Zebra stripes in the Atacama Desert: Fossil evidence of overland flow

Geomorphology 182 (2013) 157–172

Contents lists available at SciVerse ScienceDirect

Geomorphology journal homepage: www.elsevier.com/locate/geomorph

Zebra stripes in the Atacama Desert: Fossil evidence of overland flow Justine J. Owen a,⁎, William E. Dietrich b, Kuni Nishiizumi c, Guillermo Chong d, Ronald Amundson a a

Department of Environmental Science, Policy, & Management, University of California, Berkeley, CA 94720, USA Department of Earth and Planetary Science, University of California, Berkeley, CA 94720, USA c Space Sciences Lab, University of California, Berkeley, CA 94720, USA d Departamento de Ciencias Geológicas, Universidad Católica del Norte, Antofagasta, Chile b

a r t i c l e

i n f o

Article history: Received 31 July 2012 Received in revised form 6 November 2012 Accepted 7 November 2012 Available online 15 November 2012 Keywords: Hillslope Erosion Overland flow Soil formation Desert pavement Hyperaridity

a b s t r a c t Some hillslopes in the hyperarid region of the Atacama Desert in northern Chile have surface clasts organized into distinct, contour-parallel bands separated by bare soil. We call the bands “zebra stripes” due to the contrast between the darkly varnished clasts and the light-colored, salt-rich soil. Gravel that comprises the zebra stripes is sorted such that the coarsest clasts are at the downslope front and fine progressively upslope. How and when the zebra stripes formed are perplexing questions, particularly in a region experiencing prolonged hyperaridity. Using GoogleEarth, satellite imagery, and field observations, we report the first quantitative and qualitative observations of zebra stripes in order to test hypotheses of the mechanisms and timing of their formation. We consider soil shrink-swell, seismic shaking, and overland flow as possible formation mechanisms, and find that overland flow is the most likely. Based on cosmogenic 10Be concentrations in surface clasts, salt deposition rates from the atmosphere, and content in the soils, we propose that the salt-rich soils began accumulating ~106 y ago and the zebra stripes formed 103–104 y at the latest. The zebra stripe pattern has been preserved due to the self-stabilization of the clasts within the stripes and the continued absence of life (which would disturb the surface, as seen at a wetter site to the south). We conclude that the occurrence of zebra stripes is diagnostic of a set of distinct characteristics of local and/or regional precipitation, soil, hillslope form, and bedrock type. © 2012 Elsevier B.V. All rights reserved.

1. Introduction Some of the most visually striking features in the hyperarid Atacama Desert of northern Chile are distinct, contour-parallel bands of sorted gravels on soil-mantled hillslopes (Figs. 1 and 2). We refer to these as “zebra stripes” because the gravel, coated with a dark varnish, sharply contrasts with the underlying light-colored, sulfate-rich soils. This pattern is distinct from the more continuous, interlocking desert pavement found on subhorizontal alluvial fans and terraces adjoining these slopes. Zebra stripes on hillslopes appear to be a feature unique to the Atacama Desert since nothing similar has been reported to date from other deserts on Earth, despite extensive study of surface gravel in Arizona (Abrahams et al., 1986, 1992), New Mexico (Wilcox et al., 1997), California (Castle and Youd, 1972; Clark, 1972; McFadden et al., 1987; Anderson et al., 2002; Sylvester et al., 2002; Wood et al., 2005), Egypt (Adelsberger and Smith, 2009), Oman and the United Arab Emirates (Al-Farraj and Harvey, 2000), the Namib Desert (Van der Wateren and Dunai, 2001), and the Negev Desert (Yair, 1983, 1990; Yair and Kossovsky, 2002; Kuhn et al., 2004; Matmon et al., 2009). The first mention of gravel bands on hillslopes in the Atacama Desert

⁎ Corresponding author. Tel.: +1 510 642 3874; fax: +1 510 643 5438. E-mail address: [email protected] (J.J. Owen). 0169-555X/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.geomorph.2012.11.006

was in a conference abstract by Beaty (1983). He called the bands “tiger stripes” and proposed that the stripes were formed due to the shrinking and swelling of the salts in the soil, but his observations were never published. Zebra stripes may be a special case of reorganized desert pavement on hillslopes. Desert pavement is a layer of rock clasts that forms on low-gradient landforms in arid to hyperarid regions (McFadden et al., 1998). It is characterized by a single layer of rock fragments which overlie a sometimes vesicular, always gravel-poor soil horizon whose thickness varies depending on the age of the surface, rate of atmospheric dust and salt input, climate, and topography (e.g. McFadden et al., 1987, 1998; Al-Farraj and Harvey, 2000; Anderson et al., 2002; Wood et al., 2005; Adelsberger and Smith, 2009). Though several theories of desert pavement formation have been proposed, one of the best supported is that clasts “float” on the accumulation of atmosphericallyderived salt and dust such that they experience constant exposure at the surface from the time the geomorphic surface is formed or stabilized (McFadden et al., 1998). The continuous exposure of desert pavement gravels in the Mojave Desert, California, has been demonstrated by comparing their cosmogenic radionuclide (3He) ages with that of nearby exposed volcanic bedrock (the same material from which the soil and its desert pavement formed; Wells et al., 1995). Likewise, Matmon et al. (2009) measured cosmogenic radionuclide 10Be concentrations in surface gravels and gravels from within a soil profile in the Negev Desert,

158

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

Fig. 1. A drawing highlighting the main features of zebra stripes. The size of the clasts on the hillslope is exaggerated to show sorting. Wavelength was measured along the transects as the distance between fronts, regardless of where the transect intersected the zebra stripe (middle vs. toward one end).

Israel, and used modeling to calculate that the desert pavement and the underlying gravel-free soil formed over the last 1.5–1.9 My. On gently-sloping surfaces, patterns within desert pavement appear to at least partially preserve the topography and gravel distribution

prior to the accumulation of the atmospherically-derived, fine material (Wood et al., 2005; Matmon et al., 2009). In Chile, bar and swale patterns persist in desert pavements on Miocene landforms (Ewing et al., 2006), attesting to both the importance of initial gravel distribution

(a)

(b)

(c)

(d)

(e)

(f)

Fig. 2. Examples of zebra stripes. (a) Zebra stripes on a hillslope near Oficina Rosario (OR). (b) Close up of gravel sorting within a zebra stripe on OR, looking upslope. Gravels in the foreground were 4–10 cm in diameter and fined upslope towards a relatively gravel-free zone. The surface of the underlying soil is very smooth and light-colored due to the presence of sulfate salts. (c) Example of a soil profile on OR showing gray, angular rock fragments in a matrix of white gypsum-cemented fines. Brown vertical feature on the right side is a sand-filled crack. Excavation is 1-m-deep. (d) Example of the smaller zebra stripes at CH. In some cases, the fine gravels (the coarsest material) were colonized by lichens, producing the darker gray banding on the hillslope. (e) A close up of a zebra stripe, looking upslope, with an ~8-cm-long lip balm tube for scale. As at OR, particles coarsened downslope, but they were much smaller and the stripe was shorter along-contour. (f) A soil profile on CH showing the thinness of the gravelly soil mantle (~1 cm here) and the fracturing of the underlying bedrock. Fine red dust filled the fractures.

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

and its persistence over millions of years under favorable conditions. Desert pavement is poorly developed, if at all, on moderately to steeply sloping surfaces due to sporadic overland flow or other processes moving gravels downslope (Wood et al., 2005; Adelsberger and Smith, 2009). This disturbance, however, rarely results in any type of patterning of the gravels. What could disturb desert pavement on hillslopes in the Atacama to form such distinctive patterns? Three processes are possible — overland flow, salt shrink-swell, and seismic shaking. Overland flow in the Atacama is rare due to the lack of precipitation. Hyperaridity in the northern part of the Atacama has persisted for the last 2 My (Amundson et al., 2012), preceded by hyperarid periods beginning in the middle to late Miocene (Alpers and Brimhall, 1988; Sillitoe and McKee, 1996; Hartley and Chong, 2002; Dunai et al., 2005; Nishiizumi et al., 2005; Clarke, 2006; Houston, 2006; Rech et al., 2006; Kober et al., 2007). Present-day mean annual precipitation (MAP) in the Atacama is low and decreases with decreasing latitude, from ~100 mm at 29° S to ≤1 mm between 20 and 26° S (Ericksen, 1981). However, MAP does not reflect geomorphically significant rainfall events, i.e. those that create overland flow. Essentially no data on the intensity of storm events exist, even for present day storms (Ortlieb, 1995). The largest storm for which there is hourly intensity data occurred in Antofagasta (Fig. 3) in 1991 and caused debris flows (Vargas et al., 2000). Precipitation averaged 0.5–1.4 cm h−1 and peaked at 2.4 cm h−1 (Vargas et al., 2000). Thus, the best proxy for storm intensity near our area of study appears to be the sedimentary records of flooding and debris flows in Antofagasta. Debris flow deposits near Antofagasta date back to ~23.5 ka and include 45 deposits since ~5.5 ka (Vargas et al., 2006). Between ~8.4 and 5.3 ka there is a conspicuous lack of alluvial deposition in both southern Peru and northern Chile, after which debris flow activity simultaneously began in both regions (Vargas et al., 2006). The Coastal Cordillera (Fig. 3) blocks the eastward movement of many storm systems, thus the frequency with which they reach the interior of the Atacama is likely lower than that suggested by coastal debris flow deposits. Though rare, storms with intensities great enough to mobilize gravels have occurred since the mid-Pleistocene, if not earlier. Two million years of hyperaridity (Amundson et al., 2012) have allowed the accumulation of thick (2 + m) deposits of salt and dust throughout the Atacama (Ericksen, 1981; Ewing et al., 2006). The salt-rich soils often have dramatic, cm- to m-scale prismatic structure which is thought to form through the shrinking and swelling of the soils, most likely through the hydration and dehydration of soil salts (Ewing et al., 2006). Sometimes the cracks bounding these polygons are indirectly visible on the surface through variations in particle size, where coarse particles are preferentially located above the cracks. Though most common on horizontal soils, many hillslope soils in the Atacama also have prismatic structure. Thus, salt shrink-swell could disturb the surface and promote gravel organization. Polygonal structure and lobate surface features have also been attributed to periglacial processes (e.g., Benedict, 1976). Periglacial processes are very unlikely to have occurred at the low elevations where zebra stripes are observed. Andean moraines from the most extensive Holocene glaciation do not extend below ~4500 masl (meters above sea level; versus a maximum elevation of 1410 masl at our study sites) and precipitation would have been too low for ground ice accumulation based on long-term precipitation gradients (Ammann et al., 2001). Seismic shaking has caused downslope movement or displacement of gravels in the Mojave Desert on a gently-sloping (2–4°) interfluve (Michael et al., 2002; Haff, 2005); in the Anza-Borrego Desert, California, where stones on ridge tops were most likely to move (Castle and Youd, 1972; Clark, 1972); and on the steep (>30°) slopes of the Pisgah Crater, California, where rocks rolled downslope, 10 to 80-cm-tall fault scarps formed, and soils slumped (Sylvester et al., 2002). Though no similar observations of quake-driven soil movement have been made in the Atacama Desert, earthquake-shattered ground observed in northern Chile and southern Peru (Keefer and Moseley, 2004; Loveless et al.,

159

Fig. 3. Shaded relief map of northern Chile showing major physiological features (Coast Range, Central Depression, and Andes). Black circles mark the locations of OR and CH. White circles mark large cities for reference. The dashed line outlines the approximate extent of zebra stripes. Map created with GeoMapApp v.1.7.8.

2005, 2009) demonstrates the potential for seismic shaking to be an important geomorphic process in the region. Studies of particles subjected to shaking suggest that for thin layers experiencing relatively largeamplitude vertical shaking, coarse particles will migrate downwards (the “reverse brazil nut effect” of Schroter et al., 2006). While this creates some sorting in small-scale laboratory experiments, no clear mechanism exists for shaking to organize particles into bands at the hillslope scale. How and when the zebra stripes formed, and why they persist, are interesting questions involving the coupling between geomorphology and climate. We hypothesize that each of the proposed formation

160

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

mechanisms would produce zebra stripes with distinct characteristics, such that the correct process can be deduced from field observations. In this paper, we 1) make the first quantitative observations of these features and their associated hillslopes and soils, 2) consider an array of hypotheses of zebra stripe formation, and 3) conclude that zebra stripes were likely produced by overland flow and have been preserved for at least 10 3–10 4 y under hyperarid, abiotic conditions. 2. Methods To characterize zebra stripes and the conditions related to their occurrence, we surveyed, sampled, and observed one hillslope intensively, made qualitative observations of other hillslopes in the Atacama, and used satellite imagery for regional scale analysis. Our primary field site was a convexo-planar hillslope near Oficina Rosario (OR), 70 km south of Antofagasta and 50 km inland at 1400 masl (Fig. 3). This hill was selected because it is well within the zebra-striped area, easily accessible off the Pan-American Highway, and representative of hillslopes in the region with its salt-rich soil mantle and distinct zebra stripes. The hillslope is underlain by fine-grained granodiorite and abutted by an internally-drained basin in which there are Pliocene alluvium and localized sand and salar deposits (Marinovic et al., 1992). A second field site was a hill outside of the zebra stripe zone to the south (CH), located north of Chañaral, 21 km inland from the coast and 680 masl (Fig. 3). The hillslope was underlain by coarse-grained monzogranite. It was representative the hillslopes in this slightly wetter region, with a very thin (b 2 cm) soil cover and small-scale gravel bands similar to zebra stripes. 2.1. Local distribution and description of zebra stripes The distribution of zebra stripes on OR was analyzed using a topographic survey, satellite imagery and field observations. The hillslope was surveyed using a Trimble differential GPS, with data points collected approximately every 2 m. The topographic data were processed with Surfer v. 8.0 (Golden Software, Inc.) and converted to a 3.6-m grid by Kriging. This grid size was found to best capture the general topography using the same approach as Heimsath et al. (1999) and neither overly smoothes, nor puts too much weight on, small features that could be due to survey error. Contours from the topographic data were overlaid with a 25 km2, 0.6-m resolution Quickbird satellite image (Digital Globe, Inc.) (Fig. 4). Zebra stripe fronts, defined by the abrupt downslope transition from dark gravel to light soil, were digitized along eight transects extending from the top of the hillslope to the north, northeast, east, southeast, south, southwest, west, and northwest (Fig. 4). The distances between the fronts were tested for correlation with gradient, curvature, and distance downslope. In the field, particle size and percent clast coverage (C, %) were estimated along seven downslope transects. Five 50-m-long transects were located on OR (G-1, G-2, G-3, and G-4 in Fig. 4; the fifth transect (G-5) could not be located on our topographic map accurately enough to correlate it with topographic characteristics), another 50-m-long transect (G-6) was located on a saddle between OR and an adjacent hillslope to the west, and a 100-m transect (G-7) was located on the adjacent hillslope. A tape measure was laid along the ground and changes in particle size and C along the tape were recorded, dividing each transect into 40–70 distinct segments. Due to time limitations, standard pebble counts could not be made along each transect. Instead, particle size was visually estimated according to five classes which were defined by the dominant clast sizes: class 1 — fine, with an average intermediate diameter, D, equal to 1–2 cm; class 2 — fine and medium, D = 1–5 cm; class 3 — medium, D = 2–5 cm; class 4 — medium and coarse, D = 2–10+ cm, or mix, D = 1–10+ cm; and class 5 — coarse, D = 5–10+ cm. Though the classes overlap, each is distinct in which clast sizes are dominant and which sizes are mostly absent.

Fig. 4. Quickbird imagery overlaid with topographic contours. Contour interval is 5 m. North is up. Black circles labeled 1, 18, 2, 3, and 4 are the locations of soil excavations. Heavy black lines mark the transects along which gravel size and concentration were visually estimated (G-1, G-2, G-3, and G-4). Thin black lines extending N, NE, E, SE, S, SW, W, and NW from the top of hill mark the transects along which zebra fronts were identified using the Quickbird image. The dark, broad linear feature in the upper right is the Pan American highway and the two light gray linear features running parallel to it are bulldozed dirt roads. Two shallow channels run along the southern and western boundaries of the hillslope. Truck tracks are visible as pairs of faint, thin, light gray lines running parallel to the western channel and next to transect G-3. An area of concentrated rilling is visible on the east–southeast face downslope of G-2, and a large rill extends farther up the hill to the north of this area.

Zebra stripe fronts were defined along G-1 through G-7 using a clast index defined as size class × C ÷ 100. Zebra stripe fronts were defined as bands where clast index ≥2.4. The results of the clast index approach were in good agreement with visual estimates and other more subjective metrics, but the clast index approach was preferred for being more objective. 2.2. Soil, rock, and atmospheric dust sampling and analysis Desert pavement is defined as much by the soil as by the surface clasts. In order to characterize the underlying soil, five soil excavations were made along an undisturbed downslope transect on OR (Fig. 4). The soils were described following the methods of Schoenenberger et al. (2002) and were sampled by horizon for analysis of chemical composition, bulk density, and particle size. Soil thickness was defined as the average depth at which rock fragments exceeded 85% across the sampled soil excavation face and retained relict rock structure. Soil fines (b 2 mm, separated by sieving) were shaken for 1 h in deionized water and filtered, then the solute was analyzed for nitrate and chloride ions by ion chromatography (IC) by Dr. Ken Williams of the Lawrence Berkeley National Laboratory. Rock and soil total S (sulfur) measured by ALS Chemex (Sparks, NV) by LECO furnace and soil S was directly converted to sulfate because the rock contained no S. Bulk density was measured by one of two methods. If the soil was moderately cemented by salt, three cubes (~1–2 cm 3) of the soil were carved with a knife, the cubes' volumes were calculated, the soil was dried at 70 °C and weighed, and the bulk densities of the three cubes were averaged. If the soil was too strongly cemented to carve, gravel-free peds were dried at 70 °C, weighed, then dipped in paraffin wax and weighed again while submerged in water, such

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

that the difference in the two masses was equal to the volume of the ped. Soil texture was determined by hydrometer on soil fines from which salt had been removed using a weak EDTA solution (Bodine and Fernalld, 1973). Because the retention of soil salts is high, even on hillslopes, atmospheric deposition rates can be used to estimate the residence time of the soil. A deposition rate of 4 ± 2 g m − 2 y − 1 was measured by Ewing et al. (2006) at a site 70 km north of OR using passive dust collectors and geologic averages, which were found to be similar despite averaging over 1 y and 2 My, respectively. To complement and expand this data set, three passive dust traps were set near OR. At CH, three passive dust traps were emplaced but were stolen. Dust traps were made of aluminum bundt pans in which a wire mesh supported a single layer of glass marbles flush with the lip of the pan (Reheis and Kihl, 1995). The pans were put on poles 1–1.5 m above the ground to avoid the collection of saltating sand. The traps at OR were collected after ~2 y. The pans and marbles were rinsed with deionized water and the solution was filtered. Nitrate, chloride, and sulfate ions in the solution were analyzed by IC by Dr. Williams. The silicate dust from the pans was collected on filters and analyzed for major element chemistry by ICP-MS at the Desert Research Institute, Reno, Nevada. To estimate how long the surface clasts have been exposed on the soil surface, amalgamated samples of 10+ surface clasts (0.5–1 kg) were collected from each excavation site for measurement of cosmogenic 10Be concentration. Following the methods of Kohl and Nishiizumi (1992), the samples were first ground to pass through a 0.5 mm sieve. Quartz mineral grains were isolated through a series of acid leaches then dissolved with an HF–HNO3–HClO4 solution and mixed with Be carrier. Be was then chemically isolated and purified. 10Be/9Be was measured by accelerator mass spectrometry (AMS) at the Purdue Rare Isotope Measurement Laboratory. 10 Be concentration in rock, N (atom g −1), is modeled as: −hρsΛ −1

N ¼ N0 e



ρr εΛ

−1

 −1  −1 −t ðρr εΛ þλÞ þλ 1−e

ð1Þ

where N0 is the initial 10Be content of the sample, λ is the decay constant of 10Be (y −1), t is time (y), P is 10Be production rate at the surface corrected for sample latitude and elevation (at g quartz −1 y −1), h is the soil thickness above the sample (cm), ρs is the bulk density of the overlying soil (g cm −3), Λ is the mean cosmic ray attenuation length (g cm −2), ρr is the bulk density of the rock (g cm −3), and ε is the rock erosion rate (cm y −1). In the case of the surface gravels, N0, h, and ε are assumed to be zero, such that Eq. (1) can be rearranged to solve for the exposure time, t, as, t ¼ Pλ

−1

  −tλ : 1−e

ð2Þ

2.3. Sprinkling and infiltration experiments If zebra stripes form by overland flow, measuring the infiltration rate of the soil could help constrain the precipitation rate necessary to produce overland flow and consider if this rate is feasible under present conditions. Infiltration rates were measured using three sprinkling experiments and four infiltrometer experiments on OR, and three sprinkling experiments on CH. Sprinkling experiments were small scale due to limited water availability and the need for portability. The sprinkler consisted of a nozzle mounted on an adjustable tripod with a pressure gauge and compression stop valve to control water pressure (Wilcox et al., 1986). The water was pumped from a bucket using a 1/6 horsepower Simer Submersible Utility Pump which produced 8–10 psi through the nozzle. Two nozzle sizes were used: a 1/8GG-4.3 W fulljet brass nozzle and a 1/8GG-6.5 W fulljet brass nozzle, both from Spraying Systems Co. The nozzles produced precipitation at rates of about 6 and

161

55 cm h−1, respectively, which was non-uniformly distributed over a circular area ~1.2 m in diameter. Strong diurnal winds occurred at the sites so a windscreen was constructed with a PVC pipe frame and heavy plastic drop cloths. While this greatly reduced wind disturbance, the experiments were conducted at a height of ~0.5 m above the ground to avoid drift in the swirling winds. Due to the low spraying height and pressure, the spray did not have the same kinetic energy as natural rainfall, and nozzles of this size tend to produce drops smaller than natural rainfall (Wilcox et al., 1986). The lower kinetic energy likely decreased the mobilization of fine sediment, but since we were most interested in the mobilization of gravels this did not negatively affect our interpretations. Each sprinkling experiment was performed on a 1 m 2 plot that was bounded on the downslope side by a steel collection trough set into the ground (Dunne et al., 1980). The other sides of the plot were bounded by plastic lawn edging. On OR, the edging was set 5–10 cm into the soil whereas on CH, the edging was laid on the surface because the soil was too thin. The trough had a steel lip that was set into the soil to capture the surface runoff and a lid to minimize direct addition from the sprinkler. The trough had a sloped bottom to send runoff to a spout from which runoff was collected. Runoff samples were collected and their volume measured every 1 or 2 min, depending on runoff rate. Time to runoff, time to steady-state runoff (when the runoff rate was nearly constant for 4–5 sequential measurements), and duration of runoff after spray was turned off were recorded. Sprinkling experiments were run 3–27 min after steady-state runoff was achieved, lasting 10–30 min overall except for the high precipitation rate experiments. We also noted if and when sediment was mobilized. Application rate was measured by covering the plot with plastic and measuring runoff over two one-minute intervals. Infiltration rates were calculated by subtracting the runoff rate from the application rate. The application rate and runoff rate were corrected for direct input of water into the collection trough by measuring the water collected prior to runoff initiation on the plot. After each sprinkling experiment, the collection trough was removed and the face of its trench was excavated to observe how deep the water had infiltrated and if any subsurface patterns were observable. Single-ring cylinder infiltrometers provide a simple means of measuring local maximum soil infiltration rates and were used to confirm the infiltration rates determined from sprinkling experiments. Our infiltrometer was an 11-cm-diameter plastic cylinder which was comparable to the one used by Yair and Klein (1973), though smaller than what is commonly recommended (Bouwer, 1986). The infiltrometer was driven 5–7 cm into the soil, depending on the subsurface conditions, and the disturbed contact between the soil and the pipe was sealed with quick-setting caulk. The infiltrometer was filled to the lip with water and the water level was maintained at a constant level. The time it took to add additional one-liter volumes of water was recorded for 45–60 min. Following the experiment, the infiltrometer was removed and a cross-section parallel to the flow path was excavated to expose how the water had moved through the subsurface. 2.4. Regional distribution of zebra stripes The Atacama spans distinct tectonic zones and climatic gradients, such that the regional extent of zebra stripes might reflect a correlation with these features and indicate a formation mechanism. GoogleEarth was used to identify the regional extent of zebra-striping in northern Chile and to explore other parts of the world where they might occur. This was supplemented by field observations of zebra stripe occurrence made along major roads in northern Chile from the coast to the Andean foothills between ~18.4 and 23.6° S. Less intensive observations were made as far south as La Serena (29° S). Bedrock composition likely affects the size, shape, and concentration of gravels (Molnar et al., 2007; Brantley et al., 2011), all of which might affect the formation of zebra stripes. A study of the correlation

162

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

between bedrock geology and zebra stripe occurrence used data from a digital 1:100,000 Antofagasta geologic map (González and Niemeyer, 2005) and GoogleEarth. Of the digital geologic map coverage in northern Chile, the 1:100,000 Antofagasta geologic map had the most consistent high resolution imagery in GoogleEarth and was within the heart of the region where zebra stripes occur. In an area spanning 23.5–24° S and 70–70.3° W, GoogleEarth was used to draw polygons around 1) all hillslopes and 2) hillslopes with zebra stripes, with the extent of hillslopes estimated from the shaded relief of the imagery because high resolution topographic data were not available. The hillslope and zebra stripe polygons were overlaid on the geologic map using ArcView v.9.3 (ESRI) and the area of each geologic unit within the hillslope polygons and the zebra-striped polygons were calculated using tools within the software. 3. Results 3.1. Zebra stripe occurrence on OR The main features of zebra stripes – their sorting, shape, width, and wavelength – are shown in Fig. 1. Zebra stripes were one clast deep and most clasts were varnished. The clasts were 0.5–3 cm in diameter (intermediate axis), though much larger clasts (>10 cm) were also present. The clasts were sorted within the zebra stripes; a downslope front of tightly packed (high C), relatively coarse clasts graded upslope to progressively finer, more scattered clasts until another coarse front was encountered (Figs. 1 and 2). The clasts were angular except for the largest cobbles and boulders which were rounded by ventifaction and/or concentric sheeting. Almost all were composed of the underlying bedrock, b 1% were composed of indurated gypsum derived from the soil. The clasts appeared randomly oriented, though the shortest axes were usually oriented perpendicular to the surface. Most of the surface gravel was loosely resting on the soil, but some (≤ 10%) were cemented in the soil and others sat atop short (b 1 cm) salt pedestals (Fig. 1). Pedestals occurred under medium to coarse clasts when they were isolated (in areas of low C) or at the downslope front of a zebra stripe. Similarly, the soil immediately downslope of the zebra stripe front was sometimes eroded 1–2 cm, particularly where there was an abrupt transition to an area of low gravel cover, creating a small cliff at the front. Length-averaged C along transects G-1 through G-7 ranged from 40 to 61% and averaged 51%. Zebra stripe distribution was best quantified using the coarse fronts as a guide, with the fronts defined as the darkest portions of the bands in the satellite imagery and by the clast index in the field observations. The downslope fronts did not follow topographic contours perfectly and appeared to form offset, arcuate, concave-downslope forms whose lateral tips merged with other fronts or gradually decreased in concentration

(Figs. 1 and 2). These characteristics of the lateral tips made it difficult to measure individual zebra stripe length with consistency or accuracy. In satellite imagery they appeared laterally continuous for 50+m, whereas in the field they generally appeared not more than several meters long before the gravels decreased in concentration or interacted with another zebra stripe. Estimates of grain size and gravel cover revealed the somewhat periodic nature of the zebra stripes and the sorting within zebra stripes, but also demonstrated the considerable variation in gravel cover (Fig. 5). The average wavelength measured on the Quickbird transects was 6.29 ± 3.05 m, greater than (but not statistically significantly different from) the average derived from field observations of 3.2 ± 1.5 m. This difference was expected given that the resolution of the Quickbird image was coarse relative to the zebra stripe wavelength. The combined measurements from the field and Quickbird transects show that wavelength was generally between 3 and 6 m and right-skewed (Fig. 6). Zebra stripe wavelength and slope gradient were only weakly correlated (Fig. 7a). Fig. 7a also shows that no zebra stripes occurred when slope gradient was b7% or >45%. The low slopes occurred at the crest of the hill, the bottom, and along the saddle between OR and the adjacent hillslope. Though the crest of OR had been disturbed by vehicle traffic, observations of the crest of the adjacent hillslope found that the undisturbed gravels formed a tight pavement until some distance downslope (discussed below). Slope gradients >45% were not present so may not represent an upper limit on zebra stripe occurrence. Contributing area for each zebra stripe front location was approximated using distance downslope. Wavelength was weakly, positively correlated to distance downslope (Fig. 7b). This was true along seven of the eight Quickbird image transects (r 2 ≅ 0.3, Fig. 8a–h). The correlation was weaker along the field transects (r 2 ≅ 0.1, Fig. 8i–o). Wavelength along the southwest- and west-trending transects decreased or slightly increased due to the presence of the saddle and the shallow channel-rill network along the west and southwest side of the hillslope (discussed further below). On the Quickbird transects, wavelength was 2.4–9.2 m within 40 m of the summit (averaging 4.6 ± 1.5 m), and 1.9–15.2 m at 40–120 m downslope (averaging 7.4 ± 3.3 m). These values did not change if data from the west and southwest transects were omitted. Though not statistically significant, the difference in the maximum wavelength between the two sections indicates that the downslope section has much larger (~ 1.5 times) wavelengths than the upslope section. The average wavelength from the field measurements was statistically the same as that measured on the upper section of the hillslope in the Quickbird imagery. Consistent with transect data, field observations suggested the importance of hillslope location on surface clast mobilization. First, a

Fig. 5. Gravel size class and gravel cover (C ÷ 100) vs. distance along transect G-4. Size classes are 1 = 1–2 cm-diameter gravels, 2 = 1–5 cm, 3 = 2–5 cm, 4 = 2–10 + cm, and 5 = 5–10 + cm (see text). Vertical dashed lines indicate the locations of zebra stripe fronts based on the clast index (size class × gravel cover ≥ 2.4). The segments from 0 to 4 m and 20 to 22.5 m show the typical trend of increasing gravel size and cover towards the zebra stripe front.

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

25

cemented soil horizon (described below). Where rills were small and widely spaced, zebra stripes were preserved in inter-rill areas. Where rills were larger or closer together, zebra stripes in the inter-rill areas were disturbed and diffuse. In a few locations on OR and the hillslope adjacent to OR, rills appeared to emerge from small seep areas or to seep into the soil just upslope of an area with greater C only to emerge somewhat downslope.

15

3.2. Soil characteristics and accumulation on OR

10

13

12

11

9

10

8

7

6

5

4

3

2

1

0

5

Wavelength interval (m) Fig. 6. Histogram of zebra stripe wavelength frequency distribution in 0.5 m bins.

minimum distance downslope appeared necessary for zebra stripe formation. On the undisturbed hillslope adjacent to OR, gravel concentration decreased beginning ~ 5 m from the summit and formed clear bands at ~ 10 m downslope. Remnants of disturbed zebra stripes on OR were observed ~ 5–14 m from the summit, initiating closer to the crest on the northwest and southeast sides where the hillslope gradient increased most rapidly. Second, at the transition from the hillslope to the adjacent alluvial plain there was no thick accumulation of surface clasts or evidence of buried colluvial clasts. Zebra stripe distribution was also affected by slope curvature. Zebra stripes were best developed where curvature was divergent (noses) and were least clear, disrupted, or absent where curvature was convergent (hollows). Convergent topography produced channels and rills which affected zebra stripes differently. Near channels, zebra stripes tended to be diffuse, if present at all, with a less clear front, greater wavelength, and lower C overall (Fig. 1). Where rills crossed zebra stripes, the clasts were locally disordered and no longer sorted, though some of the C pattern often remained (Figs. 1 and 4). The disturbance from rills was mostly limited to the area inside the rill and ~ 1–2 cm to either side. In a few places, spillover from one rill to another scoured the soil between rills and disturbed the surface clasts. Rills occurred in three main regions on OR: 1) upslope from the channel bounding the hillslope to the south, 2) upslope from the channel bounding the hillslope to the northwest, and 3) on the east–southeast face downslope of G-2 (Fig. 4). Rills were up to 10-cm-deep and most were ~10-cm-wide, though some are up to 30 to 40-cm-wide. Their shape is irregular and their depth was often limited by a strongly

16

16

field quickbird

(a) Wavelength (m)

Soil chemistry, depth, and particle size at OR are summarized in Table 1. The soil fines (b 2 mm) had high concentrations of NO3−, Cl−, and SO42− (averaging 1.5–4% gypsum, 0.2% halite, and ~300 ppm nitrate). Salt concentrations increased with depth and with distance downslope, as would be expected for a landscape experiencing small, intermittent rainfall events over 2 My (Ewing et al., 2008). Soil thickness generally increased with distance downslope but was extremely heterogeneous, even within each soil excavation (Table 1 and Fig. 9). As another example, the photo in Fig. 2c is of a side face of the ORH-4 excavation and shows much deeper salt-rich soil accumulation compared to the upslope face of the excavation shown in Fig. 9a. The soil had low gravel content (0–40%) and the fine silicate fraction was a sandy loam (68% sand, 17% silt, and 15% clay). The near-surface salt-rich horizons were porous, with densities often b1 g cm−3, whereas deeper horizons in which salt is concentrated and the silicate fraction is higher had soil bulk densities comparable to those in other deserts (~1.4 g cm−3). The soils displayed a similar horizonation (Fig. 9). The surface gravels sat on top of 1–2 cm of light brown, soft, porous, platy, weakly salt-cemented, fine, silicate material that was nearly gravel-free (By or Byk1). On most of the hillslope, the upper 2–4 mm of soil was weakly cemented into a brittle crust. Below this platy horizon was a 2 cm-thick horizon of soft, porous, white, gravel-free, salt and dust (Byk1 or Byk2). These upper 3–4 cm of soil are comparable to the Av horizons described by Wells et al. (1995), McFadden et al. (1998), and others, though we avoid the nomenclature of “Av” since it is not officially recognized by the NRCS (Soil Survey Staff, 1999). Below ~3–4 cm depth was a porous, gravel-poor horizon of varying thickness with strong prismatic structure, mottled brown and white color, and stronger (but discontinuous) salt cementation, including veins of indurated white gypsum. As described above, this layer appeared to limit the depth of rills on the hillslope. With increasing depth, the degree of salt cementation and the size of prisms increased. The prisms were bounded by cracks filled with brown, salt-cemented sand and some gravels. The cracks were up to tens of cm wide and penetrated deep (~1 m) into the soil. The sandy fill was weakly laminated parallel to the prism walls. Slickensides were observed on the faces of some prisms (e.g., Fig. 9e). Some prisms

(b)

12

8

r2 = 0.05

4

r2 = 0.05

Wavelength (m)

Frequency

20

0

163

12

r2 = 0.26 8

r2 = 0.04

4

field

0

quickbird

0 0

0.2

0.4

Gradient (m m-1)

0.6

0

50

100

150

Distance downslope (m)

Fig. 7. Zebra stripe wavelength was weakly, positively correlated to (a) hillslope gradient, and (b) distance downslope. Field data includes only G-1, G-2, G-3, and G-4.

164

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

16

(a) N, r2 = 0.38

12

16

16

(b) NE,

12

16

(c) E, r2 = 0.31

12

r2 = 0.40

12

8

8

8

8

4

4

4

4

0

0

0 0

16

40

80

120 16

(e) S, r2 = 0.34

12

0

50

100

12

0 0

16

(f) SW, r2 = 0.00

40

80 16

(g) W,

12

r2

0

12

= 0.15

8

8

8

8

4

4

4

4

0

0

0 0

40

80

0

40

(i) G-1,

12

r2

12

= 0.13

40

80

16

16

16

(j) G-2, r2 = 0.13

16

(k) G-3, r2 = 0.23

12

12

8

8

8

4

4

4

4

16 12

0

0 0

30

60

90

0

40

(n) G-6,

12

r2 = 0.11

8

8

4

4

4

0 0

20

40

60

40

120

(l) G-4, r2 = 0.11

30

60

90

0

50

100

(o) G-7,

12

r2 = 0.03

8

0

80

r2 = 0.31

16

16

(m) G-5,

120

0 0

80

80

(h) NW,

0

8

0

40

0 0

80

(d) SE, r2 = 0.50

r2 = 0.20

0 0

20

40

0

50

100

Fig. 8. Plots of zebra stripe wavelength vs. distance downslope along each transect observed in the Quickbird imagery (a–h) and in the field (i–o). All but three show wavelength increased with distance downslope. Note that the y axes are on the same scale, while the x axes vary by the length of the transect.

had zones with 40–60% gravel but most were gravel-poor (≤10%, Table 1). The gravel was angular and lacked evidence of chemical weathering. Rock fragment content increased abruptly, though unevenly, beneath the prismatic horizons, grading to bedrock physically weathered by salt. The road cut for the Pan American Highway at the base of OR (downslope of ORH-1, Fig. 3) revealed salt-filled cracks extending >3 m into the fractured bedrock. No evidence of chemical weathering or saprolitization was observed. The excavations revealed that zebra stripe occurrence was not correlated with underlying soil structure or chemistry. A substantial portion of the soil was derived from atmospheric inputs. Using the mass of the salts in the soil and their deposition rates, we estimated a minimum time required for their accumulation. The rates of atmospheric deposition calculated from the dust traps are presented in Table 2. The average rate at OR of 3.6 ± 0.4 g m −2 y −1 was not significantly different from measurements by us ~ 70 km north (YH in Table 2) and was only slightly smaller than the rate calculated by Ewing et al. (2006). Deposition rates of NO3−, Cl −, and SO42− were significantly lower than at YH, with the rate of sulfate deposition (0.554 ± 0.013 g m −2 y −1) being the largest. The ratio of silicate material to soluble material was larger than YH, but still comparable. The difference is due to different geographic settings which control fog movement inland. YH was closer to a valley which crosses the Coast Range and provides a route inland for fog through the mountains. In contrast, OR was located on the lee side of a

laterally-continuous section of the Coast Range which blocks most fog from moving inland. We focus on the results for SO42− because it is the least soluble of the salts measured, thus the inventory in the soils is likely to be the most complete. The upper 3–4 cm of soil had 2–6 ky worth of sulfate (Table 1). The portions of the soil profiles in which gravel content was b5% had between 8 and 691 ky worth of sulfate, and the entire profiles contained 71–692 ky worth of sulfate (Table 1). The calculations are minimum estimates because salt has been transported deeper than our sampling (as seen in the road cut). Salt has also been transported downslope so the values are likely minima near the crest of the hill and overestimates at the lower positions. Taking the average of the five excavations, the hillslope had about 400 ky-worth of SO42− in the soil. Surface clast exposure ages derived from 10Be are summarized in Table 3 and were consistent with the long soil accumulation history derived from salt content. For the four samples, calculated exposure ages ranged from 841 to 1322 ky and may have slightly increased with distance downslope. The sample from ORH-1 (the lowest excavation, Fig. 4) was especially low in quartz and not enough could be isolated for 10Be analysis. The exposure age calculations are also approximate, primarily because the exposure age calculation used here (Eq. (2)) assumes a simple exposure history for the sample — no burial, no erosion, and no inherited 10Be. The soils in this region had such low density, that

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

165

Table 1 Soil characteristics, salt content, and years of salt by horizon at Oficina Rosario. Sample ID

Horizon

Depth

Soil bulk densitya

Gravel fraction

Particle size of fine fraction Clay fraction

(cm) Soil 4, x= 21.6 ORH-05-4-1 ORH-05-4-2 ORH-05-4-3 ORH-05-4-4 ORH-05-4-5

m, ∇z = 0.14 Byk1 2 Byk2 4 Byk3 7 BCyk 23 CByk 36

Soil 3, x= 34.9 ORH-05-3-1 ORH-05-3-2 ORH-05-3-3 ORH-05-3-4 ORH-05-3-5

m, ∇z = 0.09 Byk1 1 By 3 Byk2 9 BCyk 36 CBy 69

Soil 2, x= 54.7 ORH-05-2-1 ORH-05-2-2 ORH-05-2-3 ORH-05-2-4 ORH-05-2-5

m, ∇z = 0.27 Byk1 1 Byk2 3.5 Byk3 24 Bykm 68 By 75

Soil 18, x = 65.6 m, ∇z = 0.31 ORH-05-18-1 Byk1 1 ORH-05-18-2 Byk2 3.5 ORH-05-18-3 Byk3 8 ORH-05-18-4 Byk4 37 ORH-05-18-5 By 73

Soil 1, x= 87.7 ORH-05-1-1 ORH-05-1-2 ORH-05-1-3 ORH-05-1-4 ORH-05-1-5 ORH-05-1-6 ORH-05-1-7

m, ∇z = 0.36 By 1 Byk1 3 Byk2 13 Bykm1 57 BCyk 76 Bykm2 115 BCy 137

(g cm

Silt fraction

Concentration in fine fraction Sand fraction

−3

)

Years of salt by horizonb

NO3−

Cl−

SO42–

NO3−

Cl−

SO42–

(ppm)

(ppm)

(ppm)

(y)

(y)

(y)

0.8 0.17 0.32 1.43 1.37 Depth-averaged =

0.05 0 0.1 0.8 0.9

0.19 0.18 0.13 NA 0.36 0.29

0.21 0.04 0.08 NA 0.18 0.15

0.60 0.77 0.79 NA 0.46 0.55

159 103 156 1197 123

654 1378 916 2349 1014

6000 318,000 439,500 541,500 540,000 Total =

9 1 5 200 8 230

2800 1300 2200 30,000 5100 42,000

200 2000 6900 45,000 17,000 71,000

1.22 0.16 0.41 1.37 1.39 Depth-averaged =

0 0 0.05 0.8 0.9

0.16 0.18 0.08 0.19 0.23 0.20

0.14 0.03 0.15 0.13 0.21 0.17

0.70 0.78 0.77 0.68 0.56 0.64

204 769 183 134 159

1236 979 927 792 1050

16,800 333,000 400,500 492,000 460,500 Total =

9 9 16 38 28 100

4300 900 6200 17,000 14,000 42,000

400 2000 17,000 66,000 38,000 123,000

1.29 0.16 0.33 1.34 1.3 Depth-averaged =

0.05 0 0 0.04 0

0.12 0.09 0.07 0.14 0.07 0.11

0.18 0.10 0.12 0.10 0.07 0.10

0.70 0.81 0.80 0.76 0.86 0.78

143 140 193 123 964

288 1092 748 864 1445

4500 337,500 319,530 418,500 162,600 Total =

7 2 50 260 330 660

1000 1300 14,000 138,000 37,000 192,000

100 2500 39,000 427,000 27,000 496,000

1.12 0.31 0.55 1.5 1.29 Depth-averaged =

0.05 0 0.05 0.05 0.3

0.14 0.15 0.08 0.12 0.15 0.13

0.19 0.09 0.11 0.08 0.09 0.09

0.67 0.76 0.81 0.79 0.75 0.77

257 145 224 153 145

1310 948 1023 1256 2743

48,000 367,500 370,500 490,500 519,000 Total =

10 4 20 240 180 450

4000 2100 6800 147,000 252,000 412,000

900 5100 16,000 366,000 304,000 692,000

0.89 0.34 0.5 1.16 1.24 1.22 1.34 Depth-averaged =

0 0 0 0.05 0.28 0.08 0.48

0.13 0.12 0.19 0.03 0.06 0.18 0.16 0.11

0.11 0.06 0.09 0.11 0.10 0.31 0.28 0.19

0.76 0.82 0.72 0.87 0.85 0.51 0.56 0.70

578 406 442 404 1177 6730 17,742

968 890 610 1012 905 18,072 10,451

84,000 354,000 408,000 277,500 296,100 221,100 384,000 Total =

20 10 84 740 760 11,000 10,000 23,000

2400 1700 8600 139,000 43,000 2,240,000 454,000 2,890,000

1300 4300 37,000 243,000 91,000 175,000 106,000 656,000

NA = not available, x = distance downslope and ∇z = local slope gradient. a Soil bulk density is calculated for the fine fraction only (b2 mm). b Years of salt = salt concentration × horizon thickness × bulk density × (1 − gravel fraction) /atmospheric input rate.

slight burial has little effect on the calculated production rate (e.g., b 20% decrease for 35 cm burial). Thus, the calculated exposure age may be an overestimate of the time the clasts spent on the soil surface, possibly including some unknown period near, but below, the surface. Previous work showed that bedrock erosion rates in the region were b1 m My−1 (Owen et al., 2010) and an exposed boulder eroded at b0.2 m My−1 (Ewing et al., 2006). If the clasts did experience erosion, 10 Be concentrations would be lower and the calculated exposure age would be an underestimate, but this is likely a small effect. Lastly, the samples could contain 10Be produced prior to their exposure at the surface under different climatic or soil conditions, such as faster erosion rates driven by increased precipitation or a denser soil prior to the accumulation of salt. Given the long time spans considered here, this is also likely to be a small effect as the inherited 10Be signature will have largely decayed and been replaced by one reflecting more recent conditions. The similarity of values across the hillslope suggest a reliable time of exposure of ~10 6 y.

3.3. Soil characteristics and zebra stripes on CH Unlike OR, the soil cover on the hillslopes near Chañaral was thin, patchy, and nearly salt-free. There were few plants and a sparse animal population, including lizards, birds, coyotes, and guanacos (Lama guanicoe, a coastal camelid in the same genus as llamas, Lama glama). Where present, the soil was a 0.5–2 cm-thick layer composed of coarse sand, fine gravels, and fine reddish-brown dust, and which had an abrupt contact with the coarsely fractured bedrock below (Fig. 2f). The bedrock showed almost no evidence of chemical weathering, with only a few reddish rinds detected in thin section (Owen et al., 2010). Small slope parallel stripes which had some similarities with the zebra stripes in the hyperarid zone occurred in a few places on CH (Fig. 2d). They were found only on steep slopes where there was little evidence of animal traffic (guanaco paths were common on the lower portions of the hillslopes where no zebra stripes were observed).

166

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

Fig. 9. Drawings (to scale, re-drawn from photographs by J.O.) of soil profiles along the downslope transect on OR. From upslope to downslope (a) ORH-4, (b) ORH-3, (c) ORH-2, (d) ORH-18, and (e) ORH-1. Soil thickness is on the right axes and soil horizon classifications are on the left axes. Stippled areas are sandy while white areas are predominantly indurated sulfate. Linear stipples on the large upper prism in (e) represent slickensides. Solid lines define prism boundaries.

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

167

Table 2 Deposition rates of atmospheric dust and salt. Silicate dust input rate

NO3− input rate

Cl− input rate

SO42− input rate

Total input rate

(g m−2 y−1)

(g m−2 y−1)

(g m−2 y−1)

(g m−2 y−1)

(g m−2 y−1)

ORH-1 ORH-2 ORH-3 ORH average ORH st. deviation

1.572 1.127 1.509 1.403 0.241

0.260 0.335 0.195 0.263 0.070

0.004 0.003 0.003 0.004 0.000

0.543 0.568 0.552 0.554 0.013

3.942 3.155 3.760 3.619 0.412

Nearby sites YH-1 YH-2 YH-3 YH average YH st. deviation Ewing et al. (2006)

1.162 1.009 1.018 1.063 0.086 2.032

0.637 0.500 0.449 0.529 0.097 0.759

0.077 0.031 0.042 0.050 0.024 0.098

0.958 0.923 0.967 0.949 0.023 1.319

3.980 3.457 3.481 3.639 0.295 4.208

Dust trap ID

“ORH” are from Oficina Rosario. “YH” are from a site 70 km north, near that of Ewing et al. (2006), and are included for comparison.

infiltration. Infiltrometer-based infiltration rates were lowest where near-surface salt cementation was strongest and most continuous (Table 5). Excavating the soils from the infiltrometer site revealed preferential flow paths which followed soil structure. Where the lowest infiltration rate was measured (26.6 cm h−1, test 3), indurated soil was laterally continuous such that only one vertical, narrow (1–3 cm-diameter) flow path connected the upper soil horizons with a less indurated zone. In contrast, where the experiment with the second highest infiltration rate (135 cm h−1, test 2) had been, the soil was wet at the surface ≥30 cm laterally and downslope of the infiltrometer, almost continuously 20-cm-deep below where the surface had become wet, and over 40 cm into an ~5-cm-wide, sand-filled crack between indurated soil prisms. Infiltration rates determined from sprinkling experiments on CH ranged from 3.3 to 25.3 cm h−1, comparable to the rates calculated from the sprinkling experiments at OR, and were also proportional to sprinkling rate (Table 5). Runoff appeared similar to that at OR – cloudy, concentrated in lows, unable to mobilize particles larger than sand – and it was too thin to measure its depth. Exposing the downslope face of the sprinkling area showed uniform wetting through the thin soil and upper 2–5 cm of fractured rock. Below that, preferential flow through rock fractures extended irregularly more than 15 cm into the soil. The highest precipitation rate was able to flush some of the fine silicate material out of larger fractures. The soil was too thin to use an infiltrometer. Attempts to mobilize gravels where the sprinkling experiment had been by rapidly pouring water from a bucket directly on the surface managed to move some small gravels, but most of the water infiltrated and flowed through the rock fractures, suggesting a potentially very high infiltration rate.

They had an arcuate, concave downslope shape with overlapping tips, and the clasts coarsened towards the downslope front, but rarely exceeded 1 cm in diameter. No surface gravels > 10-cm-diameter were observed on the hillslopes. Surface clasts were not varnished (Fig. 2e and f), though some had dark lichen growing on them (Fig. 2d). The clasts sat loosely on top of the dusty soil and none were perched on pedestals. The stripes at CH were similar to those at OR except much smaller, finer grained, and rarer. 3.4. Sprinkling and infiltration experiments Steady-state infiltration measured by sprinkling experiments on OR ranged from 4.2 to 42.6 cm h −1. The short, high precipitation run of 58.7 cm h −1 applied nearly 8 cm of rain total whereas the low precipitation runs applied 1.7 to 3.1 cm of rain (Table 4). Total runoff for the high precipitation rate was ~ 2.8 cm and 0.2–0.5 cm for the low precipitation experiments. The depth of the runoff was too small to measure, even at the highest precipitation rates. Runoff was concentrated in low areas and was cloudy with fine sediment. Given the small area and shallow runoff, no clasts larger than sand were mobilized. During the high precipitation tests, the fine soil in a few locations was scoured 1–2 mm by localized concentrated flow. The wetting front along the downslope edge of the test plot was continuous through the upper 3–4 cm of porous, weakly cemented soil, and was irregular below that. Preferential flow paths extending 20–30 cm into the underlying soil were observed along the boundaries of prisms, particularly where the boundaries were wide (>10 cm) sand-filled cracks. The position of these boundaries did not coincide with zebra stripe front position nor did surface C correlate with underlying soil properties. Infiltration rates calculated from the infiltrometer tests were higher than those calculated by the sprinkling experiments, with values ranging from 26 to 145 cm h−1 (Table 5). The maximum infiltration was large for barren ground, suggesting the plastic edging bounding the study plots in both tests may have disturbed the surface and enabled greater Table 3 Surface clast exposure ages calculated from cosmogenic Sample ID

x (m) Latitude

ORH-05-4-G ORH-05-3-G ORH-05-2-G ORH-05-18-G

21.6 34.9 54.7 65.6

a

−24.3818 −24.3817 −24.3815 −24.3814

10

3.5. Regional distribution Contour-parallel, dark-colored bands indicative of zebra stripes were visible in GoogleEarth on hillslopes within a 400 km-long, 20 to 50 km-wide zone between 21° 46′ S and 25° 29′ S (Fig. 3), an extent confirmed by field observations. The western boundary was

Be concentrations. Be concentrationb (atom g quartz−1) ± (at g quartz−1) Exposure agec (ky) ± (ky)

Longitude

Elevation (masl) Scaling factora

10

−69.93986 −69.93980 −69.93976 −69.93972

1403.1 1408.2 1410.6 1399.2

7.40 10.31 10.34 9.58

2.119 2.116 2.108 2.102

0.15 0.28 0.32 0.25

841.85 1309.35 1322.07 1191.85

21.04 50.04 57.75 42.84

Production rate latitude and elevation scaling factor (Lal, 1991). Measured at the Purdue PRIME AMS lab with the 07KNSTD standard. c Calculated assuming sample thickness = 5 cm, sample density = 2.5 g cm−3, no topographic shielding, 10Be production rate = 5.1 at g quartz−1 y−1, muon contribution of 2.5% at sea level, and attenuation length = 165 g cm−2. b

168

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

Table 4 Sprinkling experiment conditions and results. Site

Test

Oficina Rosario Oficina Rosario Oficina Rosario Chañaral Chañaral Chañaral

1 2 3 1 2 3

Slope

Gravel cover and striping

Application rate

Steady-state runoff

Steady-state infiltration

Duration of sprinkling

Time until runoff

(m m−1)

(%)

(cm h−1)

(cm h−1)

(cm h−1)

(MM:SS)

(MM:SS)

0.21 0.34 0.27 0.38 0.38 0.38

25 no stripes 60 weak stripes 50 strong stripes 80 no stripes 85 weak stripes 80 no stripes

58.7 5.7 5.2 5.7 5.5 52.5

16.1 0.5 0.1 1.3 2.0 28.9

42.6 5.2 4.2 4.1 3.7 25.3

08:00 18:20 36:00 18:00 13:00 03:30

00:17 08:00 04:00 06:00 02:00 00:30

Steady state was inferred to occur when the values of runoff and infiltration rates became fairly stable and is the average of the last 4 to 5 min of the experiment. MM = minutes and SS = seconds.

20 to 40 km inland from the coast, on the east (lee) side of the Coast Range. Zebra stripe occurrence was patchy across the Central Depression (the valley between the Coast Range and the preAndean ranges) and decreased eastward before completely disappearing in the pre-Andean ranges. The southern extent of zebra stripes visible in GoogleEarth was ~80 km north of Chañaral, a latitude at which the soils no longer contained significant amounts of salt. This boundary was well north of CH, but the zebra stripes at CH were too small to be visible in GoogleEarth. The northern boundary was diffuse and did not correspond to any physiographic features, but did lie within the present-day transition zone from predominantly winter rainfall (south of ~22° S) to summer rainfall (north of ~22° S) (Houston, 2006). Zebra striping was not found on every hillslope within the region shown in Fig. 3, or had different characteristics from those described on OR. Some hillslopes simply had almost no surface clasts, while others were almost entirely covered with clasts. Other hillslopes had only a few zebra stripes with a large wavelength. The boundaries between zebra-striped areas and areas with few to no zebra stripes were usually abrupt and strongly correlated with geologic contacts (Fig. 10). The area of the Antofagasta geologic map in which the spatial analysis was performed contains a variety of bedrock types, including various igneous and metasedimentary rock. Hillslopes composed of monzogranite and diorite were generally the least likely to have zebra stripes, while those with bedrock of schist, gabbro, and certain monzogranites were the most likely (Table 6). Based on the unit descriptions accompanying the map (González and Niemeyer, 2005), no bedrock characteristics, such as composition, grain size, or foliation were predictive of zebra stripe occurrence, nor was there any correlation with age, dike concentration, or degree of faulting. For example, two components of the Cretaceous Piedra Grande unit were next to each other and nearly identical in age and geologic history. The monzogranite component, Kigr(c), was 82% zebra striped whereas the granodiorite component, Kigr(a), had b1% striping. This suggests that monzogranites produced gravel clasts which were more likely to form zebra stripes. When Kigr(c) was compared with another monzogranite, Jsg(c), however, they were 82% and 4% zebra-striped, respectively, despite no distinct differences between

them besides the age of the bedrock. Without additional fieldwork examining bedrock mineralogy and texture, clast size and shape, soil properties, and topography of hillslopes within the region, we cannot explain the above observations. 4. Discussion 4.1. Zebra stripe formation mechanism and hydraulics We proposed three processes which could produce zebra stripes: salt shrink-swell, seismic shaking, and overland flow. If soil shrink-swell was the cause, we expected 1) zebra stripes present on any hillslopes where there was adequate salt in the soil to create polygonal structure (and absent where soil was salt-poor), 2) zebra stripe location and scale correlated to underlying soil structure, and 3) poor sorting of gravels within bands. Our observations showed that salt was not required for zebra stripe formation, at least on a relatively small scale, as at CH, nor is zebra stripe occurrence correlated to soil prism boundaries. Additionally, the presence of salt-rich soil with prismatic structure did not always produce zebra stripes; salt-mantled hillslopes occurred well north of the northern boundary in Fig. 3.

Jsg(b)

Dst(b)

Table 5 Infiltrometer experiment conditions and results. Test

Subsurface conditions

Surface gravel

Slope (m m−1)

Average infiltration rate (cm h−1)

2

Generally soft, hard gypsum at 24 cm deep Hard gypsum Half sandy, half hard gypsum Sandy gypsum to 26 cm deep, then hard

Mostly absent

0.36

134.3

Very gravelly Very gravelly

0.31 0.29

26.6 30.6

Absent

0.27

145.4

3 5 6

Tests 1 and 4 are not reported due to breaches in the seal between the infiltrometer and the soil.

100 m Fig. 10. GoogleEarth image showing a sharp boundary between hillslopes that have zebra stripes and those that do not, located near 23.838° S, 70.129° W. Upper left portion labeled Jsg(b) is monzogranite and has no zebra stripes while the lower right labeled Dst(b) is slate and schist and has zebra stripes. The contact between Jsg(b) and Dst(b) is marked with a dashed white line. The dark material along the right border is Quaternary alluvium.

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

169

Table 6 Geology of zebra striped areas. Geologic data from González and Niemeyer (2005). Area was calculated in ArcMap v.3 (ESRI). Symbol Major rock units

Mineral grain size Rock fabric or other features

Jiqm Jmg Jsg(f)

Medium to coarse Locally foliated Medium to coarse Locally foliated Medium to coarse Locally foliated

Jsg(b) Kigr(a) Jcd Mg Jsg(c) Jsmv Jsg(d) Jsg(e) Jln MPla Jsg(a) Dst(a) Trqp Dst(b) Kigr(c) Kist

Metamorphosed diorite Diorite and granodiorite Monzogranite, granodiorite, quartz-rich diorite Monzogranite and syenogranite Granodiorite Monzodiorite and diorite Gravels Monzogranite and granodiorite Tonalite Syenogranite and monzodiorite Tonalite and granodiorite Andesite, some basalt and sandstone Gravels Monzogranite, syenogranite, and tonalite Quartz-rich sandstone and shale Tonalite Slate and schist Monzogranite Gabbro

Medium to coarse Medium Fine to medium Various Medium Fine to medium Coarse Fine to medium Various Various Medium to coarse Fine Medium to coarse Mostly fine Medium to coarse Fine to medium

Hill area Zebra area % Zebra coverage on hillsa (km2) (km2) 39.2 46.3 32.4

Phaneritic 74.6 Spherical inclusions of quartz diorite 17.7 Phaneritic 52.4 Well-rounded clasts 49.5 Speckled 14.9 Phaneritic 4.3 Many veins of diorite and rhyolite 5.0 Many diorite veins 6.0 Variety of textures (porphyretic to breciated) 94.3 Partially cemented by sulfate and halite 118.2 Phaneritic 76.5 Slightly metamorphosed 55.9 Marked foliation 101.5 Moderately metamorphosed 109.9 Locally foliated 51.6 Sheared texture 5.4

0 0 0

0 0 0

0.2 0.1 1.0 1.1 0.6 0.4 0.5 0.7 25.3 34.6 34.7 28.6 70.4 88.7 42.4 4.7

0.3 0.4 1.9 2.2 4.2 10.1 10.7 12.2 26.8 29.3 45.4 51.2 69.4 80.7 82.2 87.0

Geologic units are sorted by increasing zebra stripe coverage. a Calculated as = 100 × (zebra area ÷ hill area).

If zebra stripes were produced by seismic shaking, we expected 1) zebra stripes present in other seismically-active regions and correlated with tectonic boundaries, 2) more zebra stripes on the tops of hills where seismic energy is greatest due to topographic focusing (Clark, 1972; Bard, 1982; Bouchon and Barker, 1996; Bouchon et al., 1996), 3) gravels coarsening downslope within each zebra stripe (through some unknown version of the reverse brazil nut effect), and 4) zebra stripes associated with microtopography and with evidence of soil slumping in the soil profile. Our observations showed that none of these were true. The zone of zebra stripe occurrence (21–25° S) lay completely within the central subduction zone (15–27° S) of the Nazca tectonic plate and the boundaries between the subduction zones of the Nazca plate have likely been constant for >20 My (Jordan et al., 1983). Thus, zebra stripe occurrence was not bounded or affected by current or past tectonic regimes. Zebra stripes were common at some distance below the crest of the hillslopes and not associated with microtopography. Lastly, there was no evidence of earthquake disruption of the underlying soil — neat slickenslides, laminated material filling cracks between salt-cemented prisms, and the prisms themselves must have formed by soil shrink-swell and not seismic shaking. Our results overwhelmingly suggest that zebra stripes were produced by unconcentrated overland flow. We propose that zebra stripes are an emergent, self-stabilizing form produced by shallow sheetflow and the interaction of coarse grains traveling over a smooth bed. The grain interactions occur due to: 1) clasts of similar size shedding wakes that halt nearby traveling grains (e.g., Brayshaw et al., 1983; Malmaeus and Hassan, 2002), 2) coarser clasts tending to travel faster than smaller ones on smooth beds, resulting in a coarse front in which they interact (e.g., Whiting et al., 1988), and 3) coarse clast deposition creating an elevated frictional field that progressively stops finer clasts arriving from upslope. This stoppage would be expected to spread laterally as coarse particles are added to the edges in a process similar to the formation of stone lines observed in gravel streams (Tribe and Church, 1999; Malmaeus and Hassan, 2002). This hypothesis requires several key components for zebra stripe development. First, the initial state must be one of dispersed coarse clasts of sufficient abundance to interact and eventually create the clumped formations leading to stripes, but not so concentrated that flow velocities are too low and critical shear stresses are too high to prevent

mobilization (e.g., Bunte and Poesen, 1994). Second, the clasts must be large enough that they are marginally transportable by sheetwash. Smaller clasts will remain moving until stopped by larger ones and cannot initiate the front. Clasts that are too large will not participate. Third, a smooth surface must exist under the clasts initially and remain undissected and smooth as the clasts are arranged into stripes. The critical initial clast concentration may be approximated by the length-averaged C on OR of ~ 50%. In a modeling study of bed sorting patterns in small streams in British Columbia (Tribe and Church, 1999), C ≤ 30% created inconsistent patchiness of the stones and longitudinal stone lines, whereas C ≥ 35% created distinct transverse clusters and lines (though not fully sorted features equivalent to zebra stripes). The inferred concentration-dependence of stone line formation in streams may explain why some hillslopes do not have zebra stripes. A survey of C on hillslopes in the region in which zebra striping occurs, on hillslopes with and without zebra stripes, could constrain the transition. The variation in availability may correlate with bedrock characteristics. Rock types with different strengths (due to mineralogy, fabric, etc.) fracture differently under regional and local stresses such that the bedrock at the soil–bedrock interface is predisposed to release rock fragments into the soil with a certain size distribution. Rock fragment chemistry and fabric will also affect the persistence of clasts of certain sizes once they are incorporated into the soil (Brantley et al., 2011). The relationships between bedrock, erosion processes, and rock fragment production, however, are still poorly understood and an active area of study (Marshall et al., 2010). Nevertheless, geological control on clast properties likely explains the correlation between underlying geology and the distinct boundaries observed between zebra-striped and non-zebra striped areas. The lack of an accumulation of gravels at the base of OR or evidence that colluvial gravels were buried there suggests 1) zebra stripe formation inhibits gravel transport off the hillslope, and 2) that surface gravels traveled relatively short distances before becoming trapped in bands. Infiltration rates may also change towards the base of the hillslope as soil thickens and becomes sandier, but we did not measure at these lower locations. Indeed, if the surface clasts were being transported downslope and replaced with gravel from the underlying soil, we would likely have observed a clear increase in 10Be exposure age with distance

170

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

downslope. Instead, exposure ages show, at most, a weak trend of increasing age for three of the samples (Table 3). This indicates that the hillslope soil was likely behaving more like a desert pavement, where surface gravels were riding the accumulation of salt with little downslope transport. Zebra stripe wavelength is likely scaled in part by the supply of sediment which is controlled by both the initial availability of clasts upslope and the subsequent decline in sediment due to the formation of coarse fronts. This supply dependency should lead to some central tendency in wavelength. The cessation of sediment transport downslope of the coarse fronts will lead to scour of the smooth surface. The latter constraints suggest a significant role played by the accumulation of salt. Clasts up to a certain size will float on the accumulating salt, while salt weathering shatters larger rocks. The two effects reduce the likelihood of clasts too large to be mobilized occurring at the surface. The type of salt accumulated is also important — sulfate creates a smooth surface whereas halite roughens it. Salt weakly cements the surface such that even if it has a high sand content, there is not a significant source of sand which would increase the mobility of the surface clasts (e.g., Whiting et al., 1988; Venditti et al., 2010). This creates a smooth, resistant surface which nevertheless has a relatively high infiltration rate. The smooth surface enables the initial mobilization of relatively large clasts while high infiltration lowers the likelihood of overland flow depths great enough to mobilize the surface clasts off the hillslope. Thus, it may be exceptionally rare rainfall events that create the zebra stripes. Salt accumulation requires not only hyperaridity but a source of salt as well, which in the Atacama is a combination of salt-carrying fog and eolian transport of salar dust, with the importance of fog decreasing inland (Rech et al., 2003). Disturbance of the zebra stripes occurred in a few ways. First, small steps beneath the zebra stripe front appeared to undermine and destabilize clasts, enabling the remobilization of the remaining clasts. These steps may have formed by plunge-pool type scour or rainsplash, or some combination of both. They appeared similar to the pedestals beneath isolated clasts, which themselves were similar to rainsplash-formed pedestals described by Poesen et al. (1994). Second, rills cut directly across some zebra stripes (Fig. 1) and their concentrated flow was adequate to remove the underlying fine, uncemented soil, undermining the surface clasts and mobilizing some of them. Rill formation implies overland flow extended across the entire hillslope, from crest to footslope, and yet interill overland flow was not enough to completely disturb the zebra stripes. In fact, rills may have formed as a result of the formation of the zebra stripes which cleared parts of the hillslope of clasts and left them open to incision. Soil stirring by vegetation and animals appeared to limit the persistence of zebra stripes at CH. The preservation of the relatively large zebra stripes to the north requires the absence of life, thus, these features are distinctive on an otherwise biotic world. 4.2. Precipitation sets regional distribution of zebra stripes We described several characteristics of soil, rock, and overland flow conducive to zebra stripe formation and found that changes in these characteristics were correlated to the boundaries of zebra stripe occurrence shown in Fig. 3. The western boundary coincided with changes in MAP and soil cover. Between 20–27° S, the Coast Range rises from 1500 masl to more than 2300 masl, creating a rain shadow. The current rain shadow is small: Antofagasta, located on the coast at 23.6° S, has an MAP of 3.5 mm (based on 687 months of data from 1931 to 1989, www.worldclimate.com) whereas YH, located ~50 km inland at 24.1° S and featuring well-developed zebra stripes on hillslopes, received one precipitation event totaling 2.3 mm during 4 years of continuous monitoring (McKay et al., 2003). The Coast Range uplifted in the Oligocene (Hartley et al., 2000) and the accumulation of Miocene sediments on its eastern flanks indicate that a

rain shadow has been in place for at least that long. On the wetter western side of the Coast Range where no zebra stripes occurred, hillslopes had little to no soil mantle and no evidence of recent salt accumulation. Relatively small rainfall events (1.6–4.2 cm total precipitation) have triggered debris flows in Antofagasta (Vargas et al., 2000), suggesting the hillslopes have low infiltration rates in that area, and may be subject to more frequent overland flow events capable of mobilizing large clasts than hillslopes on the eastern side of the range. The northern boundary of the zebra-striped area lay within a zone where MAP changed little and hillslopes were consistently mantled with salt-rich soil. This boundary corresponds to a shift in the seasonality of precipitation from predominantly winter rainfall (south of ~ 22° S) to summer rainfall (north of ~ 22° S) (Houston, 2006). Assuming that the current estimated MAP patterns can be applied to the past, this suggests that winter and summer rainfall intensity and duration may be sufficiently different that, for the same salt-rich soil mantle, zebra stripes may be produced only by winter rainfall. The precipitation data available for the region were not of fine enough resolution or long enough duration to determine if this seasonality effect was due to differences in rainfall intensity or timing. The southern boundary corresponded to an increase in precipitation, an increase in biotic disturbance, and a decrease of soil salt concentration to zero. Though the zebra stripes were disturbed by animal trampling on CH, the relatively small gravels that comprise them require smaller precipitation events to be mobilized, such that the zebra stripes were easier to re-form. Whereas the zebra stripes on OR were stable, the zebra stripes on CH may be relatively transient. The presence of small zebra stripes at CH despite biotic disturbance and in the absence of salt-rich soil suggests 1) a process similar to that which formed the large zebra stripes to the north is presently active, and 2) salt is not necessary for small zebra stripe formation in spite of a relatively high infiltration rate. This site may be one of the few locations to observe zebra stripe formation under current conditions. Lastly, the eastern boundary in the pre-Andean ranges was the least distinct but corresponded to two, albeit poorly constrained, changes in precipitation. First, precipitation increased with increasing elevation, and second, 69° W marked the approximate location of the transition from winter rains (west of ~ 69° W) to summer rains (east of ~ 69° W) (Houston, 2006). These precipitation changes were accompanied by a gradual west-to-east transition from salt-rich soils to gravelly, salt-poor soils. Thus, as along the other boundaries, changes in water availability, soil infiltration, and surface particle size likely drove the transition from zebra-striped to unstriped hillslopes. The regional distribution of zebra stripes suggests that they occurred only where MAP is b 4 mm, precipitation fell primarily in winter, and the soils contained large amounts of salt or silicate dust (in the case of CH). All three conditions must have been met for zebra stripe formation. We have proposed reasons why low MAP and large salt concentrations are required, but we cannot explain the mechanism by which precipitation seasonality affects zebra stripe occurrence.

4.3. Chronology of zebra stripe formation at OR We have proposed that the zebra stripes are self-stabilizing features that, once formed, may persist for long time periods in the absence of biotic disturbance. How long is long? If the surface clasts have been floating on accumulating salt and dust as a desert pavement, then the clast 10Be concentrations suggest that they were at the surface for ~ 10 6 y, while soil sulfate concentrations suggest that these clasts were separated from the bedrock surface for at least 10 4–10 5 y, and from the top of the prismatic horizons for > 10 3 y (Table 1). When the zebra stripes formed within that history is uncertain. Here we present a hypothesis for the history of soil and zebra stripe formation on OR.

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

To begin, we look to the south where modern precipitation levels and hillslope soil characteristics were likely similar to those at OR prior to the onset of hyperaridity 2 Mya (Amundson et al., 2012). Coastal semiarid hillslopes (MAP = 100 mm) near 29°S were mantled with chemically-weathered soil and support desert vegetation, whereas CH and OR lacked chemically-altered soil or bedrock (Owen et al., 2010). Thus, increasing aridity appears to have stripped OR of this material prior to the accumulation of salt. This process is similar to the one proposed by Oberlander (1972) to explain the formation of boulder-covered slopes in the Mojave Desert, California. Beginning ~ 10 6 ya, precipitation was sufficiently low (b10 mm) for salt accumulation. Salt crystallization in bedrock fractures pried out rock fragments. At least some of these clasts would have floated on the surface, creating a surface layer of relatively coarse, unsorted clasts. Wetting and drying produced the coarse prismatic soil structure, concentrated clasts in parts of the prisms (Fig. 9), possibly by mechanisms similar to Antarctic ice prisms described by Sletten et al., 2003), and decreased the overall infiltration rate of the soil as the salt was concentrated. The diffusive effect of shrinking and swelling likely kept the surface clasts scattered and unsorted. The lack of gravels in the cracks between prisms suggests that few surface clasts were recycled back into the soil. We propose that the zebra stripes observed today may post-date the formation of the largest salt prisms and pre-date the accumulation of the vesicular horizon, such that they stabilized 2–6 kya at the latest, and likely earlier. Indeed, the consistency of the ages derived from the upper 3–4 cm in the five excavations (Table 1) indicates a relatively stable and uniform accumulation period of > 10 3 y. With the soil surface stabilized, the fine material and zebra stripes could form and persist undisturbed. If the zebra stripes have floated intact on the accumulating soil for thousands of years, then the process of soil accumulation is somewhat different from the formation hypothesis of McFadden et al. (1998). In their model, dust is trapped between the surface clasts and is worked underneath them to produce a gravel-free soil beneath the desert pavement. On a zebra-striped hill there are large areas with sparse gravel cover which would not be expected to accumulate soil. Small cracks observed in the surface crust, however, could act as traps for dust and salt. Additionally, fog adds salt through wet deposition which could contribute to surface crusting and minimize wind erosion. Thus, the continued accumulation of soil despite patchy surface gravel cover is possible. The salt accumulation time calculations upon which our chronology is based are approximate for several reasons. Though downslope transport of soil is slow by shrink-swell creep and/or overland flow, it does redistribute sulfate on the hillslope. Thus, the accumulation times will be underestimates for soils at upslope positions, and may be overestimates for soils at downslope positions. In summary, zebra stripe formation and persistence appear to require specific conditions: 1) surface clasts to some size must be present at the correct concentration, 2) the soil surface must be smooth and resistant to erosion yet have a high infiltration rate, 3) sheetwash must be adequate to mobilize most surface clasts without completely removing them from the hillslope or forming rills, and 4) disturbance by animals or vegetation must be minimal. In the Atacama, these conditions have taken ~106 y to develop, requiring prolonged hyperaridity and an almost total lack of life. Though some of these conditions are met in other parts of the world, the hyperarid region of the Atacama Desert is the only place identified so far where all occur. 5. Conclusions Zebra stripes are lateral bands of sorted, varnished gravels found on hillslopes in the Atacama Desert. Their gravel sorting, wavelength, and regional distribution indicate that they formed through the mobilization of surface clasts by sheetwash. Their formation and persistence required a distinct combination of climate and soil properties

171

including the necessary concentration and size of surface clasts for mobilization and self-stabilization, a high infiltration rate to limit sheetwash depth, a surface resistant to erosion to reduce rilling, and the absence of surface disturbance to preserve them. These conditions took ~ 10 6 y to develop in this hyperarid region and are predominantly the result of the accumulation of salt deposited by fog and eolian transport of salar dust. Similar, but much smaller features have been observed to the south near Chañaral, however, zebra stripes have not been reported in any other deserts on Earth. Thus, the climatic, geographic, and geologic history of the Atacama Desert may have produced unique conditions conducive to zebra stripe formation. This work has identified several factors important to zebra stripe formation, but more work is required to refine our understanding of the role of lithology, precipitation, and topography in producing and shaping zebra stripes. Nevertheless, zebra stripes provide long-lasting evidence of sheetwash in a place where sheetwash is presently extremely rare. Their persistence requires the absence of biotic processes that would disturb the land surface. If they were observed in other deserts or on terrestrial planets, they offer a tantalizing window into a distinct set of surface processes and environmental conditions. Acknowledgments Thanks to Sarah Reed, Brad Sutter, Robert Finkel, and Simona Balan for help with fieldwork; Peter Nelson and Mike Lamb for helpful discussions of gravel transport; and Marc Caffee and Robert Finkel for cosmogenic radionuclide analysis. This work was partially supported by a National Science Foundation Geobiology and Low Temperature Geochemistry grant to R.A. and K.N., and a NASA Graduate Student Research Program fellowship to J.O. Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version, at http://dx.doi.org/10.1016/j.geomorph.2012.11. 006. These data include Google maps of the most important areas described in this article. References Abrahams, A.D., Parsons, A.J., Luk, S.-H., 1986. Resistance to overland-flow on desert hillslopes. Journal of Hydrology 88, 343–363. Abrahams, A.D., Parsons, A.J., Hirsch, P., 1992. Field and laboratory studies of resistance to interrill overland flow on semiarid hillslopes, southern Arizona. In: Parsons, A.J., Abrahams, A.D. (Eds.), Overland Flow: Hydraulics and Erosion Mechanics. UCL Press, London, pp. 1–23. Adelsberger, K.A., Smith, J.R., 2009. Desert pavement development and landscape stability on the Eastern Libyan Plateau, Egypt. Geomorphology 107, 178–194. Al-Farraj, A., Harvey, A.M., 2000. Desert pavement characteristics on wadi terrace and alluvial fan surfaces: Wadi Al-Bih, U.A.E. and Oman. Geomorphology 35, 279–297. Alpers, C.N., Brimhall, G.H., 1988. Middle Miocene climatic change in the Atacama Desert, Northern Chile — evidence from supergene mineralization at La Escondida. Geological Society of America Bulletin 100, 1640–1656. Ammann, C., Jenny, B., Kammer, K., Messerli, B., 2001. Late Quaternary Glacier response to humidity changes in the arid Andes of Chile (18 ± 29°S). Palaeogeography, Palaeoclimatology, Palaeoecology 172, 313–326. Amundson, R., Dietrich, W., Bellugi, D., Ewing, S., Nishiizumi, K., Chong, G., Owen, J., Finkel, R., Heimsath, A., Stewart, B., Caffee, M., 2012. Geomorphologic evidence for the late Pliocene onset of hyperaridity in the Atacama Desert. Geological Society of America Bulletin 124, 1048–1070. http://dx.doi.org/10.1130/B30445.1. Anderson, K., Wells, S., Graham, R., 2002. Pedogenesis of vesicular horizons, Cima Volcanic Field, Mojave Desert, California. Soil Science Society of America Journal 66, 878–887. Bard, P.Y., 1982. Diffracted waves and displacement field over two-dimensional elevated topographies. Geophysical Journal of the Royal Astronomical Society 71, 731–760. Beaty, C., 1983. Tiger striping; a curious form of surficial patterning in the Atacama Desert, N. Chile. GSA Annual Meeting Abstracts with Programs, 15, p. 387. Benedict, J.B., 1976. Frost creep and gelifluction features: a review. Quaternary Research 6, 55–76. Bodine, M.W., Fernalld, T.H., 1973. EDTA dissolution of gypsum, anhydrite, and Ca–Mg carbonates. Journal of Sedimentary Petrology 43, 1152–1156. Bouchon, M., Barker, J.S., 1996. Seismic response of a hill: the example of Tarzana, California. Bulletin of the Seismological Society of America 86, 66–72.

172

J.J. Owen et al. / Geomorphology 182 (2013) 157–172

Bouchon, M., Schultz, C.A., Toksoz, M.N., 1996. Effect of three-dimensional topography on seismic motion. Journal of Geophysical Research, Solid Earth 101 (B3), 5835–5846. Bouwer, H., 1986. Intake rate: cylinder infiltrometer, Methods of Soil Analysis, Part 1. Physical and Mineralogical Methods, 2nd ed. : Agronomy Monograph, 9. American Society of Agronomy — Soil Science Society of America, pp. 825–844. Brantley, S.L., Buss, H., Lebedeva, M., Fletcher, R.C., Ma, L., 2011. Investigating the complex interface where bedrock transforms to regolith. Applied Geochemistry 26, S12–S15. Brayshaw, A.C., Frostick, L.E., Reid, I., 1983. The hydrodynamics of particle clusters and sediment entrainment in coarse alluvial channels. Sedimentology 30, 137–143. Bunte, K., Poesen, J., 1994. Effects of rock fragment size and cover on overland-flow hydraulics, local turbulence and sediment yield on an erodible soil surface. Earth Surface Processes and Landforms 19, 115–135. Castle, R.O., Youd, T.L., 1972. Engineering geology. The Borrego Mountain Earthquake of April 9, 1968. : Geological Survey Professional Paper, 787. USGS, pp. 158–174. Clark, M.M., 1972. Collapse fissures along the Coyote Creek fault. The Borrego Mountain Earthquake of April 9, 1968. : Geological Survey Professional Paper, 787. USGS, pp. 190–207. Clarke, J.D.A., 2006. Antiquity of aridity in the Chilean Atacama Desert. Geomorphology 73, 101–114. Dunai, T.J., Gonzales Lopez, G.A., Juez-Larre, J., 2005. Oligocene–Miocene age of aridity in the Atacama Desert revealed by exposure dating of erosion-sensitive landforms. Geology 33, 321–324. Dunne, T., Dietrich, W.E., Brunengo, M.J., 1980. Simple, portable equipment for erosion experiments under artificial rainfall. Journal of Agricultural Engineering Research 25, 161–168. Ericksen, G.E., 1981. Geology and origin of the Chilean nitrate deposits. Geological Society Professional Paper, 1188. United States Government Printing Office, Washington, DC, p. 37. Ewing, S.A., Sutter, B., Owen, J., Nishiizumi, K., Sharp, W., Cliff, S.S., Perry, K., Dietrich, W., McKay, C.P., Amundson, R., 2006. A threshold in soil formation at Earth's arid–hyperarid transition. Geochimica et Cosmochimica Acta 70, 5293–5322. Ewing, S.A., Yang, W., DePaolo, D.J., Michalski, G., Kendall, C., Stewart, B.W., Thiemens, M., Amundson, R., 2008. Non-biological fractionation of stable Ca isotopes in soils of the Atacama Desert, Chile. Geochimica et Cosmochimica Acta 72, 1096–1110. González, G., Niemeyer, H, 2005. Cartas Antofagasta y Punta Tetas, Región de Antofagasta. Servicio Nacional de Geología y Minería, Geologic Map of Chile No. 89, 1:100 000, 1 sheet, 35 p. Haff, P.K., 2005. Response of desert pavement to seismic shaking, Hector Mine earthquake, California, 1999. Journal of Geophysical Research, Earth Surface 110 (F02006). http:// dx.doi.org/10.1029/2003JF000054. Hartley, A.J., Chong, G., 2002. Late Pliocene age for the Atacama Desert: implications for the desertification of western South America. Geology 30, 43–46. Hartley, A.J., May, G., Chong, G., Turner, P., Kape, S.J., Jolley, E.J., 2000. Development of a continental forearc: a Cenozoic example from the Central Andes, northern Chile. Geology 28, 331–334. Heimsath, A.M., Dietrich, W.E., Nishiizumi, K., Finkel, R.C., 1999. Cosmogenic nuclides, topography, and the spatial variation of soil depth. Geomorphology 27, 151–172. Houston, J., 2006. Variability of precipitation in the Atacama Desert: its causes and hydrological impact. International Journal of Climatology 26, 2181–2198. Jordan, T.E., Isacks, B.L., Allmendinger, R.W., Brewer, J.A., Ramos, V.A., Ando, C.J., 1983. Andean tectonics related to geometry of subducted Nazca plate. Geological Society of America Bulletin 94, 341–361. Keefer, D.K., Moseley, M.E., 2004. Southern Peru desert shattered by the great 2001 earthquake: implications for paleoseismic and paleo-El Nino-Southern Oscillation records. Proceedings of the National Academy of Sciences of the United States of America 101, 10878–10883. Kober, F., Ivy-Ochs, S., Schlunegger, F., Baur, H., Kubik, P., Wieler, R., 2007. Denudation rates and a topography-driven rainfall threshold in northern Chile: multiple cosmogenic nuclide data and sediment yield budgets. Geomorphology 83, 97–120. Kohl, C.P., Nishiizumi, K., 1992. Chemical isolation of quartz for measurement of in situproduced cosmogenic nuclides. Geochimica et Cosmochimica Acta 56, 3583–3587. Kuhn, N.J., Yair, A., Grubin, M.K., 2004. Spatial distribution of surface properties, runoff generation and landscape development in the Zin Valley Badlands, northern Negev, Israel. Earth Surface Processes and Landforms 29, 1417–1430. Lal, D., 1991. Cosmic-ray labeling of erosion surfaces — in situ nuclide production-rates and erosion models. Earth and Planetary Science Letters 104, 424–439. Loveless, J.P., Hoke, G.D., Allmendinger, R.W., González, G., Isacks, B.L., Carrizo, D., 2005. Pervasive cracking of the northern Chilean Coastal Cordillera: new evidence for forearc extension. Geology 33, 973–976. Loveless, J.P., Allmendinger, R.W., Pritchard, M.E., Garroway, J.L., Gonzalez, G., 2009. Surface cracks record long-term seismic segmentation of the Andean margin. Geology 37, 23–26. Malmaeus, J.M., Hassan, M.A., 2002. Simulation of individual particle movement in a gravel. Earth Surface Processes and Landforms 27, 81–97. Marinovic, N., Smoje, I., Maksaev, V., Hervé, M., Mpodozis, C., 1992. Hoja Aguas Blancas, Región de Antofagasta. Servicio Nacional de Geología y Minería, Geologic map of Chile No. 70, 1:100 000, 1 sheet, 150 p. Marshall, J.A., Sklar, L.S., Riebe, C.S., 2010. Effects of elevation, lithology, climate, and erosion on hillslope rock fragment distribution and abundance in diverse settings. Geochimica et Cosmochimica Acta 74 (Supplement, A670). http://dx.doi.org/10.1016/j.gca.2010. 04.039. Matmon, A., Simhai, O., Amit, R., Haviv, I., Porat, N., McDonald, E., Benedetti, L., Finkel, R.C., 2009. Desert pavement-coated surfaces in extreme deserts present the longest-lived landforms on Earth. Geological Society of America Bulletin 121, 688–697. McFadden, L.D., Wells, S.G., Jercinovich, M.J., 1987. Influences of eolian and pedogenic processes on the origin and evolution of desert pavements. Geology 15, 504–508.

McFadden, L.D., McDonald, E.V., Wells, S.G., Anderson, K., Quade, J., Forman, S.L., 1998. The vesicular layer and carbonate collars of desert soils and pavements: formation, age and relation to climate change. Geomorphology 24, 101–145. McKay, C.P., Friedmann, E.I., Gómez-Silva, B., Cáceres-Villanueva, L., Andersen, D.T., 2003. Temperature and moisture conditions for life in the of observations including the El Niño of 1997–1998. Astrobiology 3, 393–406. Michael, A.J., Ross, S.L., Stenner, H.D., 2002. Displaced rocks, strong motion, and the mechanics of shallow faulting associated with the 1999 Hector Mine, California, earthquake. Bulletin of the Seismological Society of America 92, 1561–1569. Molnar, P., Anderson, R.S., Anderson, S.P., 2007. Tectonics, fracturing of rock, and erosion. Journal of Geophysical Research 112, F03014. Nishiizumi, K., Caffee, M.W., Finkel, R.C., Brimhall, G., Mote, T., 2005. Remnants of a fossil alluvial fan landscape of Miocene age in the Atacama Desert of northern Chile using cosmogenic nuclide exposure age dating. Earth and Planetary Science Letters 237, 499–507. Oberlander, T.M., 1972. Morphogenesis of granitic boulder slopes in Mojave Desert, California. Journal of Geology 80, 1–20. Ortlieb, J., 1995. Eventos El Niño y episodios lluviosos en el Desierto de Atacama. El registro de los últimos dos siglos. Bulletin de l'Institut Français d'Etudes Andines 24, 519–537. Owen, J.J., Amundson, R., Dietrich, W.E., Nishiizumi, K., Sutter, B., Chong, G., 2010. The sensitivity of hillslope bedrock erosion to precipitation. Earth Surface Processes and Landforms. http://dx.doi.org/10.1002/esp. 2083. Poesen, J.W., Torri, D., Bunte, K., 1994. Effects of rock fragments on soil erosion by water at different spatial scales: a review. Catena 23, 141–166. Rech, J.A., Quade, J., Hart, W.S., 2003. Isotopic evidence for the source of Ca and S in soil gypsum, anhydrite, and calcite in the Atacama Desert, Chile. Geochimica et Cosmochimica Acta 67, 575–586. Rech, J.A., Currie, B.S., Michalski, G., Cowan, A.M., 2006. Neogene climate change and uplift in the Atacama Desert, Chile. Geology 34, 761–764. Reheis, M.C., Kihl, R., 1995. Dust deposition in southern Nevada and California, 1984–1989 — relations to climate, source area, and source lithology. Journal of Geophysical Research—Atmospheres 100 (D5), 8893–8918. Schoenenberger, P.J., Wysocki, D.A., Benham, E.C., Broderson, W.D. (Eds.), 2002. Field Book for Describing and Sampling Soils, Version 2.0. Natural Resources Conservation Service, National Soil Survey Center, Lincoln, NE. Schroter, M., Ulrich, S., Kreft, J., Swift, J.B., Swinney, H.L., 2006. Mechanisms in the size segregation of a binary granular mixture. Physical Review E 74, 011307. http:// dx.doi.org/10.1103/PhysRevE.74.011307. Sillitoe, R.H., McKee, E.H., 1996. Age of supergene oxidation and enrichment in the Chilean porphyry copper province. Economic Geology and the Bulletin of the Society of Economic Geologists 91, 164–179. Sletten, R.S., Hallet, B., Fletcher, R.C., 2003. Resurfacing time of terrestrial surfaces by the formation and maturation of polygonal patterned ground. Journal of Geophysical Research 108, 8044. Soil Survey Staff, 1999. Soil Taxonomy, 2nd ed. : USDA Agriculture Handbook, 436. Sylvester, A.G., Burmeister, K.C., Wise, W.S., 2002. Faulting and effects of associated shaking at Pisgah Crater Volcano caused by the 16 October 1999 Hector Mine Earthquake (Mw 7.1), Central Mojave Desert, California. Bulletin of the Seismological Society of America 92, 1333–1340. Tribe, S., Church, M., 1999. Simulations of cobble structure on a gravel streambed. Water Resources Research 35, 311–318. Van der Wateren, F.M., Dunai, T.J., 2001. Late Neogene passive margin denudation history — cosmogenic isotope measurements from the central Namib desert. Global and Planetary Change 30, 271–307. Vargas, G., Ortlieb, L., Rutllant, J., 2000. Aluviones históricos en Antofagasta y su relación con eventos El Niño/Oscilación del Sur. Revista Geológica de Chile 27, 157–176. Vargas, G., Rutllant, J., Ortlieb, L., 2006. ENSO tropical–extratropical climate teleconnections and mechanisms for Holocene debris flows along the hyperarid coast of western South America (17°–24°S). Earth and Planetary Science Letters 249, 467–483. Venditti, J.G., Dietrich, W.E., Nelson, P.A., Wydzga, M.A., Fadde, J., Sklar, L., 2010. Effect of sediment pulse grain size on sediment transport rates and bed mobility in gravel bed rivers. Journal of Geophysical Research 115, F03039. http://dx.doi.org/ 10.1029/2009JF001418. Wells, S.G., McFadden, L.D., Poths, J., Olinger, C.T., 1995. Cosmogenic 3He surfaceexposure dating of stone pavements — implications for landscape evolution in deserts. Geology 23, 613–616. Whiting, P.J., Dietrich, W.E., Leopold, L.B., Drake, T.G., Shreve, R.L., 1988. Bedload sheets in heterogeneous sediment. Geology 16, 105–108. Wilcox, B.P., Wood, M.K., Tromble, J.T., Ward, T.J., 1986. A hand-portable single nozzle rainfall simulator designed for use on steep slopes. Journal of Range Management 39, 375–377. Wilcox, B.P., Newman, B.D., Brandes, D., Davenport, D.W., Reid, K., 1997. Runoff from a semiarid ponderosa pine hillslope in New Mexico. Water Resources Research 33, 2301–2314. Wood, Y.A., Graham, R.C., Wells, S.G., 2005. Surface control of desert pavement pedologic process and landscape function, Cima Volcanic field, Mojave Desert, California. Catena 59, 205–230. Yair, A., 1983. Hillslope hydrology water harvesting and areal distribution of some ancient agricultural systems in the northern Negev Desert. Journal of Arid Environments 6, 283–301. Yair, A., 1990. Runoff generation in a sandy area — the Nizzana sands, western-Negev, Israel. Earth Surface Processes and Landforms 15, 597–609. Yair, A., Klein, M., 1973. The influence of surface properties on flow and erosion processes on debris covered slopes in an arid area. Catena 1, 1–18. Yair, A., Kossovsky, A., 2002. Climate and surface properties: hydrological response of small arid and semi-arid watersheds. Geomorphology 42, 43–57.