Seismic surveys

We attempted seismic reflection and refraction soundings using a 12-channel Bison^® digital seismograph with geophone spacing of 7–15 m. For sources we used 300- to 500-grain blank shotgun shells discharged within the debris mantle just below the surface. Geophones were placed in small pits and sandwiched between flat rocks in an attempt to reduce noise. Shots were made at various positions along and off the ends of transverse and longitudinal arrays.

No obvious reflections were identified, probably because of poor source coupling and a low signal-to-noise ratio. We suspect that larger explosive sources may have penetrated the debris mantle more effectively. However, refraction analysis (Fig. 2) indicated a consistent pattern of first arrivals and suggested the presence of three layers. Three longitudinal arrays on the main trunk and one on the middle tributary indicated a debris-mantle seismic velocity of 400–500 ms^–1, with the second and third layers having velocities of about 1700–2300 and about 4500 ms^–1, respectively (Fig. 2). We estimate a debris-mantle thickness of 2–4 m, which is similar to that observed by Reference Elconin and LaChapelleElconin and LaChapelle (1997) in moulins and crevasses. We estimate the depth of the second discontinuity to be about 14–23 m. Extrapolations of valley-wall geometry suggest that the base of this second layer represents a discontinuity within the rock glacier, rather than the contact of valley bottom or the valley walls. This is discussed further in Reference Bucki and EchelmeyerBucki and Echelmeyer (2004). However, the difficulty experienced with source/geophone coupling at the debris mantle precludes detailed resolution of subsurface structure, including the basal layer.

Fig. 2. P-wave first arrival times from an array along the main-trunk center line. Layer 1 is the debris mantle; layers 2 and 3 may represent a discontinuity within the ice—rock mixture. V_i = (slope of linear segments)^–1.

Radio-echo sounding

Thirty ice radar soundings were made with an avalanche-style radio-echo sounder (Reference Watts and EnglandWatts and England, 1976). The antennas were orientated perpendicular to flow, with a spacing of 30–50 m. The observed waveforms showed a distorted signal that appears to be caused by the overlap of the airwave and a return wave (Fig. 3). Such a waveform is typically observed when sounding a very shallow ice glacier. Because of this interference, it was difficult to determine the return signal. However, in some cases we were able to estimate the travel time of the first return, typically in the range 0.12–0.29 μs. Using an electromagnetic wave velocity of 170 m μs^–1 (clean ice; Reference Petrenko and WhitworthPetrenko and Whitworth, 1999) we obtained a reflector “depth” of 10–25 m. These possible returns were not observed everywhere and exhibited no obvious spatial relationship, so they were not used to define rock-glacier geometry.

Fig. 3. Examples of RES: (a) an airwave with no return signal; (b) interference of the return signal and the airwave creates a distorted signal. (For a deeper rock glacier the return signal would occur after the airwave in each of these panels.) Travel times are determined by measuring the point that the airwave begins to distort as shown by the arrow in (b).

Electrical methods

There are various electrical resistivity methods that can be used to investigate the subsurface of a rock glacier. We used both d.c. resistivity and TEM, which is an electrical induction method, to investigate Fireweed rock glacier. To aid in our interpretation of these soundings, we made laboratory measurements of the electrical resistivity of rock samples from the debris mantle and others inferred to underlie the main trunk. Our sampling indicates the debris mantle on the west flow contains 80% intrusive rock, much of which is thermally altered. The mantle of the middle flow contains equal amounts of intrusive rocks and mudstone, and the mantle on the east flow comprises two distinct sections. These sections vary in their amounts of intrusive rocks and mudstone; the easternmost part of this flow contains less of the thermally altered rock. We note that these compositions may not represent the relative concentrations of these rock types within the ice—rock mixture.

Electrical properties of rock types

Electrical resistivities of each sample were measured in a laboratory using conventional time-domain methods (Zonge Engineering; Table 1). The altered intrusive rocks had quite low resistivities (~800 Ωm), while the non-altered igneous rocks and both mudstones had higher resistivities (>2000 Ωm), similar to those measured by Reference Keller, Hohmann and NabighianKeller (1991). We also determined the resistivity of the periglacial stream water. It was quite high (~8500 Ωm), which indicates that it does not contain significant amounts of dissolved ions. These measured resistivities are likely to be somewhat different than in situ values because of fractures, interstitial fluids and water flowing within the rock glacier, so we use them only as guidelines when interpreting our geophysical data.

Direct-current resistivity

Most often used in rock-glacier surveys are d.c. resistivity methods (Reference Fisch, Fisch and HaeberliFisch and others, 1977; Reference Evin, Fabre and JohnsonEvin and others, 1997; Reference HauckHauk, 2001; Reference Hauck, Guglielmin, Isaksen and Vonder MühllVonder Mühll and others, 2001), in which the resistivity structure is determined by applying a voltage between electrodes placed in the substrate. Good electrical contact between these electrodes and the substrate is important for this technique. This can be limiting on rock glaciers because the electrodes must be placed in the coarse and often dry, unconsolidated debris mantle. On Fireweed rock glacier we attempted d.c. resistivity measurements with dipole–dipole arrays (Reference ReynoldsReynolds, 1997). In order to overcome the sensitivity that this array type has to electrode contact resistance (Reference HauckHauck, 2001) we used brine-soaked sponges at each electrode. In spite of these efforts we did not acquire any interpretable data.

Transient electromagnetic (TEM) methods

TEM methods can be used to determine resistivity distribution at depth by measuring the decay of an induced magnetic field (Reference NabighianNabighian, 1979; Reference Kaufman and KellerKaufman and Keller, 1983; Reference Nabighian, Macnae and NabighianNabighian and Macnae, 1991). These methods are not often used in glaciology but have been used in permafrost studies and on a few rock glaciers (Reference HauckHauck, 2001; Reference HauckHauck and others, 2001). Here we provide a brief summary of the basic principles of this method.

Unlike d.c. methods, TEM uses an inductive source and does not require direct electrical-current injection into the debris mantle. Applying a current to a TEM transmitting loop generates a primary magnetic field. Abruptly shutting off this transmitter-loop current induces currents in the substrate. These induced currents generate secondary magnetic fields that decay in proportion to the electrical resistivity of the substrate. These decaying secondary magnetic fields are measured with a receiver loop while the transmitter current is off; the geometry used in this study is shown in Figure 4.

Fig. 4. Array geometry used in our survey (NanoTEM^®, Zonge Engineering), which collects a series of data at 31 progressive time windows per measurement, and 400–1000 individual measurements (made at 32 Hz) stacked to compose a sounding. For each measurement, data collection begins at about 1.5 μ and extends to 3 ms after transmitter turn-off.

The effective depth of investigation for TEM sounding depends on the size of the transmitter loop, background noise and resistivity of the substrate. The rate at which the current can be shut off in the transmitter loop prior to measurements is a limiting factor in resolving shallow depths. Rock glaciers can be considered shallow in the context of TEM soundings and we are interested in near-surface structure, so rapid termination of the transmitter current is required. Systems that employ such a rapid turn-off (~1.5 μs) are capable of depth resolution of a few meters, even in resistive substrates such as ice.

It has been shown by Reference NabighianNabighian (1979) that the combined effect of all induced currents in a uniform half-space can be approximated by a single current filament moving downward with a velocity υ given by

where ρ is the resistivity of the half-space, μ is its magnetic permeability and t is the time since the turn-off of the primary magnetic field. This current filament expands like a “smoke ring” downward at an angle of 35–45° from the transmitter loop. In a layered half-space the current filament moves with varying velocity as it crosses layer boundaries. If it encounters a layer with a very low resistivity, its velocity is reduced (Reference Nabighian, Macnae and NabighianNabighian and Macnae, 1991). ATEM sounding is a measurement of the time rate of change of the secondary magnetic field generated by this downward-propagating “smoke ring”.

Reference KaufmanKaufman (1979) gives the response for a circular in-loop array at the surface of a uniform half-space. This is shown in Figure 5a for a transmitter with radius 11m and a current of 3.5 A. The transient response can be divided into three stages: early time, intermediate time and late time. In the early time the rate of change of the magnetic field is nearly constant, with a value that is proportional to the resistivity of the half-space (early-time segment in Fig. 5a). Intermediate time is the transition from early time to late time. The resistivity of the half-space influences how quickly this transition is made. This is also illustrated in Figure 5a, where two half-spaces having 90 and 20Ωm resistivities are compared. At late time the transient response can be approximated as

where k = 2.5 for a uniform half-space, as shown in Figure 5, while for a layered geology k can vary from about 1.5 to 3.5. A ₀ encompasses physical constants and geometry. Changes in the slope of the transient-decay curve can be used to infer the presence of layers of different resistivity. However, the relationship is not unique.

Fig. 5. Time rate of change of the secondary magnetic field is measured (dH_z/dt) vs time from the shut-off of the primary magnetic field. (a) Responses from two half-spaces after Reference KaufmanKaufman (1979) with an 11m transmitter radius and current of 3.5 A. “Early”, “intermediate” and “late” correspond to decay stages and are shown for the 90Ωm half-space. (b) Complete response in the presence of a conductor can be modeled as the superposition of a power-law response from a half-space and the exponential response of a conductor (Equations (2) and (3)). Late time is divided into late time I and late time II.

At some point in time, the secondary magnetic fields decay to background noise levels. This causes increased scatter and a leveling in the TEM transient curves (Fig. 5a). However, this decay to background noise levels is not observed if nearby geology includes a sufficiently low resistivity unit, i.e. a “conductor”. In the context of TEM modeling, one-dimensional (1-D)means a model with horizontal layers. Two-dimensional (2-D) means a model of geologic structure with arbitrary variation in cross-section, but no variation along strike. Three-dimensional (3-D) modeling incorporates arbitrary variation both in cross-section and along strike. Two-dimensional conductive bodies, such as cylinders or slabs with infinite strike extent, and 3-D conductive bodies, such as spheres or rectangular block-shaped geologic features, generate an exponentially decaying secondary field (Equation (3)) rather than the power-law decay characteristic of the half-space (Equation (2)).

where A ₁ encompasses geometric parameters and τ is the characteristic time of the conductor. The characteristic time contains information on the dimensions and resistivity of the conductive object (Reference Nabighian, Macnae and NabighianNabighian and Macnae, 1991). When a conductive body is situated within a half-space, the TEM decay curve has contributions from both the half-space and the conductor. A conductor within a resistive half-space can be recognized if its response is sufficiently strong relative to background. To illustrate this, Equations (2) and (3) can be superimposed, as shown in Figure 5b for various values of τ and A ₁ = 30. Here, we distinguish between two stages: late time I and late time II. In late time I the TEM transient follows a power law (Equation (2)). In late time II the transient decays into background noise levels or, if a conductive body is present, has an exponential late-time decay (Equation (3)).

Transient-decay curves contain a significant amount of detail about the subsurface resistivity structure; however, this structure is not necessarily directly apparent. To more clearly express the resistivity structure of the subsurface, the transient-decay curves can be re-parameterized into resistivity–depth curves through inversion. Converting the decay curve into either an apparent resistivity curve or a smoothly varying resistivity model improves the expression of geologic structure. Apparent resistivity curves are determined by fitting a uniform half-space response (Reference Kaufman and KellerKaufman and Keller, 1983) to each data point of the transient-decay curve. Smooth-model inversion assumes a horizontally layered resistivity structure and solves for a model of resistivities that vary smoothly from layer to layer and that generates a calculated transient-decay curve that matches the measured data as closely as possible (Reference Ward, Hohmann and NabighianWard and Hohmann, 1987). This later approach can provide more detail about the subsurface than apparent resistivity calculations because it uses a more complete model of subsurface resisitivity. Smooth-model resistivities are based on a complete solution to the layered-earth TEM response, and they provide more realistic values of the true resistivity structure than apparent resistivity curves. Alternatively, a layered inversion with fewer layers but with no restrictions on resistivity changes between layers can be used. This type of inversion allows specification of the number of layers, layer resistivity and layer thickness. Zonge Engineering developed the inversion software used in this study. Modeling results from a transect of soundings can be plotted as a pseudo-section showing contours of resistivity vs distance along line and depth from surface.

Errors in TEM measurements occur if there are deviations from the assumed shape and relationship of the transmitter and receiver loops. For this reason, it is important to maintain consistent loop geometries throughout a survey. Topographic effects can also introduce error in TEM soundings, but accounting for the relative position of each sounding reduces these errors.

Summarizing the expected structure of the TEM transient-decay curves, we note there are several key features. The nearly constant level of the early phase and its duration give some information about the overall resistivity of the upper layers. A change in the slope of the late-time I stage represents resistivity changes with depth, whereas the late-time II stage can indicate the presence of conductive geologic features following Equation (3). These guidelines apply when 1-D assumptions are valid. We note that 2-D and 3-D structures can influence any portion of the curve.

TEM data

Ten soundings were performed along the center line of the main trunk (Fig. 1b). The first sounding was made about 120 m up-glacier from the terminus, and subsequent soundings extended to the uppermost main trunk. An additional 30-sounding transect was made across the main trunk and included soundings off the rock glacier on or near bedrock exposures. This transect crossed the main trunk about 150 m up-valley (Fig. 1b). Figure 6a shows three typical decay curves from the rock glacier and two from the bedrock.

Fig. 6. (a). Decay curves from the transverse transect; three soundings are from on the rock glacier (r.g.) and two are from on the bedrock (b.r.). Error bars reflect the standard deviations of the stacked records in each case. A relatively slow decay after the arrow occurs in many of the r.g. soundings, but not for those soundings on b.r. The slow decay is characteristic of a conductor. (b) Shows how this conductive response varies along the transverse transect. We use window 17 (indicated in (a) as “Win. 17”) for all soundings in this transect. Plotting the magnitude graphically shows how the conductive response varies across the rock glacier. It is interesting to note that the strongest response is offset to the west.

Discussion of TEM results

The decay curves in Figure 6a are, at first viewing, quite similar to each other, especially for times before late II. Close examination of the late II times highlights differences that represent variations in resistivity structure between the bedrock and the rock glacier. At these times, the bedrock curves rapidly become noisy, whereas the rock-glacier curve shows a characteristic exponential decay (Equation (3)). This slow, less noisy decay is observed in all curves from the center-line profile and in most curves taken on the rock glacier from the transverse profile. This behavior suggests that there is a conductive feature within, on or under the rock glacier that is not present in the bedrock. To illustrate this we have plotted the magnitudes of window 17 (the data point at approximately 0.1 ms in Fig. 6a) for all soundings in the transverse profile (Fig. 6b). This window (and most others in late II) is near zero and sometimes negative for soundings taken off the rock glacier. However, for those soundings on the rock glacier (between vertical arrows in Fig. 6b) this window is always positive and often high-magnitude. Figure 6b shows how the conductive response varies across the rock glacier and that there are distinctions among soundings off and on the rock glacier. We note that the low-resistivity response is greatest just to the west of the center line.

One-dimensional smooth-model and layered inversions are used to determine subsurface resistivity structures (Fig. 7). Both types of inversion techniques show the same basic resistivity structure for the rock glacier and the same for the bedrock. The rock-glacier soundings (R.G. A–D, Fig. 7) consistently show a surface layer a few meters thick with resistivity values near 100–300 Ωm. This layer overlies a more resistive region that is 60–70 m thick and ≥1000 Ωm. The inversions for most of the rock-glacier soundings are clearly distinguished from bedrock soundings by a low-resistivity structure that underlies the highly resistive region (compare R.G. A–C with B.R. A and B). The resistivity of this structure does vary among the rock-glacier soundings. For example, R.G. C in Figure 7a shows a much lower resistivity at depth than R.G. D. The modeled structure in R.G. D is of higher resistivity than the other rock-glacier soundings but it is still an order of magnitude lower than the bedrock resistivity at depth. These model results are consistent with what would be expected from the raw data (Fig. 6) and our preliminary analysis above.

Fig. 7. (a). Smooth-model inversions of rock-glacier soundings (R.G. A–D) and bedrock soundings (B.R. A and B). (b) Three-layer inversions for the same soundings as those in (a). Layered inversions are typically used for picking depths when geology is 1-D.

Figure 7b shows 1-D layered inversions for the same soundings as in Figure 7a. In a layered inversion the number of layers is fixed and the inversion is used to solve for thickness and resistivity. In Figure 7b, horizontal line segments represent the contacts of these layers. Based on our smooth models (Fig. 7a) we used a three-layer model to represent an upper layer, which is perhaps the debris mantle, a second layer thought to be an ice—rock mixture, and a third, lowermost layer, for which resistivities and thickness were not assigned. Layered inversions produced the same basic resistivity structure as the smooth models but provide an estimate of layer thickness and rock-glacier depth. For most of the rock-glacier soundings, the debris-mantle thickness is well resolved through the layered-inversion method and is similar to that determined with seismic methods (2–4 m; Fig. 2).

The conductive structure (<80Ωm) at depth is evident in most of the rock-glacier inversions, but is not evident in the bedrock inversions. This conductive structure is thought to represent a till-like layer. Elconin (1995) described the existence of a <2 m thick wet mud layer between the bed of the rock glacier and the bedrock. He describes an exposure of this material between the eastern bedrock wall and the rock glacier as containing mostly clay-, silt- and sand-size material with angular clasts 5–150 mm in diameter. The material is described as being colored yellow, red and orange-brown, with a fetid odor. Elconin (1995) notes the presence of alder roots oriented in the direction of flow, which may indicate deformation within this layer. This mud may be analogous to a deformable subglacial till (e.g. Reference PatersonPaterson, 1994). It was also reported that within the rock-glacier/ ice—rock mixture there was significant silt and clay (Reference Elconin and LaChapelleElconin and LaChapelle, 1997). From these observations, it is quite likely that there exists a significant amount of wet, fine-grained material beneath the rock glacier. However, some of the geophysical evidence and the analysis of rock-glacier motion (Reference Bucki and EchelmeyerBucki and Echelmeyer, 2004) seems to indicate that a shear plane exists at 20–28 m depth where most of the motion of the rock glacier probably occurs.

It is unlikely that this conductive response at depth is the result of a different type of bedrock found under the rock glacier than along its margins. The conductive response could also represent water-saturated bedrock or water flowing under the rock glacier. However, if this is the case, then this water must not be connected to the highly resistive water flowing in the periglacial stream (Table 1).

The modeled depth for the till-like layer near the terminus is deeper than the thickness of the rock glacier measured at the terminus exposure (58 m). This suggests that our layered inversions produce an overdeepened depth-to-bed model. This may be a result of applying a 1-D model to a TEM response of a 2- or 3-D feature. To investigate this we used a 2-D TEM forward model (Reference Rijo and HohmannRijo and Hohmann, 1999; Zonge Engineering) to calculate a TEM response over the center of a hypothetical rock-glacier-filled channel. Figure 8 shows the expected responses for a simple 1000Ωm half-space, a 1000Ωm channel within a 500Ωm bedrock half-space and this same channel with a ribbon of till situated at the bottom of the channel. These 2-D features create significant deflections from the simple half-space model (Fig. 8) at early times. Using the forward modeling results for the channel plus the till model, we used the 1-D inversion to reconstruct the resistivity. We found that the 1-D inversion successfully recovers the low-resistivity feature but places it at a greater depth. This is consistent with our observations near the terminus and we expect that it would be true for all the soundings up the center line of the rock glacier, since the entire main trunk seems to be situated within a trough (Reference Bucki and EchelmeyerBucki and Echelmeyer, 2004).

Fig. 8. Two-dimensional forward models. A simple half-space, a channel and a channel with a conductive till at the bottom center.

We cannot reliably use layered inversions to determine rock-glacier depth because of the negative 2-D effects discussed above. Instead, we use the measured 58 m thickness at the terminus to calibrate the smooth-model inversion of the nearest terminus center-line sounding. This calibration defines the “depth to till”, which we specify to be ~900Ωm. This calibration is then extended up-glacier for the remainder of the center-line soundings. We note that the resolution of the basal discontinuity is limited by the averaging effect of the smooth-model inversion. To pick the depth, we use the depth that corresponds to the 900Ωm level in all of the soundings.

The smooth-model inversions for each of the ten soundings along the center line are plotted along section in Figure 9a, where the location of each sounding is shown along the top axis. The data are then contoured to produce a pseudo-section along the rock glacier from about 120 m to 530 m up from the terminal face (located at 0 m), which we interpret as an approximate profile of rock-glacier thickness. This profile shows structure in the basal topography that is similar to that observed in the surface topography and marginal features. For example, Figure 1 shows the location of two snow-filled depressions along the margins (“D”), where we might expect a deeper rock-glacier bed. There is a corresponding depression in the pseudo-section at a longitudinal position of 340 m (“D” in Fig. 9a), and a corresponding change in surface slope, consistent with a rock glacier flowing over a dip in its bed.

Fig. 9. (a). NanoTEM smooth-model resistivity cross-section; “D” corresponds to the feature marked with the same letter in Figure 1b. (b) Center-line profile determined from NanoTEM soundings by methods described in the text. (c) Transverse parabolic cross-section 170 m from terminus, determined from surface topography and center-line NanoTEM depth. Parabolic geometry is inferred. Hae is height about the ellipsoid.

From the center-line profile data, the corresponding thickness and the slope of the valley walls, it is possible to construct an approximate transverse cross-section that conforms to reasonable channel geometry, as shown in Figure 9c. We do know through observations at the terminus (Reference BuckiBucki, 2002; Reference Bucki and EchelmeyerBucki and Echelmeyer, 2004) that the actual channel geometry may be slightly asymmetrical, where the deepest part is actually offset to the west.