, the inherited Quaternary geomorphology imposes local channel gradient and degree of coupling (Figure 9a), which in turn at the channel reach scale control hydraulics, unit stream power, channel morphology and in-channel habitat characteristics. At the watershed scale this influences hydrologic response, downstream variation of stream power, as well as material transfer dynamics (e.g., wood and sediment cascades) and the transmission of impulses of any kind along the system (see “transmission resistance” in the work by Brunsden and Thornes [1979]).

[35] In summary, the glacially inherited topography induces a peculiar degree of geomorphic coupling (connectivity) between hillslope and channel processes. This in turn has a major impact on the delineation of process domains in slope-area plots by generating a transitional process domain (sink colluvial) whose boundary with alluvial environments is an inverse function of local slope and contributing area. We expect to see different colluvial-alluvial transition trends in different glaciated landscapes, depending on pre-glacial litho-topographic properties and style of glaciation.

[36] In watershed analysis, GIS is an extarordinary valuable tool for quantifying topographic spatial variability and for integrating field and remotely sensed data. From a watershed management standpoint, assessing how well topographic signatures and process domains may be captured solely from DEM-extraction is very critical. In slope-area plots of transverse transects (Figure 3) GIS-derived slopes (DEM and digitized) exhibit considerably higher scatter than field data, making the identification of inflection points more difficult. Along longitudinal transects (Figure 4) inflections and reversals observable in field-based slope-area relations are reproduced reasonably well by those extracted with GIS. The hanging valley at East Cap Creek headwaters is an exception and is not appropriately captured. We attribute such a discrepancy to the presence of two (about 20 m long) relict meltwater channels and their associated morphological jumps.

6. Conclusions

[37] Our results portray a more complex geomorphic picture than that described for unglaciated catchments [e.g., Montgomery and Foufoula-Georgiou, 1993; Stock and Dietrich, 2003]. By considering the inherited effects of glacial geomorphic processes we extend previous results and characterize process domains and topographic signatures in glaciated mountain environments. Specifically, Quaternary climate changes in glaciated British Columbia have left a landscape where process-specific topographic signatures rarely match the domains of currently active geomorphic process.

[38] Transverse and longitudinal slope-area transects indicate widespread process-form disequilibrium resulting from the superimposition of ongoing Earth surface processes on a glacial palimpsest, so that glacial topographic signatures override those produced by contemporary geomorphic processes. Particularly relevant is the finding that longitudinal profiles of channels periodically scoured by debris flows (which we define as source colluvial domains) remain largely controlled by the inherited glacial topography (Figure 3). Along the direction of formerly active ice flows, relict glacial cirques and hanging valleys enclose typical hanging fluvial domains and characterize the morphological structure of the landscape (Figure 4). Likely, these will persist as first-order landforms until the end of this interglacial period, if any glacial period is to occur. In general terms, this study shows a clear causal linkage across spatial scales. Glacial macro-forms (coarse scale level) drive the active processes that dominate mass transport at the finer local level, such as the channel reach scale (see Figures 9a and 9b in conjunction with Table 2).

[39] Although topographic signatures are imposed by glacial macro-forms the concept of process domains appears to apply. However, an important deviation from the unglaciated model [Montgomery and Foufoula-Georgiou, 1993; Sklar and Dietrich, 1998] is observed: the transition between colluvial and alluvial channels, which occurs at a constant slope of about 0.2 for small drainage areas, starts declining systematically for contributing areas larger than about 1 km2. This trend is explained by the glacially imposed degree of coupling (sink colluvial channels), which is controlled by the morphometric characteristics of glacial troughs and adjoining valley walls. Therefore the spatial distribution of process domains on a slope-area plot may be to some degree similar in landscapes with different history (i.e., unglaciated versus unglaciated) but where the topography plots within these process domains varies between landscape types (see Figures 1 and 9b).

[40] From a watershed management perspective, assessing the ability to identify process domains from GIS-extracted slope-area plots is critical for predicting patterns of natural and anthropogenic disturbances, as well as in-stream habitat conditions. The 25-m DEM currently available for glaciated coastal British Columbia does not ensure an automated discrimination of process domains (see Figures 5a and 5b) and signatures (Figure 4d) that we can detect in the field.

[41] After about 14 ka since the last ice retreat occurred locally, the landscape we have studied has not recovered significantly from past glaciations and is out of balance with prevailing conditions. Given how little modification of the glacial signature has occurred in the 14 ka since deglaciation, and the typical duration of interglacial periods (10–50 ka [e.g., Berger and Loutre, 2002]), glacial signatures will likely persist until the onset of the next glacial period.


[42] We thank Garry Clarke, who provided critical comments on an earlier draft of the paper, Robert Anderson, Raphael Bras, Alex Densmore, and Kelly MacGregor for thoughtful reviews. Discussions with Mike Church, David Montgomery, Olav Slaymaker, and Kelin Whipple helped organizing the presentation of the results. Sally Hermansen and Ian Bryor kindly proofread the paper. Jason Rempel helped during fieldwork. Eric Leinberger prepared all figures. The project was funded through University of British Columbia Graduate Fellowships awarded to F. B., and a NSERC grant to M. H.