Volume 47, Issue 3 e2019GL084484
Research Letter
Open Access

Experimentally Derived Thresholds for Windblown Sand on Mars

C. Swann

Corresponding Author

C. Swann

Sediment Dynamics Section, Seafloor Sciences Branch, Marine Geosciences Division, U.S. Naval Research Laboratory, Stennis Space Center, MS, USA

Correspondence to: C. Swann,

[email protected]

Search for more papers by this author
D. J. Sherman

D. J. Sherman

Department of Geography, University of Alabama, Tuscaloosa, AL, USA

Search for more papers by this author
R. C. Ewing

R. C. Ewing

Department of Geology and Geophysics, Texas A&M University, College Station, TX, USA

Search for more papers by this author
First published: 22 October 2019
Citations: 39


Aeolian sand transport frequently occurs on Mars despite wind speeds rarely exceeding predicted thresholds for motion. This dissonance is problematic for understanding aeolian processes and the resulting geomorphologic responses in contemporary and ancient atmospheres. To address the apparent discrepancy between speeds required to initiate sand motion and transport observed from orbital and in situ observations, we conducted a series of wind tunnel experiments at Martian atmospheric pressures and used buoyancy as the similitude parameter to simulate the force required to initiate particle motion on Mars. Shear velocities were derived from velocity profiles corresponding to the onset of sporadic transport, which can induce downwind cascading motion. Here, we find that threshold shear velocities are slower than previously thought by a factor of 1.6 to 2.5. Measured wind speeds on Mars exceed our observed thresholds, thus offering one mechanism behind the dissonance between observations of transport below previous thresholds of motion.

Key Points

  • Threshold wind speeds for sand transport on Mars are slower than previously inferred
  • Martian winds exceed the threshold for cascading motion
  • Overprediction of thresholds lead to underprediction of the frequency of sand transport by Martian winds

Plain Language Summary

Robotic lander and satellite images show sand movement by wind frequently occurs on Mars. Martian wind speeds, however, rarely exceed the minimum wind speed thought to be required to move sand. To resolve this dilemma, we conducted wind tunnel experiments to determine if sand could move at slower wind speeds than previously predicted. Previous experiments used the onset of continuous motion throughout a wind tunnel to define the minimum wind speeds required to move sand on Mars and ignored sporadic transport that can occur at slower wind speeds. Here, we use the onset of sporadic bursts of sand movement in a wind tunnel to define the minimum wind speed for sand motion on Mars. These bursts can induce a cascade of motion that develops from discrete patches and grows exponentially into continuous transport, which has the potential to produce significant landform change on Mars. We find the minimum speeds necessary to initiate sporadic bursts of motion are slower than previous model estimates by a factor of 1.6 to 2.5. Our results offer one explanation for abundant ripple and dune movement and dust emission under current thin-atmosphere climate conditions and suggest winds have sculpted the Martian landscape over billions of years.

1 Introduction

Aeolian processes play an important role in shaping ancient and modern landscapes on Mars (Bridges et al., 2012; Greeley & Iversen, 1985; Sullivan et al., 2008). Sandstones formed by migrating dunes indicate the ancient Martian atmosphere mobilized sand effectively (Banham et al., 2018; Grotzinger et al., 2005; Lapotre et al., 2016). Recent remote sensing images indicate ripples are currently migrating up to 4.5 m/year (Bridges et al., 2012) and dunes are migrating on an average of 0.5 m/year (Chojnacki et al., 2019), making it clear that contemporary Martian winds move substantial amounts of sand (Baker et al., 2018). Yet, in situ wind speed observations (Newman et al., 2017; Viúdez-Moreiras et al., 2019a) have not exceeded predicted threshold shear velocities. Current models estimate that threshold shear velocities between ~1.09 and 1.73 m/s are required to initiate transport of fine-to-medium size grains (grain size, d = 200 μm; Bagnold, 1936; Iversen & White, 1982; Kok & Renno, 2006; Shao & Lu, 2000), and using von Karman's Law of the Wall, these shear velocities translate to minimum and maximum modelled threshold wind speeds of 26 (Bagnold, 1936) and 42 m/s (Iversen & White, 1982) at 1.5 m above the surface. These values are greater than or equal to the maximum measured wind speed of 23 (±3.45) m/s by the Viking Lander 2 (Lorenz, 1996) and faster than reliably measured wind speeds from the Rover Environmental Monitoring Station wind sensor on the Curiosity rover (Gómez-Elvira et al., 2014; Newman et al., 2017). Thus, windblown sand occurs on Mars despite wind speeds not reaching predicted thresholds for motion. The inconsistencies between threshold wind speed models, maximum measured wind speeds, and observed sand movement preclude accurate estimations of dune and ripple movement, dust lifting events via the impact of saltating grains, and bedrock erosion rates. To bridge this fundamental gap, we reevaluated the fluid threshold of motion for windblown sand on Mars.

The fluid threshold is the shear velocity, u*t, or equivalent wind speed at height z, ut, required to entrain particles by fluid drag alone. The fluid threshold typically exceeds the impact threshold, which sustains particle motion via both the shear imparted by the fluid and momentum from the impact of saltating grains (Bagnold, 1936; Kok et al., 2012). Once the fluid threshold is exceeded, transport at a point will continue until the wind speed drops below the impact threshold. On Mars, the impact threshold is estimated to be significantly slower than the fluid threshold because of the amount of energy saltating grains impart on the surface due to the low atmospheric density (Kok, 2010). Using low-pressure wind tunnel experiments, we aimed to identify wind speeds that coincided with the fluid threshold at the onset of sporadic grain movement. Such movement is initiated by turbulent structures whose transient velocities can be significantly faster than the mean wind speed, which is typically used to resolve threshold shear velocity. Sporadic grain movement can develop into cascading transport, which is one element of the commonly known fetch effect in which sporadic grain movement triggers a “cascade” of saltation that leads to downwind saturated transport over sandy surfaces (Bauer et al., 2009; Carneiro et al., 2015; O'Brien & Neuman, 2018; Parteli & Herrmann, 2007; Sullivan & Kok, 2017). Recently, cascading transport was recognized as potentially important phenomena on Mars because its low atmospheric density facilitates high-energy grain interactions with the bed (Sullivan & Kok, 2017). These high-energy impacts on the bed cause saltating grains to sustain and increase transport over much longer distances than on Earth before attaining an equilibrium or saturated state (Kok, 2010; Rasmussen et al., 2015; Sullivan & Kok, 2017). In the case of grains saltating across immobile surfaces, this same process can sustain transport over long distances (Rasmussen et al., 2015).

Despite the potential significance of cascading saltation to explain sand movement on Mars, this mode of entrainment has yet to be considered in observations during Martian threshold experiments. In the original MARtian Surface Wind Tunnel (MARSWIT) experiments (Greeley et al., 1976; Greeley et al., 1980; Iversen & White, 1982), the condition intended to define the fluid threshold shear velocity for sand motion (u*t) was rooted in the early experiments of Bagnold (1941), and described as “… (following Bagnold, 1941) the movement of particles over the entire test bed” (Greeley et al., 1976, p. 418). This characterization was, however, a misunderstanding of Bagnold's definition; Bagnold used this criterion to describe the impact threshold (Bagnold, 1941, p. 32). To predict motion over the entire test bed, three conditional threshold expressions (Greeley & Iversen, 1985) were developed from the original experiments. Further, their derived models contained an empirical constant that was not dimensionally corrected to account for differences in fluid properties and gravity on Mars, a step that would have produced slower threshold shear velocities. Overprediction of u*t leads to underprediction of the frequency and magnitude of sand transport by Martian winds. Here, we derive threshold shear velocities from low-pressure wind tunnel experiments at the Planetary Aeolian Laboratory's MARSWIT at NASA's Ames Research Center. Distinct from the previous MARSWIT experiments, our focus was on recognizing and quantifying conditions for the initiation of sporadic grain movement, applying dimensional transformations to account for differences between terrestrial and Martian environments, and using the resulting experimental data to develop new threshold expressions.

2 Materials and Methods

2.1 Experimental Design

Experiments were conducted in the MARSWIT, a 13-m-long, open-circuit boundary-layer wind tunnel housed within a chamber where atmospheric pressure can be reduced to an equivalent of that on the surface of Mars. We conducted experiments at air pressures, P, between 500 and 1,000 Pa and air densities, ρ, between 0.007 and 0.011 kg/m3. Following the methods from Greeley et al. (1976), we used buoyancy similitude during tests conducted on Earth to simulate Mars. Test sands consisted of ground walnut shells whose density, ρs, of 1,110 kg/m3 compensates for the gravitational difference between Earth (g = 9.81 m/s2) and Mars (g = 3.71 m/s2) in order to obtain buoyancy similitude for particles at rest. Using buoyance as the similitude parameter for Martian atmospheric conditions is required because we cannot recreate Mars' gravity in the MARSWIT facility, but sand particles with a density reduced to the proportion between Earth and Mars gravity (where Mar's gravity is 0.38 of Earth's gravity) can be used to simulate the force required to initiate particle motion (Greeley et al., 1976, p. 418). The test bed was approximately 6.5 m in length, 1.3 m wide, and located downwind of a roughness plate that was used to trip a turbulent boundary layer. Sand beds were 20 mm deep, and the surface was leveled from the end of the roughness plate to a position downwind of instrumentation. We tested threshold conditions with three surfaces comprising unimodal grain size distributions, with average sizes and standard deviations of medium sand, 310 and 80 μm (Surface 1), coarse sand, 730 and 20 μm (Surface 2), and very coarse sand, 1,310 and 300 μm (Surface 3; Figure S1 in the supporting information). Grain size statistics were calculated using the arithmetic Method of Moments (Blott & Pye, 2001) and with data obtained with a Retsch Camsizer grain size analyzer.

Before each set of experiments, the sand surface was manually smoothed and then dynamically conditioned by briefly increasing the wind speed to achieve full saltation to remove perched grains. Then the wind speed was reduced to below the threshold of motion to begin each experimental run. During a run, the mean tunnel wind speed was increased in steps of ~5 m/s and held constant for 30 s. Velocities were measured at elevations of 4, 66, 160, and 320 mm above the initial surface elevation (Figure S2 in the supporting information) using independently sampled, differential capacitance manometers (Baratron® MKS 226A) connected to stainless steel pitot tubes (Dwyer 166-12). Particle movement was observed and recorded at 30 fps using infrared HD IP video cameras (Figure S2 and Movie S1 in the supporting information). Pressure and temperature were monitored continuously through the experiments. The elevations of pitot tubes were measured before and after each set of experiments and when differences were found, linear interpolation was used to estimate changes through time for the purpose of fitting log-linear velocity profiles (Tables S1 and S2).

2.2 Defining the Threshold

To define a new u*t, we identified transitions in the bed state as wind speed increased. Sporadic grain movement was defined as the transition from a state of no motion to one of minimal movement, in which individual grains or small grain clusters rolled or jumped across a mostly static bed and sporadic clusters of saltating grains were being lifted from numerous areas of the bed (Movie S1 in the supporting information). Continuous transport, which was the previously defined u*t, occurred when saltating grains were continuously visible across the entire bed. This latter threshold is analogous to the general threshold from flume experiments where continuous transport along the flume is used to define a maximum threshold condition (Casey, 1935; Kramer, 1935); herein, our estimates of u*t corresponding to continuous transport throughout the wind tunnel is referred to as the general threshold. Because the bursts of motion that manifest prior to a state of continuous motion are thought to be capable of leading to cascading transport (Bauer et al., 2009; Carneiro et al., 2015; O'Brien & Neuman, 2018; Parteli & Herrmann, 2007; Sullivan & Kok, 2017), we adopt the onset of sporadic bursts as the new criterion to define the Martian fluid threshold, u*t (section 2.2).

Visual observations of motion were recorded during the wind tunnel experiments based on video observations of motions (Movie S1 in the supporting information) by the same observer for the duration of the experiments. Visual percentages of the bed were estimated as the bed transitioned from intermittent, sporadic motion (fluid threshold) to continuous transport (previously defined, general threshold). The onset of single saltating grains prior to bursts of motion were not considered as the cascading threshold. These visual percentages (Tables 1 and 2) occur over a limited range of shear velocities. The range of shear velocities recorded is due to visual percent delineations (i.e., 5–30% of motion for the cascading, fluid threshold and 40–60% for continuous, general threshold motion), conditioning of the bed after multiple threshold exceedances, and the stochastic nature of the threshold of motion (Lavelle & Mofjeld, 1987).

Table 1. Data Summary for Establishment of Fluid Threshold
Surface urn:x-wiley:00948276:media:grl59650:grl59650-math-0001 T (°C) P (mb) Visual motion (%)

u*t tunnel




u*t Mars


Rep Mars AFluid D*
S1 dEarth = 310 μm 0.009 15.2 8.0 20 1.26 200 0.69 0.21 0.06 1.56
0.009 15.6 8.0 30 1.23 200 0.68 0.21 0.06 1.56
0.009 15.9 7.9 20 1.00 200 0.55 0.17 0.05 1.56
0.009 16.4 8.1 30 1.31 200 0.72 0.22 0.07 1.56
0.009 17.0 8.1 30 1.25 200 0.69 0.21 0.06 1.56
0.009 17.3 8.0 30 1.19 200 0.65 0.20 0.06 1.56
S2 dEarth = 730 μm 0.007 16.5 6.7 20 2.46 430 1.12 0.73 0.07 3.09
0.007 17.0 6.5 5 2.16 430 0.98 0.64 0.06 3.09
0.007 17.1 6.8 10 2.65 430 1.21 0.79 0.08 3.09
0.007 18.0 6.6 10 2.37 430 1.08 0.71 0.07 3.09
0.007 18.4 6.9 20 2.88 420 1.25 0.80 0.08 3.01
0.007 18.7 6.9 20 2.70 420 1.17 0.75 0.07 3.01
0.006 18.9 6.3 5 1.75 420 0.76 0.49 0.05 2.86
0.007 9.0 6.6 40 2.37 420 1.03 0.66 0.07 3.01
S3 dEarth = 1,310 μm 0.01 14.8 8.7 5 3.98 780 1.89 2.27 0.09 6.3
0.01 15.1 8.8 5 3.61 780 1.71 2.06 0.08 6.3
0.01 15.2 8.9 30 3.87 780 1.83 2.20 0.09 6.3
0.008 16.5 7.8 10 3.64 750 1.60 1.86 0.08 5.63
0.009 16.6 8.2 30 4.00 750 1.76 2.04 0.08 5.85
0.009 17.0 8.4 5 4.20 750 1.85 2.14 0.09 5.85
0.009 17.0 8.5 20 4.34 750 1.91 2.22 0.09 5.85
0.01 17.1 8.7 30 4.48 750 1.97 2.28 0.09 6.06
0.01 17.4 8.9 30 4.50 750 1.98 2.29 0.09 6.06
0.01 18.0 9.2 5 4.63 750 2.04 2.36 0.10 6.06
0.01 18.0 9.4 20 4.78 750 2.11 2.44 0.10 6.06
0.01 18.2 9.5 20 4.94 750 2.18 2.52 0.10 6.06
  • Note. Visual motion refers to % of bed active. Subscripts indicate the representative atmospheric boundary condition (e.g., u*t tunnel and u*t Mars). Tunnel corresponds to the experimental conditions. D* is the dimensionless grain size used to model AFluid at the onset of sporadic motion, that is, the cascading threshold. Slight variability in dmars, derived from equation 6, is due to changes pressure, density, and temperature during experiment runs.
Table 2. Data Summary for Establishment of General Threshold
Surface urn:x-wiley:00948276:media:grl59650:grl59650-math-0002 T (°C) P (mb) Visual motion (%)

u*t tunnel




u*t Mars


Rep Mars AFluid D*
S1 dEarth = 310 μm 0.009 15.9 8 60 1.36 200 0.75 0.23 0.07 1.56
0.009 16.2 8.1 60 1.31 200 0.72 0.22 0.07 1.56
0.009 16.5 8.2 50 1.48 200 0.81 0.25 0.07 1.56
0.009 16.7 8.2 50 1.49 200 0.82 0.25 0.08 1.56
0.009 17 8.2 60 1.46 200 0.8 0.24 0.07 1.56
0.009 17.3 8.1 40 1.44 200 0.79 0.24 0.07 1.56
S2 dEarth = 730 μm 0.007 16.4 7 60 2.85 430 1.3 0.85 0.08 3.09
0.008 17.1 7.1 40 3.18 430 1.45 0.95 0.09 3.23
0.007 17.6 7 40 2.99 430 1.36 0.89 0.09 3.09
0.007 18 6.9 40 2.95 430 1.34 0.88 0.08 3.09
0.008 18.5 7.3 50 3.4 420 1.48 0.95 0.09 3.15
0.007 18.7 7.1 40 3.05 420 1.33 0.85 0.08 3.01
0.008 18.7 7.3 60 3.33 420 1.45 0.93 0.09 3.15
0.075 19 7.2 40 3.33 420 1.45 0.93 0.09 3.01
S3 dEarth = 1,310 μm 0.01 14.9 8.9 40 4.12 780 1.96 2.36 0.09 6.3
0.01 14.9 9.2 40 4.23 780 2.01 2.41 0.09 6.3
0.01 15.2 9.2 50 4.03 780 1.91 2.3 0.09 6.3
0.01 16.7 8.7 50 5.13 750 2.26 2.62 0.11 6.06
0.01 17.1 9 50 4.96 750 2.18 2.53 0.1 6.06
0.01 17.4 9.1 40 4.75 750 2.09 2.42 0.1 6.06
0.11 17.5 9.7 60 5.58 750 2.46 2.85 0.12 6.26
0.01 18.1 9.6 40 5.22 750 2.3 2.66 0.11 6.06
0.01 18.2 9.6 40 5.05 750 2.22 2.57 0.11 6.06
  • Note. Visual motion refers to % of bed active. Subscripts indicate the representative atmospheric boundary condition (e.g., u*t tunnel and u*t Mars). Tunnel corresponds to the experimental conditions. D* is the dimensionless grain size used to model AFluid at the onset of continuous motion, that is, the general threshold. Slight variability in dmars, derived from equation 6, is due to changes pressure, density, and temperature during experiment runs.

2.3 Dimensional Transformation

Bagnold's threshold shear velocity model uses A, an empirical constant to differentiate between fluid and impact thresholds, u*t = A(((ρs − ρ)/ρ)gd)½, where ρs is sediment density (3,200 kg/m3 for basalt), ρ is atmospheric density (0.02 kg/m3) on Mars, g is gravitational acceleration (3.71 m/s2), and d is sand grain size. As an empirical factor derived under the influence of Earth's gravity and air composition, a dimensional transformation needs to be operated in order to estimate its value on Mars.

Experimental u*t estimates were derived by regressing the estimates of A with dimensionless grain size, urn:x-wiley:00948276:media:grl59650:grl59650-math-0003 where γ is the submerged weight of the grain (g(ρs − ρ)), and μ is the absolute viscosity of the fluid, 1.5 × 10−5 kg·m−1·s−1, on Mars. We estimate u*t from our data following methods to dimensionally transform experimental observations in the wind tunnel to Martian equivalents. Dimensional transformation is appropriate when predicting transport in environments with fluid conditions that differ from the experimental conditions (Le Roux, 2005; Southard, 1971; Southard & Boguchwal, 1990). We follow the conceptual framework of Southard (1971) and Southard and Boguchwal (1990) and utilize dimensionless versions of threshold shear velocity, urn:x-wiley:00948276:media:grl59650:grl59650-math-0004, and grain size, d0, to convert our wind tunnel data to Mars equivalents by compensating for differences in fluid density, viscosity, and gravity:
where γ = g(ρs − ρ) is the submerged weight of the grain, μ is the absolute viscosity of the fluid, and the subscripts T and M refer to Earth wind tunnel and Mars values. Experimental values for sand, gravity, and fluid parameters (ρs = 1,100 kg/m3, g = 9.81 m/s2, μ = 0.000018 kg·m−1·s−1, and ρ) were used in equation (1). Martian basaltic sand, gravity, and average fluid parameters (ρs = 3,200 kg/m3, ρ = 0.02 kg/m3, g = 3.71 m/s2, and μ = 0.000015 kg·m−1·s−1) were used to convert the MARSWIT data to obtain the Martian equivalent u*tM and dM (Tables 1 and 2). Herein, u*t = u*tM and d = dM. Martian atmospheric viscosity was estimated using Sutherland (1893):
where a = 0.555T0 + C, b = 0.555T + C, T0 is a reference temperature (527.67 °R), T is the air temperature, C = 240 is Sutherland's constant for CO2, and μ0 is the reference viscosity (0.0148 for CO2 in centipoises). Variations in atmospheric density in the test chamber during experimental runs (Tables 1 and 2) result in five Martian equivalent grain sizes: 200, 420, 430, 750, and 780 μm. Variations in atmospheric density in the test chamber during experimental runs along with the dimensional transformation result in variations in Martian equivalent grain size. There is a slight decrease between 750 and 780 μm in threshold shear velocity observations which is a function of variations in temperature, pressure, and atmospheric density during experimental runs (Tables 1, 2, S1, and S2).

2.4 Deriving a Threshold Curve from Experimental Observations

The slopes, m, of the velocity profiles, obtained from least-squares regression lines, were used to estimate shear velocities according to u* = κm, where κ is the von Kármán constant and m is the slope of the regression line. Profiles where the coefficient of determination was less than 0.90 were not used because of the large statistical uncertainty that would be associated with the estimates of u*. Forty-nine velocity profiles met our quality control threshold (Tables 1 and 2); raw wind speed and elevation data are reported in Tables S1 and S2.

We first empirically determine A via wind tunnel observations of u*t using equation 3 for the threshold of motion:
and then derive two expressions for Bagnold's A parameter to predict the threshold shear velocity on Mars (Figure 2):
where D* is the dimensionless grain size:

Empirical estimates of A were derived from 26 shear velocity estimates for the cascading threshold and 23 for the general threshold (Tables 1 and 2). Power regressions to arrive at threshold expressions for AFluid and AGeneral have root-mean-squared errors of 0.007 and 0.008, respectively, standard estimates of the error of ~0.001, and R2 values of 0.74 and 0.70, respectively.

Predicted threshold shear velocities were calculated using equations 46 and Bagnold's model, equation 7:

Atmospheric densities estimated from Curiosity observations (Ullán et al., 2017) ranging from 0.013 to 0.025 kg/m3 on the floor of Gale Crater which is ~4.5 km below the datum (Tables S3 and S4 in the supporting information), were used to predict threshold shear velocity for any given grain size.

We converted our minimum threshold shear velocities for fluid and general thresholds for 200-μm basalt sand particles to equivalent threshold wind speeds, u, at a height, z = 1.5 m (Rover Environmental Monitoring Station sensor) or 1.6 m (Viking Lander 2) above the bed using the Law of the Wall:
where κ is the von Karman constant of 0.40 and z0 is the extrapolated height above the bed at which the Law of the Wall would predict wind speed to be zero. Here z0 is assumed to 0.0001 m based on recent estimates (Bridges et al., 2017).

3 Results

Our threshold expressions indicate that the Martian fluid threshold for 200-μm particles ranges from 0.63 to 0.81 m/s (0.013 > ρ > 0.025 kg/m3). For 200-μm grains at an average surface density of ρ = 0.02 kg/m3 our model predicts u*t = 0.68 m/s, a threshold slower than that predicted by traditional models by a factor of 1.6 to 2.5 (u*t Bagnold = 1.09 m/s; u*t Iversen & White = 1.73 m/s; Figures 1 and 2). Our equivalent threshold wind speed for these conditions is about 40% less than Bagnold's original estimate for continuous motion where A = 0.1. For the same grain size and range of densities, the threshold shear velocities we obtained for the general threshold range from 0.76 to 1.00 m/s which are slower than those predicted by previous models (Figure 2 and Table S3 in the supporting information). This is due to the absence of dimensional transformation in previous Martian experiments. Equivalent wind speed ranges, depending on atmospheric density, for the two thresholds are 15.07–19.51 and 18.31–24.09 m/s, respectively, at 1.5 m above the surface (Figure S3 and Tables S3 and S4 in the supporting information).

Details are in the caption following the image
Comparison of threshold models with our wind tunnel results for fluid and general thresholds. Our observed general and fluid thresholds are well below the predicted values for 200- and 420-μm grains but conform reasonably with Bagnold's model for larger grain sizes. Martian basaltic sand, gravity, and average fluid parameters (ρs = 3,200 kg/m3, ρ = 0.02 kg/m3, and g = 3.71 m/s2) were used to derive model estimates for each model.
Details are in the caption following the image
New fluid and general threshold models. The maximum Viking Lander 2 wind speed is derived from hourly averaged winds. The Rover Environmental Monitoring Station instrument onboard Curiosity was calibrated to accurately measure winds up to 20 m/s. Our predicted thresholds fall below the maximum winds at Viking for grain sizes smaller than 300 and 250 μm for Curiosity. Stippled lines below 200 μm linearly extrapolate our model down to 100 μm and shows shear velocities slower than 0.5 m/s for noncohesive grains and nearly 1 m/s for cohesive grains, based on Shao and Lu (2000). The cross-hatched region denotes uncertainty based on the degree of cohesion.

Cohesion between particles becomes increasingly important as grain sizes decrease from about 200 μm (Iversen & White, 1982; Merrison et al., 2007; Shao & Lu, 2000), the lower limit of our experimental range. Martian dune sands at Bagnold Dunes in Gale Crater and ripple sands at El Dorado in Gusev Crater have been found to have sizes closer to 100 μm (Ewing et al., 2017; Sullivan et al., 2008; Weitz et al., 2018). Further, Sullivan et al. (2008) found sand to be cohesionless. Exploring the uncertainty of thresholds for smaller grain sizes by extrapolation provides useful bounds for predicting the ranges of potential transport frequency of these smaller, cohesionless sands. If cohesion from interparticle forces plays a significant role in resisting movement, then the fluid threshold for 100-μm grains would be approximately equal to or greater than that for 200-μm grains (Figure 2). Conversely, Sagan and Bagnold (1975) linearly extrapolate their experimentally derived law to finer grain sizes assuming the absence of cohesion on Mars. Linear extrapolation of our fluid model indicates that 100-μm grains at densities ranging from 0.013 to 0.025 kg/m3 should be entrained from u*t = 0.36 to 0.46 m/s (Figure 2).

4 Discussion

Unlike previous models, our threshold curves produce estimates of u*t with equivalent wind speeds that are within the range of wind speeds measured on Mars (Lorenz, 1996; Newman et al., 2017; Viúdez-Moreiras et al., 2019a; Viúdez-Moreiras et al., 2019b). Five-minute averaged wind distributions observed at 0.25 Hz by Curiosity at z = 1.5 m predict wind speeds up to 20 m/s (Newman et al., 2017; Viúdez-Moreiras et al., 2019a; Viúdez-Moreiras et al., 2019b) and hourly averaged Viking Lander 2 wind speeds at z = 1.6 m (Lorenz, 1996) predict maximum speeds of 23 m/s. These measured Martian wind speeds exceed our minimum wind speed for sporadic, cascading motion for 200-μm grains, 15.07–19.51 m/s, while our minimum wind speed for the onset of continuous motion, 18.31–24.09 m/s, is only exceeded at the higher range of atmospheric densities (Figure S4). Thus, our model indicates Martian winds are capable of exceeding both thresholds for motion, but the threshold for sporadic, cascading motion is likely exceeded more often. Given saltation hysteresis and the substantially slower impact threshold on Mars (Kok, 2010; Kok et al., 2012), transport events, once initiated by sporadic, cascading motion, may be relatively long-lived and proficient at bedform migration and dust emission. These new threshold estimates support observations of wind as an effective agent of landscape change on Mars as noted by ripple and dune migration observed on Mars (Baker et al., 2018; Bridges et al., 2012; Bridges et al., 2017; Chojnacki et al., 2011; Chojnacki et al., 2019; Hansen et al., 2011; Runyon et al., 2017; Silvestro et al., 2013).

While our threshold curve predicts slower transport thresholds, it does not account for the varied nature of the Martian surface that can greatly affect the wind speeds at which sand moves. Thresholds vary with surface roughness, crusting, moisture content, electrostatics, particle type, and surface composition that act to increase threshold shear velocities, which are not accounted for here (Li et al., 2014; Sherman et al., 2013). Increased roughness from bedforms (e.g., meter-scale ripples (Ewing et al., 2017; Lapotre et al., 2016)) or rocks can also increase threshold values. Further, our observations are for a monodisperse grain distribution with relatively angular particles. While monodisperse distributions are consistent with sands found in the Bagnold Dune Field (Weitz et al., 2018), they do not represent thresholds for heterogeneous beds typical of many Martian surfaces (Sullivan et al., 2008). Additionally, many particles on Mars are well-rounded, unlike our angular walnut shells, which may result in variations in the thresholds.

5 Conclusions

Fluid threshold shear velocities for the onset of cascading transport indicate that Martian winds have the potential to move sand, relaxing the need for faster-than-measured wind speeds to explain present-day sand motion on Mars. Our new threshold curve offers an explanation for the dissonance between rover and orbital observations of aeolian transport and previous model-predicted values for the initiation of sand motion by wind on Mars. This new expression for sand transport, equation 4, is in agreement with the observation that the aeolian landscape on Mars is active and can be reshaped over annual to decadal timescales. Our results highlight the significance of determining the threshold for cascading grain motion in thin atmospheres and show how overestimates of the transport threshold greatly minimize the potential work done on planetary landscapes by wind and sand. Recognition of windblown features on the surfaces of Pluto, Triton, and 67P/Churyumov–Gerasimenko may signal that cascading grain motion arising from turbulent bursts plays a large role in shaping landscapes on other celestial bodies. Our threshold curves are consistent with mechanistic models, numerical experiments, and observations of transport on Mars (Kok et al., 2012; Sullivan & Kok, 2017) and give rise to a picture of Mars in which frequent cascading sand transport sculpted the Martian landscape over billions of years.


We would like to thank the staff at NASA/ASU's Planetary Aeolian Laboratory, in particular the MARSWIT manager Ken Smith, for assistance in acquiring optimal experimental data. We would also like to thank two anonymous reviewers for comments that greatly improved this manuscript. This study was funded by NASA Mars Fundamental Research Program Grant NNX14AO10G. C. Swann contributions were conceptualization, data curation, formal analysis, investigation, methodology, writing-original draft, and writing-review and editing. D. J. Sherman contributed via formulation, investigation, methodology, and writing-review and editing. R.C. Ewing contributed via conceptualization, funding acquisition, investigation, project administration, supervision, and writing-review and editing. Authors declare no competing interests. Our experimental data is available in the text and supporting information and in the Planetary Data System (PDS).