Wind-formed gravel bed forms, Wright Valley, Antarctica


Corresponding author: J. A. Gillies, Division of Atmospheric Sciences, Desert Research Institute, 2215 Raggio Pkwy., Reno, NV 89512, USA. (


[1] Bed forms composed of gravel size particles (≈50% of particles >4 mm) are observed in the Wright Valley of the McMurdo Dry Valley system in Antarctica. These bed forms are characterized by a very asymmetrical shape with a mean aspect ratio of 0.025 (standard deviation 0.005), mean wavelength of 2.7 m (±0.49 m), and a mean height of 0.06 m (±0.01 m). Particle size analysis of the bed form sediments shows bimodality with a peak near 9 mm and another between 0.5 mm and 0.25 mm. Time-integrated sediment trap samples of horizontal saltation and creep flux indicate the flux of particles ≥4 mm during the two-year monitoring period was extremely low. Measurements of the horizontal displacement of tracer particles (14 mm, 12 mm, 10 mm, 8 mm, and 6 mm diameter) placed onto the bed forms corroborate the low particle flux measurements and limited movement of particles. The bed forms share form and grain size characteristics with both ripples and mega-ripples, showing poor sorting of particles across a single wavelength except for a slight coarsening at the crest similar to ripples, but their sinuosity suggest that transverse instabilities affect their formation similar to mega-ripples. Based on the data for the prevailing environmental conditions it can be argued that the Wright Valley form is an expression of gravel particles moved solely by highly intermittent creep processes. This also argues for the need for a very long period of time for their evolution, on the order of centuries.

1. Introduction

[2] Unusual bed forms composed of mainly gravel-sized material (2.0 mm–16 mm diameter [Folk, 1974]), formed by wind-driven sediment transport and of limited spatial distribution, are observed in the Wright Valley of the McMurdo Dry Valley system in Antarctica (Figure 1). The morphology of these bed forms is very different when compared with ripples that form on subaerial exposures of sand [Sharp, 1963], and the larger aeolian bed forms known as mega-ripples [Yizhaq, 2004, 2008; Milana, 2009]. They also have form and particle size characteristics different from other ripple-like bed forms composed of coarse particles observed in other locations in the Dry Valleys [e.g.,Selby et al., 1974; Ackert, 1989; Henderson et al., 2002]. The bed forms of interest in the Wright Valley, previously identified as mega-ripples [Lancaster, 2002], are long, low, sinuous-crested and reminiscent of small water waves frozen in place on a very flat beach, and appear on visual inspection to be composed of relatively homogeneous gravel particles (Figure 2). Despite the size of the material that comprises these bed forms, they appear to be of aeolian origin although the exact process responsible for their formation remains uncertain. It is most likely that the gravel particles are moved by the impact of saltating grains that are derived from the deposited sediments of the Onyx River, as are the gravel particles themselves. Because of the well-packed gravel surface, saltating grains rebound with highly elastic collisions reaching heights in excess of 2 m, resulting in a substantial increase in particle velocity and subsequent momentum transfer to the bed [Bagnold, 1941; McKenna Neuman, 1990]. Nearby structures are stripped of paint on their sides well above 2 m, similar to the results McKenna Neuman [1990] obtained using abrasion plates in the Arctic.

Figure 1.

Location of the Wright Valley in the McMurdo Dry Valley system and the position of the bed forms.

Figure 2.

The Wright Valley wind-formed gravel bed forms from two perspectives.

[3] Smaller-scale terrestrial bed forms of wind-blown sediments span a continuum from ripples through mega-ripples. Sand ripples typically have wavelengths between 13 mm to 300 mm and an amplitude (i.e., height) range of 0.6 mm to 10 mm [Bagnold, 1941; Sharp, 1963; Lancaster, 2009]. Ripples have uni-modal grain size distributions with little difference in the distribution as a function of position on the form except for coarsening of particles at the crest [Lancaster, 2009]. These ripples form by grains that are splashed by saltation [Anderson and Bunas, 1993]. Although an understanding of ripple and mega-ripple formation is still incomplete, it is fairly certain that sand ripple wavelengths scale primarily as a function of grain size and wind speed [Sharp, 1963; Werner, 1990; Andreotti et al., 2006], while mega-ripple wavelengths have been hypothesized to scale as a function of the mean saltation hop length and also as a result of interactions of different size ripples as the form evolves [Yizhaq, 2004].

[4] Bagnold [1941]specified that three conditions are necessary for mega-ripple formation: 1) a constant supply of sand for saltation, 2) a sufficient supply of coarse grains with diameters 3–7 times larger than the main population of saltating grains, and 3) wind speeds that remain below the threshold of the coarse grains so they are not removed from the crest. InBagnold's [1941]model mega-ripples can grow indefinitely as long as the supply of particles is maintained.Isenberg et al. [2011]provide evidence that the growth of mega-ripple height is self-limiting due to two possible mechanisms: 1) the increasing shear stress at the crest as the form protrudes higher above the surface, which upon reaching a critical threshold will entrain particles from the crest thus limiting further growth, and 2) as wind speed increases, saltator impact speed also increases and the high speed tail of the distribution of saltator impact speeds may remove coarse grains from their crest position. According toYizhaq et al. [2008], there is ample documentation of mega-ripple forms in the field, which demonstrates their morphological diversity.

[5] Mega- or granule ripples have been described in many places including the Libyan Desert [Bagnold, 1941], California [Sharp, 1963], the northern Sinai [Tsoar, 1990; Yizhaq et al., 2008; Isenberg et al., 2011], the Namib Desert [Fryberger et al., 1992], Iceland [Mountney and Russell, 2004], and Brazil [Yizhaq, 2008]. Milana [2009]reported the largest yet-described mega-ripples high in the Argentinean Puna Plateau. Reported wavelengths of mega-ripples are between 0.3 m to 43 m and have heights ranging from <0.4 m to 2.4 m [Bagnold, 1941; Ellwood et al., 1975; Milana, 2009]. Characteristics typical of sand and mega-ripples morphometric parameters are provided inTable 1.

Table 1. Summary of Typical Ranges of Ripple and Mega-ripple Morphometric Parameters
Ripple TypeStoss Angle (Sα) DegreesLee Angle (Lα) DegreesHeight (H) (m)Wavelength (λ) (m)RIRSI
Impact (normal)1.7–1020–30<0.01–0.10<0.1–0.252–70>1.0

[6] For mega-ripples bi-modality of the transported sediments is considered a diagnostic feature and a necessity for formation [Yizhaq, 2004]. Development of large mega-ripples has been suggested to require decades to centuries [Bagnold, 1941], while smaller sized examples may take hours to days [Sakamoto-Arnold, 1981], weeks [Sharp, 1963] or years [Yizhaq et al., 2008] to form depending upon the conditions. Only recently have rigorous empirical research methods been applied to study the evolution of mega-ripple formation [e.g.,Jerolmack et al., 2006; Isenberg et al., 2011; Zimbelman et al., 2009; Yizhaq et al., 2012], but a convincing mathematical model has yet to be developed [Yizhaq et al., 2008].

[7] The Wright Valley bed forms provide the opportunity to investigate several aspects of the evolution of coarse-grained aeolian bed forms that, as yet, are not well constrained or understood including the length of time that may be required to form such features under present-day wind regimes [Qian et al., 2012], the apparent need of bi-modality in the sediments [Yizhaq et al., 2008; Isenberg et al., 2011], and the fact that these features do not appear to be connected, repetitive bed forms similar to other mega-ripple type aeolian bed forms. To examine where these bed forms fit on the continuum of aeolian bed forms, a two-year study was undertaken, beginning in January 2008, to obtain data on their morphology, particle size and movement characteristics. The purpose of this paper is to report these data, which can subsequently form the basis for the development of a model to explain how these aeolian bed forms develop their characteristic form, and migrate. Here we offer an estimate of the time scales required to move the gravel-sized particles over length scales equivalent to the mean wavelength of the bed forms and the span of the field of bed forms, and we present a preliminary explanation for their formation.

2. Materials and Methods

2.1. Research Area

[8] The location of the gravel bed forms of interest is in the Wright Valley (77° 31.109 min. S, 161° 51.045 min. W). The field of these bed forms extends approximately 2 km down the valley and in the direction roughly parallel to their crests extends a distance of 2.4 km, which covers an area of approximately 5 km2. The seasonally flowing Onyx River lies approximately 1 km to the south of where the gravel-covered surface begins to express a wave-like form and the bed forms terminate just before the bedrock becomes exposed and begins to slope upwards toward Bull Pass in the Olympus Range (Figure 3).

Figure 3.

Schematic cross section of the Wright Valley showing the relative positions of the Onyx River floodplain, the location of the bed forms, and the point of termination.

2.2. Methods

[9] To obtain the necessary data to examine morphology and movement of these bed forms a multipronged approach was employed, which included meteorological observations, leveling, measurements of sediment flux and tracer particle movement, and particle size analysis. The forms were surveyed using a Trimble Spectra Precision Laser LL400 and HL700 Laserometer receiver attached to a stadia rod. This laser allows for rapid data capture over very long transects and is rated to a precision of ±1.5 mm at 30 m. A Sokkia C310 Level was used to mark the endpoints of each survey section to ensure each transect was straight. Azimuths of the crest directions were determined using a handheld Global Positioning System (Garmin eTrex). Seven transects varying in length from 90 m to 296 m were surveyed.

[10] A 4 m high tower erected adjacent to saltation traps, held a combination wind speed and direction propeller type anemometer (Model 05103, R.M. Young, Inc., Traverse City, MI) at the top of the tower, and three cup anemometers (Model 1900 #40C, NRG Systems, Hinesberg, Vt.) spaced logarithmically between 1.2 and 4.2 m. Air temperature and relative humidity were measured at a height of 2 m. The output from each meteorological instrument was measured at 1 Hz and the ten minute average values stored on a data logger (CR1000, Campbell Scientific, Inc., Logan, UT).

[11] Annual fluxes of the saltation and traction loads were measured using two different types of custom-designed traps (Figure 4). Mean horizontal saltation flux was measured at two locations in the approximate center of the study area and in close proximity to each other (≈10 m). Modified versions of the self-orienting passive sediment trap originally designed byFryrear [1986]were deployed at the designated sampling location. This type of trap has been used extensively for time-integrated measurements of sand-size particles moving in saltation in harsh environments [e.g.,Gillette et al., 1997a, 1997b, 2001; Gillette and Chen, 1999]. These passive collectors maintain a collection efficiency of ≈90% for a wide range of wind speeds [Shao et al., 1993]. For use in Antarctica the traps (Figure 4) were modified to withstand the expected high wind speeds and sediment transport rates. Each trap had four wedge-shaped catchers spaced logarithmically at heights between ≈0.2 m and ≈1.3 m. The opening of each catcher is 0.02 m × 0.05 m. Unlike otherFryrear [1986] type trap arrays that have been used to collect saltation samples at multiple heights [e.g., Gillette and Chen, 1999] the four catchers in this trap design are not allowed to move independently of each other. They are connected through the tail fin assembly, which keeps them all facing the same direction and aligned into the wind.

Figure 4.

(a) The self-orienting saltation trap shown with inlet covers in place. (b) The traction load trap.

[12] Overfilling of a trap would result in loss of information for estimating integrated sediment flux, as visiting the sites during the year was limited to the time frame of the month of January in each year of the project this problem needed to be addressed. We constructed a trap closing mechanism that is triggered by the filling of the lowest receptacle to a set capacity. Within the bottom catcher a light emitting diode (LED) and sensitive photocell were set in place so that they were slightly lower than the plane defined by the bottom of the trap opening. The total collection volume is approximately 0.0025 m3. Under the control of a data logger the LED is turned on at the beginning of each hour for ten minutes. If the photocell recorded light during the illumination then no action is taken. If the photoreceptor did not register the light from the LED for six consecutive hours, a signal was sent from the data logger to a linear actuator that lowers four covers over the trap openings. At the completion of the closing of trap one, a signal is sent to the linear actuator on the second trap, which raises the covers on that trap allowing collection to begin. The time of the closing/opening defining the end and beginning of the collection intervals for traps 1 and 2, respectively, is recorded by the data logger. The second trap would close if its LED/photoreceptor pair became buried by the sediment collected in its bottom catcher, and that time would be recorded to define the end of the collection interval.

[13] The mass of sand in the catchers and the period of duration of collection are used to interpolate the integrated horizontal saltation mass flux using the empirical formula of Shao and Raupach [1992]:

display math

where a, b, and c are constants and z is height above the ground. The total mass of transport was defined for a sampling interval for the ≈1.36 m layer represented by the trap height (defined as the height of the geometric mean of the top trap opening) as:

display math

where zi = i (m), and Δzi = 0.01 m. The dimensions of Qs are kg m−1 in the defined collection period t. To measure localized saltation activity (frequency and event duration) at the selected sites two Saltation Flux Impact Responders (Safires) were deployed. Baas [2004]evaluated the performance of these piezoelectric crystal type instruments and found that the Safire presents a minimal obstruction to the wind flow and provide high-frequency omni-directional measurements at a relatively low cost compared with other piezoelectric type sensors.

[14] To provide an estimate on the mass flux of particles moving below the lowest catcher of the saltation traps (≤0.24 m) and to obtain a representative sample for particle size analysis for this component of the sediment in transport, six traction load traps were installed to the east of the meteorological tower, scattered throughout the field of bed forms. These traps were mounted flush with the surface presenting an opening 0.027 m × 0.255 m that was aligned parallel to the local crest of the form (Figure 4). The total collection volume of a trap was 0.0036 m3.

[15] To acquire data on the horizontal displacement of representative particles in the bed forms, five sizes (14 mm, 12 mm, 10 mm, 8 mm, and 6 mm) of tracer particles (faceted glass beads, ρt ≈ 1.2 g cm−3) were placed on the surface (Figure 5). The spherical shape of the tracer particles does not match exactly the range of forms observed in the gravel particles, but neither the tracer nor the gravel particles are smooth spheres. Over 1000 tracers were placed onto the surface at three positions: the space between the crest and the beginning of the identifiable stoss slope, the stoss slope, and the crest, for 35 positions in total. Small pins were set into the surface at a distance of approximately 2 m apart parallel with the crest at each measurement position. A string was stretched between the pins and elevated from the surface. Tracers were placed under the string (large to small), separated by 0.05 m spacing such that the string bisected the middle of the particle when looking from above. Tracers were not placed preferentially on the surface with respect to their exposure. They could be placed in interstitial areas between surrounding mineral particles or not depending on the arrangement of particles at the placement position. The pins were left in place and the string removed. Upon returning to the field the string was replaced and the distance the tracers had moved perpendicular to the reference line was measured using a tape to millimeter precision. Tracers were set in their initial positions in both of the study years, and in the second year no 8 mm particles were used. For the tracers that were left for two years, the total distance traveled from the reference string for the entire test period was recorded.

Figure 5.

(a) The tracer particles in place on a crest. Location of tracers is indicated by arrows. (b) Measuring horizontal displacement of tracer from the white string that marked the initial position.

[16] In addition to the surveying and process measurements, sediment samples (several kilograms of material) of the surface (78 sampling locations) and material removed at depth (>0.01 m to >0.15 m, below surface layer to lacustrine unconformity) were collected from the stoss slopes, the crests, and the intervening surfaces between crests and the beginning of stoss slopes to determine the particle size distribution properties of the form. Multiple pits were excavated to examine the stratigraphy of the sediments. The material collected in the sediment traps was also analyzed for particle-size distribution characteristics for comparison with the surface sediments and those collected at depth.

[17] Prior to sieving the samples collected from the surface and at depth were allowed to air dry for several days. Water content was not deemed an overly important issue due to the dryness of the environment and the overall lack of very fine material contained in the samples. Samples were dry sieved with a mechanical shaker at 1/2φ intervals from −4.5φ to 4.0φ (26.5 mm to 0.0625 mm) and weighed to 0.001 g precision.

[18] The same sieving procedure was carried out for the samples collected in the saltation and traction traps. An additional step was carried out to characterize the particle size distribution for the ≥6 mm sized particles in the traction trap sediments for comparison with the saltation trap and tracer particle data. The material in each traction trap ≥6 mm was removed and each grain was measured using calipers to an accuracy of 0.01 mm. Particle size distributions were developed from these measurements.

3. Results

3.1. Environmental Conditions During the Study Period

[19] Wind and temperature conditions for the two years of monitoring are summarized in Table 2as a function of season, and for those periods when saltation was observed. The distribution of wind speed as a function of direction and for the combined seasons of spring-summer and fall-winter are presented as wind roses inFigure 6. As these figures show, there are seasonal changes between the warmer and colder periods. Principally, the frequency of occurrence of winds with easterly to northerly components dominate in the spring and summer periods while during the fall and winter there is an increase in the frequency and magnitude of southwesterly winds. Maximum 10 min mean wind speed at 4.2 m approached 23 m s−1 in the fall and winter periods in both years of the study.

Table 2. Wind Speed and Temperature Conditions in the Wright Valley for Each Season, in Each Study Year, and for the Same Periods When Saltation Was Observed
 Wind Speed (m s−1) at 4.2 mTemperature (°C)
Minimum (10 min. Mean)Maximum (10 min. Mean)Mean (Seasonal)Minimum (10 min. Mean)Maximum (10 min. Mean)Mean (Seasonal)
Summer 2008–2009013.84.4−39.09.7−8.1
Fall 2008021.14.9−48.60.3−27.4
Winter 2008022.94.9−54.9−5.8−32.0
Spring 2008017.45.3−33.99.7−7.9
Summer 2009–2010013.84.6−36.05.4−8.0
Fall 2009016.92.7−51.3−4.5−29.8
Winter 2009021.44.2−41.5−2.2−28.8
Spring 2009017.94.8−44.87.4−11.2
 Wind Speed (m s−1) at 4.2 m During SaltationTemperature (°C) During Saltation
 Minimum (10 min. Mean)Maximum (10 min. Mean)Mean (Seasonal)Minimum (10 min. Mean)Maximum (10 min. Mean)Mean (Seasonal)
Summer 2008–20097.313.810.9−
Fall 20083.521.111.7−27.50.2−12.1
Winter 20084.322.912.7−38.0−5.8−19.0
Spring 20083.917.411.1−21.09.7−4.0
Summer 2009–20106.213.810.5−13.34.7−1.9
Fall 20093.016.911.0−26.0−4.5−12.5
Winter 20093.021.412.4−33.3−3.1−17.5
Spring 20095.217.911.4−21.26.5−5.4
Figure 6.

Wind roses for the summer-spring and fall-winter seasons combined for the Wright Valley study site.

[20] The summer high temperature approached 10°C in both years with low values in the winter below −50°C. No transport of sand was recorded by the saltation sensors when temperatures dropped below −38°C. Saltation was observed to begin as wind speed measured at 4.2 m exceeded 10 m s−1 at temperatures >−38°C.

3.2. Bed Form Morphology

[21] To quantify the morphology of the bed forms the parameters of: 1) ripple height (H), 2) stoss slope length (SL), 3) lee slope length (LL), 4) stoss slope angle (Sα), and 5) lee slope angle (Lα) were calculated for each form in each transect by approximating each as a stylized triangle (Figure 7). The coordinates of the vertices were used to calculate the morphological parameters. Spacing (λ) of the bed forms was calculated as the crest-to-crest distance, the ripple index (RI) as λ/H [Tanner, 1967], and the ripple symmetry index (RSI) as SL/LL [Tanner, 1967].

Figure 7.

Idealized form for a ripple-like bed form.

[22] Based on data from the survey lines these bed forms are characteristically asymmetric and are not regular in their distribution as are, for example, well-developed sand ripples. Individual forms are often discrete such that the lee slope of one is not necessarily connected to the stoss slope of the next. The profiles do not reveal a sinusoidally oscillating form, but rather an irregular oscillation, with stoss slopes much longer than lee slopes, creating an elongated form. Stoss and lee slope lengths can exceed 1 m and 0.5 m, respectively, but there is considerable variability and inconsistency between transects. Mean crest height, by transect, ranges from 0.05 m to 0.08 m with standard deviations 0.02 m to 0.03 m, which is approximately half the height of the pebble ripples in the Victoria Valley described bySelby et al. [1974] and the gravel ripples described by Henderson et al. [2002] on the Bennett Platform, a snow free table northeast of the summit of Mt. Black on the west side of the Shackleton Glacier (85°13′00.0″S, 177° 50′00.0″W). Mean spacing by transect ranges from 2.10 m to 3.21 m with standard deviations 0.74 m to 1.03 m, which is similar to the mean wavelength of 3.5 m for the pebble ripples in the Victoria Valley [Selby et al., 1974]. RI ranges between 36 to 54, which is uncharacteristically large for aeolian bed forms, and the mean transect values have large standard deviations of RI (7.7 to 25.7) indicating high inconsistency between transects. RSI values are all >1 indicating pronounced asymmetry in the form. The mean spacing between the lee and subsequent stoss slopes ranges from 2.0 m to 5.5 m for the different transects, but the range of observed values is from zero to >40 m, indicating this aspect of these bed forms is highly variable. This variability makes it difficult to define a characteristic wavelength for this form, which suggests that the RI may not be an appropriate metric to calculate and compare with RI values from other studies.

3.3. Sediment Fluxes

[23] The mass of sediment collected in each trap as a function of height, for each collection period, showed a strong exponential decrease providing confidence that applying the method of Shao and Raupach [1992] to calculate horizontal flux was appropriate (Figure 8). Collection orifice height was defined following the protocols of Ellis et al. [2009]. In 2008–2009 there was an insufficient amount of sand collected in the bottom receptacle of trap number one to close the trap door and open the second trap. The Safire recorded 827 h of saltation activity during the 347 days of sampling in 2008–2009, which allows for the calculation of the horizontal flux of sand for a 24 h period of saltation of 6.81 kg m−1 day−1. In 2009–2010, trap one reached its capacity after 290 days had elapsed (14-1-2009 to 31-10-2009) and closed, which triggered trap two to open for an additional 70 days (31-10-2009 to 1-14-2010). The horizontal flux of sand for a 24 h period of saltation for trap one was 34.70 kg m−1 day−1 and for trap two 113.41 kg m−1 day−1, based on 693 and 170 h of saltation activity, respectively. Combining the mass data for both traps and using the total hours of saltation time (i.e., 863 h) the horizontal flux was 49.5 kg m−1 day−1 (per 24 h of active saltation).

Figure 8.

Mass of sediment in saltation trap catchers as a function of height for the sampling periods in the two study years.

[24] Based on the saltation activity and the wind direction data the directional characteristics of the sediment transport can be apportioned by season for the two study years. There is a distinct seasonality to the sediment transport system as shown in Figures 9 and 10. In the spring and summer periods saltation activity is observed to occur for easterly winds (67°–112.5°) accounting for 43% and 44% of all saltation activity in 2008–2009 and 2009–2010, respectively. During the spring and summer saltation activity from the southwest (202.5°–247.5°) occurred for 44% of the total saltation time in 2008–2009 and 42% in 2009–2010. During the fall and winter periods saltation activity is observed to occur for winds with a strong southwesterly direction (202.5°–247.5°) accounting for 91% and 87% of all saltation activity in 2008–2009 and 2009–2010, respectively. Saltation with an easterly component occurs approximately 2.5% of the time in the fall and winter. Based on the orientation of the crest-lines of the bed forms it would appear that they are most closely aligned with wind and sand transport conditions associated with the fall and winter seasons.

Figure 9.

Saltation frequency as a function of direction for the spring and summer seasons 2008–2009 and 2009–2010.

Figure 10.

Saltation frequency as a function of direction for the fall and winter seasons 2008–2009 and 2009–2010.

[25] The traction traps were deployed for a total of 364 days in 2009–2010 (they were not deployed the first year) and accumulated between 0.765 kg and 1.099 kg of sediment. Assuming that the period that saltation occurred for all traps was 863 h, the flux ranged from 0.084 kg m−1 day−1 to 0.120 kg m−1 day−1 with a mean value of 0.107 kg m−1 day−1 (standard deviation = 0.016 kg m−1 day−1) for the equivalent of 24 h of continuous saltation. Although the units of flux are the same for the saltation and traction load traps, the length dimension for the saltation trap represents a vertical column 1 m wide through the saltation layer (i.e., height integrated sample), whereas for the traction trap the length term is associated with sediment moving through a meter wide space in contact with and close to the surface (although saltating particles can also enter the opening).

3.4. Particle Size Distributions

3.4.1. Stratigraphic Samples

[26] The grain size statistics for the samples removed from the bed forms and at depth were computed using the GRADISTAT program [Blott and Pye, 2001]. Images of the surface grains and grains at depth are shown in Figure 11. Grain size distribution data for the samples of sediment removed from the surface of the bed forms show remarkable similarity when compared by position and between transects (Table 3). Bimodality is observed with a large peak near 9 mm and a smaller peak around 0.4 mm corresponding to gravel and coarse sand, respectively. It should be noted that as most samples show bimodality, the calculated moments for the entire particle size distribution are diminished in value (Table 3).

Figure 11.

Stratigraphic section through the Wright Valley bed forms showing the coarse surface layer and the unsorted and unstratified sediments at depth.

Table 3. Grain Size Statistics for Sediments Removed From the Surface of the Bed Forms at Three Positions
Transect No. inline image (mm)σ (mm)Median (mm)SkKs
StossCrestIntervening SurfaceStossCrestIntervening SurfaceStossCrestIntervening SurfaceStossCrestIntervening SurfaceStossCrestIntervening Surface
Transect No.Mode 1 (mm)Mode 2 (mm)D10 (mm)D90 (mm) 
StossCrestIntervening SurfaceStossCrestIntervening SurfaceStossCrestIntervening SurfaceStossCrestIntervening Surface   

3.4.2. Trap Samples

[27] The range of mean particle diameter of the sediment in the saltation traps is 0.70 mm–0.28 mm, which is in the medium to coarse sand category [Folk, 1974]. In the bottom receptacle of the traps the largest particle diameter observed was between 5.65 mm and 4 mm (−2.5ϕ to −2ϕ), which accounted for 0.5% of the mass for trap one and 7% in trap two in the 2009–2010 sampling period. In the first year 0.5% of the mass represents ≈10 particles and in the second year 7% represents ≈500 particles, assuming spherical particles of density 2,650 kg m−3.

[28] The traction load traps collect a sample of the particles that are creeping and/or reptating along the surface but also include particles that are saltating and have trajectories that would carry them into the open slit of the trap. The sediment in the traction load traps was dominated by sand-sized particles with a mean diameter of approximately 1 mm.

[29] Over 400 particles ≥6 mm were separated from the sediment collected in the six traction load traps and measured with calipers to characterize the distribution of grain diameters that corresponded to the gravel sized component of the bed forms and the tracer particles. The mean number of particles in this size range entering a traction trap was 67 with the largest particle being 16 mm.

[30] Particles ≥6 mm in diameter contributed 2.48% to the total mass collected in the traction load traps. Apportioning the mass by the particle size class equivalent to the tracer particle diameters is: 0.14% for 6 mm, 0.6% for 8 mm, 0.7% for 10 mm, 0.5% for 12 mm, 0.28% for 14 mm, and 0.23% for particles >14 mm. Based on the assumption that the particles are spherical and of density 2650 kg m−3 the horizontal flux of the gravel size particles (as would be collected in the traction load trap) for the total time that saltation was observed (i.e., 863 h) can be estimated. The flux rates by particle size are: (6 mm) 0.031 kg m−1, (8 mm) 0.139 kg m−1, (10 mm) 0.169 kg m−1, (12 mm) 0.114 kg m−1, (14 mm) 0.063 kg m−1. The total flux for the gravel particles of equivalent size range of the tracer particles is 0.570 kg m−1 during the 863 h of saltation in the one year period the traps were deployed.

3.5. Horizontal Tracer Displacement

[31] During the study period less than 10% of the tracer particles were lost. This rate of recovery provided the opportunity to develop a robust data set with which to assess horizontal displacement distances as a function of particle diameter. Table 4 provides measures of central tendency for horizontal displacement distances for each tracer particle size as a function of position for the one year samples and the two year sample. For tracers left for one year mean displacement distance ranges between 0.014 m and 0.29 m, and after two years this increased to 0.34 m to 0.6 m. Annual mean displacement generally increases with decreasing particle size and the tracers placed initially on the crest show enhanced displacement compared to the middle and trough positions. The standard deviation of mean displacement distance also indicates tracers moved from the crests have the greatest variability.

Table 4. Measures of Central Tendency for Horizontal Displacement Distances for Each Tracer Particle Size as a Function of Position for the 1-Year Samples and the 2-Year Sample
Tracers out for 1 Year Beginning 2008Intervening Surface Particle Diameter (mm)Middle of Stoss Slope Particle Diameter (mm)Crest Particle Diameter (mm)
inline image Horizontal Displacement (m)0.0220.0220.0260.0360.0920.0160.0170.0210.0300.1210.0550.0470.0410.1110.046
σ Horizontal Displacement (m)0.0340.0410.0570.1190.1620.0350.0370.0640.0350.2120.1380.1110.0890.2140.118
CV Horizontal Displacement (m)0.0060.0050.0050.0030.0060.0040.0050.0030.0090.0060.0040.0040.0050.0050.004
Min. Horizontal Displacement (m)000000000000000
Max. Horizontal Displacement (m)0.1310.1700.2800.6530.8370.1670.1350.3000.1320.8580.5440.4110.4850.8130.651
Med. Horizontal Displacement (m)0.0050.0030.0060.0190.0430.0040.0030.0080.0180.0420.0050.0030.0080.0310.013
Norm. inline image Horizontal Displacement1.
( inline image displacement/d)               
σ Norm. Horizontal Displacement2.43.45.714.927.
Tracers out for 1 Year Beginning 2009Intervening Surface Particle Diameter (mm)Middle of Stoss Slope Particle Diameter (mm)Crest Particle Diameter (mm)
141210 6141210 6141210 6
inline image Horizontal Displacement (m)0.0210.0220.065 0.0140.0720.4050.070 0.2860.0350.0930.073 0.147
σ Horizontal Displacement (m)0.0270.0250.046 0.0210.0680.2370.078 0.3780.0960.1350.131 0.265
CV Horizontal Displacement (m)0.0080.0090.014 0.0070.0110.0170.009 0.0080.0040.0070.006 0.006
Min. Horizontal Displacement (m)000 0000 0.002000.010 0.020
Max. Horizontal Displacement (m)0.0920.0810.217 0.8150.1070.1940.270 1.1150.4360.4380.564 1.199
Median Horizontal Displacement (m)0.0070.0090.050 0.3460.0050.0040.042 0.1530.0050.0310.034 0.067
Norm. inline image Horizontal Displacement1.51.27.2 24.4
(mean displacement/d)               
σ Norm. Horizontal Displacement2.01.76.8 63.06.811.213.1 44.1
Tracers out for 2 Years (2008–2010)Trough Particle Diameter (mm)Middle Particle Diameter (mm)Crest Particle Diameter (mm)
inline image Horizontal Displacement (m)0.0340.0510.0560.1620.3020.0390.0560.1050.1430.3120.0790.0840.2030.1710.606
σ Horizontal Displacement (m)0.0470.1020.0750.1630.2790.0510.0920.2140.2140.3890.1460.1590.4620.2062.346
CV Horizontal Displacement (m)0.0070.0050.0070.0100.0110.0080.0060.0050.0070.0080.0050.0050.0040.0080.003
Min. Horizontal Displacement (m)000000000000000
Max. Horizontal Displacement (m)0.1730.4510.2920.6511.2810.1650.3681.111.091.5720.60.7572.4150.86514.65
Median Horizontal Displacement (m)0.0180.0140.0340.1180.2670.0080.0050.0470.0910.1910.0130.0240.0370.0770.080
Norm. inline image Horizontal Displacement2.54.35.620.350.42.84.710.517.852.
(mean displacement/d)               
σ Norm. Horizontal Displacement3.38.57.520.446.53.67.721.426.864.810.413.346.225.7391.0

[32] The tracer data show skewed distributions for displacement of particles, with the tail of the distribution associated with the largest movements. In the 2009, one-year sample ≈92% of the tracers moved between 0 and 0.15 m. In the 2010 one-year sample, ≈80% all particles moved between 0 and 0.15 m. That a greater percentage of particles moved distances >0.15 m in the 2010 sample is likely a result of a more active saltation system at the site, as described earlier. The two-year sample, which is the tracers placed initially in 2008, shows that movement is still dominated by small amounts of horizontal displacement and that there were particles in each size class that still had not moved from their initial positions.

[33] Accounting for all tracer particle sizes and positions for both measurement years combined, horizontal displacement as a function of particle size is predictable, as shown in Figure 12. This figure shows that the normalized horizontal displacement (mean displacement distance/particle diameter) increases as a power function of decreasing normalized particle diameter (particle diameter/largest particle size). The power law form suggests a self-similar process of displacement for particles as function of their size at least for the range of sizes measured. There will be limits on the applicability of this relationship to characterize particle displacement as a function of particle size. This limit will be for particles that are too large to be moved by the forces acting to move the gravel particles as well as minimum size limit where particles will be susceptible to transport by saltation.

Figure 12.

The relationship between normalized horizontal displacement distance (mean displacement distance/particle diameter, NHDD) and normalized particle diameter (particle diameter/largest particle diameter, NPD) for all positions and for both study years combined.

4. Discussion

4.1. Bed Form Morphology

[34] The spatial variability of the bed form geometry and spacing between successive crests is quite large both along transects and among transects, which results in no or very weak relationships between form parameters (i.e., H, SL, LL, Sα, Lα, and λ) and distance from the emergence of the bed forms to the point of termination of the field. These parameters plotted as a function of distance (not shown) suggest that the bed form becomes more elongated, taller, and develops steeper stoss slopes as a function of distance, indicating some evolution of the form is occurring.

[35] Each stoss/lee unit appears to be discrete and the nature of the spacing between the units is not well constrained. The variability in the spacing between the lee slope and the beginning of the next stoss slope makes it difficult to compare wavelength-based indices with other types of bed forms. The meanRIvalues range between ≈29 and ≈51 for the Wright Valley transects, which are plotted along with other available data for mega- and granule ripples reported in the literature inFigure 13. As Figure 13 shows, the Antarctic bed form data (including data from the same bed form field from Lancaster et al. [2002]) are very scattered and include a number of points that plot much lower than other reported forms, which results from their low mean height and large mean spacing.

Figure 13.

The relationship between bed form amplitude and wavelength for aeolian mega- or granule ripples as compared with the range of values observed for the Wright Valley bed forms (open white circles (this study) and gray solid circles [Lancaster et al., 2002]. The location codes and references for the other data are as follows: Victoria Valley [Selby et al., 1974], Asgard Range [Ackert, 1989], Bennett Platform [Henderson et al., 2002], Edwards (Edwards Air Force Base, California [Ward and Greeley, 1984]), GSDNM (Great Sand Dunes National Monument, Colorado [Williams et al., 2002]), Coachella (Coachella, California [Sharp, 1963]), Kelso (Kelso Dunes National Park, California [Sharp, 1963]), NK (Nahal Kasuy, Israel [Isenberg et al., 2011]), and Med-Argentina and Large-Argentina (Puna Plateau, Argentina [Milana, 2009]).

[36] As the Antarctic bed forms lack a regular wavelength, comparison with other studies is problematic. The mean aspect ratio (H/(LL + SL)) may be a more appropriate metric in this case to compare among ripples, mega-ripples and other fluid-formed bed forms. The mean aspect ratio for the Antarctic bed forms is 0.025 with a standard deviation of 0.005. Drawing from graphical data presented inZimbelman et al. [2009, Figure 4] for Great Sand Dunes National Park and Preserve and Nahal Kasuy granule ripples [Isenberg et al., 2011, their Figure 3] an estimate of aspect ratios for these forms is ≈0.07 and ≈0.05, respectively. Based on Ackert's [1989]claim that the largest Asgard Range ripple-forms are 0.5 m high and 8 m wide, the maximum aspect ratio of that form is ≈0.06.Bourke et al. [2006, Figure 4] present a topographic profile of Martian transverse aeolian ridges (TARs) from which a mean aspect ratio can be conservatively estimated to be ≈0.08. For aeolian sand ripples Andreotti et al. [2006] report a typical aspect ratio is 0.04, and for subaqueous ripples aspect ratios are typically ≥0.05 [Venditti et al., 2005]. Even from this small sampling it appears that the Wright Valley bed forms presents a unique aspect ratio for smaller scale bed forms, being a third less than aeolian ripples, half as much as subaqueous ripples, and approximately three times lower than the examples presented for terrestrial granule ripples and Martian TARs.

4.2. Particle Size Distribution

[37] Wind-formed impact ripples and mega-ripples are commonly reported as having finer-grained particles in their trough compared with coarser grains at their crests, with mega-ripples showing a distinct bi-modality in their particle size distribution (PSD). Particle size analysis of the sediment samples removed from the Wright Valley bed forms and in the intervening space between consecutive forms reveals that no grain size segregation exists as a function of position for the mean and median grain sizes, similar to the pebble ripples in the Victoria Valley described bySelby et al. [1974]. There is no difference in the gravel and coarse sand mode sizes as a function of position either. There is a significant difference in sorting between the material lying on the surface between the end of the lee slope and the beginning of the stoss slope and the particles at the crest (as determined by ANOVA) with the crest material being better sorted, and there is a small but statistically significant increase in the D90particle size at the crest as well. This coarsening may reflect preferential removal of smaller grains from the crest zones creating a more well-sorted sediment at this position on the form.

[38] The removal of the finer grains is likely due, as described by Isenberg et al. [2011], to high-speed saltating grains impacting crest grains with sufficient force to dislodge particles and send them over the brink. IfBagnold's [1941] observation that saltating particles, upon impact, can move particles that are six times their size, the limit of particles able to be moved, based on the largest mean particle size range measured in the saltation traps (0.44 mm), would be ≈3 mm. This increases to ≈6 mm based on the mean particle size of 0.96 mm measured in the traction trap. The traction trap and tracer granule data both indicate however, that particles larger than 6 mm are being moved during saltation, but they are a very small proportion of the sediment being transported. The saltation trap data indicate that at heights ≥0.24, but <0.54, the largest particle size in saltation was between 5.65 mm and 4 mm.

[39] The characteristic of bi-modality in the particle size distribution of mega- or granule ripples is considered a diagnostic feature and the surface is typically covered by a layer of coarse grains with a core of sediments that are sand-sized [e.g.,Zimbelman et al., 2009]. This is clearly not the case for the Wright Valley bed forms as evidenced by the stratigraphic profile (Figure 11), and the particle size data from the surface to the unconformity, which shows a decrease of the mean diameter with depth due to the increasing content of sand with depth. Comparing the particle size distribution of the samples taken at depth with samples removed from an exposed sandbar in the Onyx river floodplain (Figure 14) revealed that floodplain is the sand source.

Figure 14.

Comparison of particle size distribution for sediment samples removed from the stoss slope, crest, and intervening surface between forms (Transect 4) and a sample collected from the Onyx River floodplain.

[40] Comparing the mean particle size distribution of the fine mode in the surface samples collected from the bed forms with the mean particle size distribution of the material in the traction load traps is also instructive (Figure 15). The percent composition in each size class in the fine mode (0.71 mm to 0.063 mm) of the surface samples correlates very highly with the same diameter size classes in the traction trap material. This suggests that the sand in the bed forms is representative of the material moving over the surface and is incorporated into the form once the wind drops below threshold. Some of this material will become trapped among the gravel-sized particles and is synonymous with whatFryberger et al. [1992]called the “poured-in” texture. This material however, does not play any essential role in the morphology of the bed forms except for its function to deliver energy to the surface during saltation to move the larger particles in the creep mode. The lack of sorting as a function of position with a slight coarsening at the crest is similar to aeolian sand ripples, but the movement of the large particles by impact from the saltating grains is similar to aeolian mega-ripples, but the particles do not reptate, they only creep.

Figure 15.

Comparison of particle size distribution for sediment samples removed from the surface of the bed forms (mean of all positions, black bars) and the mean particle size distribution of the sediment collected in the traction trap (white bars).

4.3. Particle Flux Rates

[41] The horizontal particle flux rates, as a function of particle size, as determined from the traction load traps, provides clear evidence that the annual throughput of gravel sized particles that form the surface layer is extremely low. Zimbelman et al. [2009], Jerolmack et al. [2006], and Isenberg et al. [2011] estimated creep fluxes (Qc) for granule ripples and found Qc ranged between 5 × 10−5 kg m−1 s−1 and 9 × 10−5 kg m−1 s−1 for granule ripples 0.03 m to 0.1 m in height. Our traction traps provide a means to estimate the creep flux for the Wright Valley bed form, which for all particles ≥6 mm diameter in equivalent time units is 1.1 × 10−7 kg m−1 s−1.

[42] Zimbelman et al. [2009] and Isenberg et al. [2011] provide data on granule movement rates that provides a second means to compare creep fluxes between locations. Zimbelman et al. [2009] reported that approximately 45 granules of 1.5 mm diameter passed over a 1 m length of ripple crest in 1 s, while for Isenberg et al.'s [2011] ripples, approximately 88 granules, 0.78 mm in diameter, passed over a 1 m length of ripple crest in 1 s. By comparison, in the Wright Valley an estimated 263 particles passed through a 1 m width during 863 h of saltation, giving a flux in comparable time units of 8.5 × 10−5 granules s−1. Factoring in the size of the particles for each location the actual mass flux rates are not too dissimilar (1.67 × 10−4 kg m−1 hour−1, 1.29 × 10−3 kg m−1 hour−1, and 3.96 × 10−4 kg m−1 hour−1). The important difference is the number of particles being transported during periods of saltation.

[43] In the Wright Valley the flux of gravel particles (from the traction trap data) is four orders of magnitude less than the horizontal saltation flux determined from the vertical traps. By comparison, Jerolmack et al. [2006]report the creep flux to be two orders of magnitude less than their trap-measured horizontal saltation flux. The saltation system in the Wright Valley for the period studied moved only a fraction (0.07) of the amount of sand in 24 h of continuous saltation as compared with the flux rate reported byJerolmack et al. [2006] for White Sands, NM. The lower saltation flux and much larger particles of the Wright Valley bed forms create conditions that lead to very low rates of creep transport.

[44] If we assume an average bed form width (LL + SL) for these bed forms of 2.66 m, with unit crest length (1 m) and a bulk density of 2,080 kg m−3 (estimated from the samples returned to the lab), a 0.01 m thick layer of surface material would contain 55.3 kg of sediment. If we then assume that in 35.1 days of active saltation per year, 0.57 kg of material passed through the 1 m unit crest length, it would take ≈99 years to accumulate enough saltation days to move the ≈55 kg of material the average width of the bed forms. By extension it would take approximately 37,294 years to move this material over 1000 m, roughly the width of the field of bed forms. This assumes that the environmental conditions are reflective of the conditions that prevailed during our brief study period. The estimate must be considered conservative as it is not known what wind speeds and saltation fluxes are required to actually move particles ≥6 mm, although it must be within the measured ranges. It is likely that particles do not move every time sand saltates over the surface, and based on the orientation of the crests it appears that the process of formation and evolution of the form has been driven by saltation events confined to a narrow wind direction. Even if the saltation hours that represent events that are outside a south to westerly direction are eliminated and the total saltation hours in 2009–2010 are reduced to 667, the effect on moving the observed gravel particle flux is minimal, reducing the time to move the mass that represents a unit length and average form width from ≈37 to ≈36 years, which translates into 36,516 years to move that mass 1000 m distance. Time appears to be of critical importance to allow the Wright Valley bed forms to achieve their current state.

4.4. Tracer Particles

[45] The tracer particles provide another means to constrain the amount of time that it takes for the gravel sized particles to move through the bed form field, assuming that the two years of measurements are representative of a much longer time frame. Based on the measured horizontal displacement data, a simple discrete time Markov Chain rule program was developed to evaluate how much time would be required for 99% of a sample of all tracer particle sizes to move 1 m. For every time step the following conditions were assumed: 1) the probability of activation is the sum of the percentages of particles that showed nonzero movement in the field, 2) each particle has the same probability for activation, 3) there is no layering of particles, and 4) there is no attempt to account for lost tracer particles. Tracers could have exited the measurement domain, been missed during the measurements, or have been covered by other advancing particles and as such lost from observation, so their actual fate is unknown.

[46] This simple model indicates that it would take 42 years to move all the 14 mm diameter tracer particles 1 m, 32 years to move the 12 mm, 30 years to move the 10 mm, 18 years to move the 8 mm, and 14 years to move the 6 mm particles. The greatest loss of particles is for the 10 mm and 6 mm diameter tracers, which lose 29% and 28% of their initial sample size over 30 years and 14 years, respectively. The other sizes lose ≤7% of their initial number. The times to move 14 mm and 6 mm particles 1000 m are 42,000 years and 14,000 years, respectively, which compares favorably with the creep flux based estimate of ≈37,000 years to pass the amount of material representing a 2.66 m × 1 m × 0.01 m volume of bed forms surface material 1000 m (refer to section 4.3). The movement characteristics of the tracer particles matches very well with the estimate of particle movement based on the traction trap data.

[47] The mass flux of sediment across the bed forms and the tracer particle movement have been referenced to the total time that saltation was observed by making the assumption that movement can only occur during saltation. It should be noted however, that the duration of recorded saltation events at the Wright Valley site is variable. Saltation events were recorded as 10 min mean values of millivolts from the Safire instruments (this was due to data storage considerations), so the actual duration of events less than this is unknown, but essentially they are extreme transients. Longer events when saltation was sustained continually for over 12 h were observed, and other events although not sustained continually (with one to several hours of no observed saltation) stretched over two to three days. During periods when saltation is active, but intermittent, the amount of mass flux per unit width of distance parallel with the bed forms is small. For example, 48 h of continuous saltation would produce a total mass flux of gravel-sized particles (6 mm–14 mm) of 0.0003 kg m−1, which amounts to approximately one 6 mm diameter particle. Forty-eight hours of continuous saltation is not sufficient to accumulate the mass of a single particle >6 mm. In addition, the tracer data shows very limited amounts of forward movement by particles in a two-year time frame.

5. Conclusions

[48] The Wright Valley gravel bed forms described here appear to be different in form from other small-scale bed forms, either subaerial or subaqueous. They share characteristics with aeolian sand ripples, mega-ripples, and subaqueous ripples, but also have features unlike any of those forms. In particular the low aspect ratio of these forms (0.025) does not align with any other small-scale bed forms, being lower by a factor of 1.6 compared to aeolian sand ripples and a factor of two compared to fluvial ripples. The mean aspect ratio is much lower than mega- or granule ripples (≈2.5–≈3).

[49] In terms of grain size distribution, the characteristics of these bed forms more closely resemble ripples in that there is no sorting by position, and only a slight coarsening of the crest. The sinuosity of the crest-lines however, is better associated with mega- or granule ripples, which indicate there are transverse instabilities that cause portions of the form to advance more quickly than other parts [Yizhaq et al., 2012]. There is bi-modality in the sediments that compose the bed forms, but the fine mode represents essentially a lag of sediments left behind as winds cease to transport sand. This material is subsequently incorporated into the bed forms, but it appears to play no role in influencing the morphology of the form. What process scales the wavelength or spacing between crests remains unknown.

[50] The evolution of this form under the environmental conditions that prevailed during the study period would appear to require long periods of time (tens of centuries), but it also occurs due to the accumulation of very small, discrete movements of the particles that make up the bed forms. At no time is there continuous particle motion for the entire population of grains sizes as is the case for ripples and mega- or granule ripples (at least on the scale of seconds) as observed byZimbelman et al. [2009], Jerolmack et al. [2006] and Isenberg et al. [2011]. Accounts of particle movement and flux rates for other areas are unfortunately lacking. With so little movement of the gravel particles occurring annually, it seems doubtful that the pattern could emerge quickly as the availability of mobile gravel is very limited. This process of terrestrial bed forms forming during a highly intermittent sediment transport process may make these Antarctic bed forms, and other variants within the Dry Valleys important analogs for understanding the development of aeolian bed forms on Mars.

[51] Based on form characteristics, the lack of evidence that gravel-sized particles ever lose contact with bed, and that the material in saltation does not play any role in the morphology of the bed forms, it can be argued that the Wright Valley form is an expression of gravel particles moved solely by intermittent creep processes, with no role for splash processes. The actual physics of the process that leads to the emergence of the Wright Valley bed forms remain to be elucidated. The gravel-sized particles creep forward by the impact of grains in saltation (and perhaps to a much lesser degree by reptation impacts), and saltation may be intensified due to the highly elastic collisions with the gravel surface. The cold climate environment, with denser but less viscous air promotes a greater transfer of energy from the fluid to the surface resulting in entrainment of sand sized particles at lower shear stress [McKenna Neuman, 2003] that possess more momentum and a greater ability to do work on the surface for a given wind speed, compared to warmer environments, generating creep of the larger particles. Based on the few published papers on aeolian bed forms of the McMurdo Dry Valleys [e.g., Selby et al., 1974; Ackert, 1989; Henderson et al., 2002] this environment gives rise to a variety of coarse-grained types, which offer the opportunity for further study to enhance our understanding of sediment transport processes, by wind in particular, and the emergence and evolution of bed forms.


coefficient of variation


particle diameter (mm)


crest height (m)


lee slope length (m)


lee angle


horizontal saltation flux (kg m−1 day−1)


creep flux (kg m−1 day−1)


ripple index


ripple symmetry index


stoss slope length (m)


stoss angle

inline image

horizontal displacement distance


height above ground (m)


unit of grain size


wavelength or spacing (m)


particle density (g cm3)


tracer particle density (g cm3)


standard deviation of the mean or particle sorting (mm)


[52] This research was supported by the U.S. National Science Foundation's Office of Polar Programs, Grant ANT-063621 to J. A. Gillies. W. G. Nickling would also like to acknowledge support to him from the National Sciences and Engineering Council, Canada. We are also grateful for the outstanding logistical support provided by the United States Antarctic Program. We would also like to thank Ms. Becky Peace (BFC, USAP) for her invaluable assistance in the field, Dr. Joanna Nield (Department of Geography and Environment, University of Southampton) for fruitful discussions, and finally Mr. Mario Finoro (Department of Geography, University of Guelph) for his outstanding technical assistance in helping to create the instruments that withstood the Antarctic weather. We would also like to thank the reviewers, the Editor, and the Associate Editor for their helpful insights and suggestions that greatly improved this paper.