Mechanisms of dust emission from Pleistocene loess deposits, Nebraska, USA

Authors


Abstract

[1] Saltation bombardment is commonly believed to be the primary mechanism by which large quantities of dust can be generated. Direct entrainment of silt, on the other hand, is thought to generate minimal quantities of dust. If so, loess landscapes should rarely be major sources of dust unless disturbed by human activity. To test the mechanisms by which loess can be eroded by the wind, we used the Portable In-Situ Wind Erosion Lab (PI-SWERL). PI-SWERL tests were conducted on benches that were carved into loess at field locations. Coarse to fine loess was also collected and tested in a controlled setting using (1) dry, unconsolidated loess and (2) crusted loess. Results indicate that in most cases, the threshold friction velocity for silt is less than that for sand and that most dust generation occurs by direct entrainment of dust. Saltation, even in coarse loess, was nonexistent to intermittent. In a few cases, saltation of sand or soil aggregates preceded dust generation or enhanced emissions after direct entrainment began. Emission fluxes of loess are potentially high in both proximal and distal settings, suggesting that loess is easily entrainable if vegetation density is low. We hypothesize that an arid and windy climate of the late Pleistocene, paired with lower vegetation density, facilitated large-scale erosion of Peoria Loess in Nebraska to generate wind-aligned ridges and troughs. More broadly, our results indicate that loess landscapes have the potential to be major sources as well as sinks of dust.

1 Introduction

[2] The classic mechanism for dust entrainment involves saltation bombardment, where the impacts of saltating grains liberate dust-sized particles for suspension transport [Bagnold, 1941; Gillette et al., 1972; Shao et al., 1993; Rice et al., 1996]. Dust emissions are sustained once the threshold velocity for saltation is reached and exceeded by the wind [Bagnold, 1941; Shao et al., 1993]. Dust composed of clay- and silt-sized particles, in theory, require greater threshold velocities for entrainment on smooth surfaces due to the binding energies of these particles [Iversen and White, 1982]. Atmospheric dust loading models generally calculate dust emissions based on the assumption that dust production occurs through saltation bombardment [Marticorena and Bergametti, 1995; Alfaro and Gomes, 2001; Tegen, 2003; Zender et al., 2003].

[3] Bagnold [1941, pp. 90–91] suggested that a relatively greater threshold velocity for direct entrainment makes dust deposits relatively immobile in the absence of saltating sand, based on a wind-tunnel experiment and qualitative field observations. This view is consistent with the emphasis on the role of saltation in much of the research on dust emission since Bagnold's time. Recently, researchers have documented direct aeolian entrainment of dust into suspension without saltation [Loosmore and Hunt, 2000; Kjelgaard et al., 2004; Roney and White, 2004; Shao, 2004; Sharratt et al., 2007; Macpherson et al., 2008]. In conditions where silt and clay are not entirely bound in crusts, the threshold velocity for direct entrainment of dust tends to be less than the threshold velocity for that of sand, but the reported magnitudes of dust generation by direct entrainment are much lower than dust generation rates from saltation bombardment [Loosmore and Hunt, 2000; Macpherson et al., 2008]. Dust produced by direct entrainment contributes to ambient dust loading [Macpherson et al., 2008] and is potentially significant over geologic timescales.

[4] Thick loess deposits in semiarid regions are net sinks of dust during major periods of loess accumulation; however, they contain vast quantities of sediment that can potentially be reentrained as suspended dust, if exposed to the wind by human land use or climatically driven vegetation change. Furthermore, many spatial patterns of thick loess accumulation can be explained by assuming that loess is relatively stable where it is shielded from saltating sand by barriers such as entrenched stream valleys [Mason et al., 1999; Mason, 2001; Sweeney et al., 2005]. If saltation bombardment is required for significant dust entrainment from loess deposits, then resuspension could be limited mainly to areas where aeolian sands encroach upon the loess, e.g., at the northern margin of the Chinese Loess Plateau [Ding et al., 1999; Rokosh et al., 2003] or at boundaries between dune sand and loess in the central Great Plains [Smith, 1965].

[5] On the other hand, coarse loess deposits can have particle size modes within or just below the range of 60–120 µm [Mason et al., 2003], for which the wind entrainment threshold is at a minimum for loose particles [Iversen and White, 1982]. If the loess is not too cohesive, mobilization of the coarser particles within it could initiate wind entrainment of the sediment as a whole, even without abrasion by saltating sand from an external source. In that case, environmental change that reduces vegetation cover could trigger deflation and atmospheric dust loading over extensive loess landscapes. Landforms observed in some loess-mantled landscapes suggest large-scale deflation of loess after its initial accumulation, including troughs and scarps in the central Great Plains that are apparently wind aligned (Figure 1) and streamlined, wind-aligned ridges in other regions of thick loess [Lewis, 1960; Shawe, 1963; Flemal et al., 1972; Leger, 1990].

Figure 1.

(a) Map depicting locations of in situ bench tests (black circles), loess collected for controlled tests (white circles), and sand dune sites (white square). Labeled sites are a = Orr, b = Fanning, c = Foran, d = Murphy, e = Broken Bow, f = Wauneta, g = Judkins, h = Ansley, i = Prairie Lake, j = Obert, and k = Barta sand dunes. Black lines are contours of Peoria Loess thickness (m), interpolated from 518 point measurements, showing decreasing thickness along transport path from northwest to southeast [Mason, 2001]. Contours do not extend into northeastern and western Nebraska because of high local thickness variability there. (b) Troughs cut in thick Peoria Loess and apparently aligned with northwesterly winds, western Nebraska; location shown in Figure 1a. Black dot marks site a.

[6] To address these issues, we used the Portable In-Situ Wind Erosion Lab (PI-SWERL) to investigate thresholds for saltation and dust entrainment and rates of dust emission from late Pleistocene Peoria Loess deposits in the central Great Plains of Nebraska. The results allow assessment of the relative importance of direct entrainment and saltation bombardment in dust formation. Peoria Loess is the most extensive loess stratigraphic unit in North America and forms a mantle up to 50 m thick in our study area [Bettis et al., 2003]. Experiments with the PI-SWERL were carried out in the field on benches carved into undisturbed loess and indoors using plots of loess samples of varying grain size.

2 Methods

[7] The PI-SWERL was developed to provide an index of dust emission potential from soil surfaces [Etyemezian et al., 2007] and has been used to assess both natural and anthropogenic dust emissions from a variety of soil and geomorphic conditions [Sweeney et al., 2008; Kavouras et al., 2009; Goossens and Buck, 2009; Bacon et al., 2011; King et al., 2011; Sweeney et al., 2011]. The PI-SWERL mobilizes sand and dust by the shearing action of a rotating annular blade contained within a cylindrical metal chamber open to the ground surface. The PI-SWERL can be programmed to operate at variable rotations per minute (RPMs) that translate into specific friction velocities (u*, m s−1) based on shear stress measured in the lab on a smooth surface using skin friction meters [Etyemezian et al., 2007]. The dust generated by shear stress is measured as particulate matter < 10 µm (PM10) using real-time instrumentation (DustTrak TSI Model 8520) in mg m−3. Fresh air is continuously fed into the chamber at a known rate. An emission flux (mg m−2 s−1) at a specific RMP level i can be calculated using the equation

display math(1)

where C is the dust concentration (mg m−3), F is the blower flow rate for fresh air entering the PI-SWERL chamber (m3 s−1), Aeff is the effective test area underneath the PI-SWERL annular ring (m2), and t is the time (s) at the beginning (tbegin,i) and ending (tend,i) of each RPM step level, i [Kavouras et al., 2009]. The PI-SWERL has a diameter of approximately 0.5 m and can easily be placed in natural conditions on bare soil. Because of the small size of the area tested by the PI-SWERL, and the fact that it does not generate straight-line winds, an extensive comparison was carried out between the emissions data collected by the PI-SWERL and in a large field wind tunnel [Sweeney et al., 2008]. That study showed an approximately 1:1 relationship between dust emissions measured by the two methods at similar friction velocities.

[8] Sand movement sensors mounted on the inside of the PI-SWERL chamber count sand grains, including sand-sized aggregates, that pass through an optical gate measured every second with an infrared light sensor (V. Etyemezian, personal communication, 2012). The optical gates used in this study are positioned approximately 0.035 m above the testing surface. Sand movement is interpreted as saltation and can provide a general sense of the magnitude of saltation, but a sand flux cannot be calculated. At zero and low RPMs, the sand movement sensors typically register < 7 counts per second (cps) which are interpreted as background noise when little or no saltation is occurring. A threshold friction velocity for saltation (u*t) is estimated when sand movement begins to increase above that background noise. This estimate is made using a 20-point (20 s) moving average of sensor data. The smoothed curve allows identification of the inflection point, after which saltation is sustained and increasing (V. Etyemezian, personal communication, 2012). We interpret intermittent saltation when saltation counts sporadically jump above 7 cps, but are not sustained. Intermittent saltation may occur on supply-limited surfaces, where abundant saltators are not always available, or may occur due to fluctuations in wind strength [Stout and Zobeck, 1997]. Our tests do not simulate fluctuating wind speeds, so when intermittent saltation occurs it is likely due to supply limitation. Intermittent saltation may occur before the true threshold of sustained saltation.

[9] Two types of PI-SWERL tests were conducted on the loess. A “ramp test” gradually increases the RPM of the PI-SWERL, and therefore the friction velocity, to a predetermined maximum. The ramp test is used to determine the threshold velocities for both dust and sand entrainment. During a “step test,” RPMs are increased in a stepwise fashion and are held at each step to measure the response of the surface to several constant friction velocity values. The step test is used to calculate the emission flux at a specific friction velocity (equation (1)). Both types of tests record the potential for a surface to emit dust and were performed on most field surfaces, except for a few cases. The field testing procedure for PI-SWERL can easily be modified to accommodate an appropriate range of RPM to constrain the threshold velocities required for initiation of dust emission and/or saltation on Peoria Loess surfaces.

[10] The threshold friction velocity for particle entrainment, u*t, is a function of several factors including particle size, soil moisture, soil crusting, size distribution of soil aggregates, and surface roughness [Zobeck, 1991; Ravi et al., 2011]. Belly [1964] demonstrated that an increase in soil moisture resulted in an increase in u*t. High atmospheric humidity has also been shown to increase u*t [Ravi and D'Odorico, 2005]. Binding of surface sediment by salts, other cementing agents, or surface crusts made of silt and clay increase the threshold for emissions or suppress particle movement altogether [Nickling and Ecclestone, 1981; Gillette et al., 1982; Rice et al., 1997]. Soils with sand-sized, silt-clay aggregates have a high potential for dust emissions because they can be entrained at velocities similar to sand grains, contributing dust via saltation bombardment or by breaking apart during transport [Alfaro et al., 2004]. Surface roughness imparted by vegetation cover can also decrease wind erosion potential [Wolfe and Nickling, 1993].

[11] PI-SWERL tests were conducted on bare surfaces cut into in situ loess at five sites (Foran, Fanning, Orr, Murphy, Ansley; Figure 1a) and in a more controlled indoor setting using plots constructed from disturbed loess samples. In both cases, sites or samples were selected to represent a range from coarse to fine loess; not all loess types were testable in the field because of limited access. Field tests were carried out by carving benches that were the approximate size of the PI-SWERL footprint (0.25 m2) into loess outcrops below the surface soil A and B horizons. The benches were carefully smoothed to create a flat surface for testing, but loose aggregates were often present. The benches were allowed to dry for at least 1 day before testing. Moisture content of sediment within 5 mm below the surface was measured gravimetrically. In a few cases, a brief rain shower followed by several days of drying created a crust on the benches before testing.

[12] Indoor plot tests were conducted in a large, well-ventilated warehouse with two surface conditions: loose and crusted. The loose surface was meant to represent freshly deposited, uncrusted loess. Two 5 gallon (18.9 L) buckets of loess were collected from each of five localities (Wauneta, Judkins, Broken Bow, Prairie Lake, Obert; Figure 1a) representing a range from coarse to fine loess. The loess was passed through a 2 mm sieve to break apart any large aggregates and spread to a uniform thickness (~0.06 m) in a 1 × 1 m shallow box forming a plot on which four PI-SWERL tests could be conducted. The plots were allowed to air-dry for at least 1 day and smoothed with a trowel but not compacted before testing occurred. Moisture contents were measured as with the field sites. Additional tests were performed by mixing and resmoothing the loess in each plot. Following PI-SWERL tests on loose loess, the loess in each plot was mixed and smoothed and then uniformly watered using a watering can from a height of about 2.5 m, as a crude simulation of rainfall. The goal was not to reproduce actual rainfall intensities but simply to create crusts similar to those that commonly form on bare loess surfaces in the field. The loess was air-dried for 2 days until crusts, mud cracks, and raindrop impressions were visible. PI-SWERL tests were then conducted on the crusted loess. To interpret major differences in PI-SWERL results on crusted surfaces, new crusts were developed on them using the same method, and shear vane tests were used to measure crust strength. At least 14 tests were performed on each loess type, using a Torvane pocket shear vane, which measures the stress in kg cm−2 required to break surface crust. For comparison, field PI-SWERL experiments were also carried out on dune sand in the Nebraska Sand Hills, using bare, dry surfaces of sand in blowouts and other exposures at the Barta Brothers Ranch, a research site operated by the University of Nebraska–Lincoln.

[13] Samples of loess from field and control plots were collected for particle size analysis and estimation of aggregate content. Particle size was measured with laser diffraction using a Malvern Mastersizer 2000. A procedure described by Mason et al. [2011a] was adapted to characterize aggregation and observe the disintegration of aggregates with wetting and mechanical energy application. The dry, untreated sample was suspended in deionized water, and repeated measurements of the particle size distribution were made as the sample circulated continuously through the particle size analyzer for 60 min (not 180 min as in Mason et al. [2011a] because these samples obtained a stable particle size distribution by 60 min). Aggregates were initially measured as sand- or silt-sized particles, and their subsequent disintegration into primary particles was evident in the changing size distribution. To obtain a fully dispersed particle size distribution, two final particle size measurements were made after (1) addition of 10 mL 50 g L−1 sodium hexametaphosphate dispersant to the circulating sample and (2) sonication for 3 min [Mason et al., 2011a].

3 Results

3.1 Particle Size, Aggregation, and Moisture Content of Loess

[14] Our loess samples, just after immersion in deionized water, had modal diameters between 85 and 50 µm (Figure 2, near 0 min circulation time) and textures of very fine sandy loam to silt loam with clay contents of 2% – 4% (Table 1). Over 60 min of circulation, modal diameters gradually become smaller, ultimately reaching 70 to 35 µm (Figure 2, points connected by smoothed line). After full chemical dispersion and sonication, modal diameters are even smaller (Figure 2, points at the right not connected by line); textures are mostly silt loam because of a decline in the sand fraction and include up to 6% clay (Table 1). Note that for Nebraska loess samples laser diffraction typically measures much less clay than other methods [Mason et al., 2007].

Figure 2.

The change in modal particle size with time for the loess tested in this study. The samples fine over time in response to the breakup of silt-clay aggregates during testing in deionized water. Points not connected by line are results after full chemical dispersion and sonication.

Table 1. Location and Grain Size Data for Loess Samples
 UTM CoordinatesMinimal DispersionMaximum DispersionAverage wt % H2O
ZoneEastingNorthingMode (µm)% PM10% Sand% Silt% ClayMode (µm)% PM10% Sand >63 µm% Silt 31–63 µm% Silt 16–31 µm% Silt 2–16 µm% Clay <2 µm
Prairie Lake14542990448785950.66.235.263.01.932.432.17.324.723.638.46.05.5
Obert14661083472301651.75.133.265.21.645.426.315.532.918.127.56.13.3
Broken Bow14459273458379954.23.735.263.31.545.917.319.735.820.120.73.75.4
Judkins Table14400263459548863.83.646.552.21.258.115.431.434.913.117.63.14.8
Wauneta14294973448605074.21.960.338.90.865.810.641.735.29.011.32.74.4
Ansley14473960456963356.23.241.757.01.346.821.418.335.318.223.94.35.0
Murphy14412925457653861.93.945.952.71.351.920.924.032.915.723.04.44.7
Fanning14290701448574772.53.258.840.30.958.819.629.431.413.122.23.82.0
Foran14398818459613972.13.655.543.51.065.915.837.333.117.023.22.96.2
Orr13746223457048074.34.955.143.61.358.619.430.328.614.722.14.23.6

[15] The changes in particle size with increasing circulation time indicate the presence of sand-sized aggregates (and possibly silt-sized as well) which slowly break down into primary mineral particles in response to wetting and mechanical stress. At least some of these aggregates may disintegrate under rainfall, contributing to dense crust formation. The additional changes after full dispersion demonstrate the presence of more persistent aggregates that may be intact and available for wind erosion even after repeated wetting. These results are consistent with previous work indicating that late Pleistocene loess in Nebraska has a significant content of aggregates, including many that are transported with the rest of the sediment [Mason et al., 2003].

[16] The modal diameters of the loess samples fall in the same order from coarse to fine (top to bottom in Figure 2) at all stages of disaggregation. This trend is consistent with other data indicating fining [Mason et al., 2003] and thinning (Figure 1a) of Peoria Loess southeastward along its regional transport path [Mason, 2001]. The coarser samples (e.g., Foran) have the least apparent aggregate content, indicated by the smallest changes in modal size. Prairie Lake loess has the largest change in modal size, about 20 µm, suggesting a larger component of aggregates in the finer loess. In discussing dust emissions of loess below, we group the samples based on particle size: (1) Wauneta, Judkins, Orr, Foran, and Fanning are coarse, proximal loess and (2) Prairie Lake, Broken Bow, Obert, Murphy, and Ansley are finer, more distal loess. Loess in both groups contains both primary sand grains and sand-sized aggregates, but aggregates are likely to make up a much larger portion of potential saltators in the finer loess.

[17] Moisture contents of loess on surfaces used for PI-SWERL tests ranged from 2.0% to 6.2% by weight (Table 1). We found no significant relationships between moisture content and dust emissions or thresholds for sand or dust entrainment; sites or plots with high emissions or low thresholds included some with relatively high moisture content. Although Bisal and Hsieh [1966] found that 4% moisture was enough to prevent particle entrainment, emissions likely occurred from the upper few millimeters of loess that were dry and persisted as the uppermost part of the loess dried during testing.

3.2 Character of Emissions

[18] Dust emissions in experiments on loess can be characterized as follows:

  1. [19] Direct suspension, supply limited (Figure 3a). No saltation occurs. High- to low-magnitude, temporary peaks in dust emissions occur at the beginning of each increase in friction velocity and rapidly decline to background levels. This type of response is common at lower friction velocities but does occur at higher friction velocities on crusted surfaces. These temporally limited peaks in dust are interpreted as direct aeolian entrainment on a supply-limited surface [Macpherson et al., 2008].

  2. [20] Saltation, supply limited (Figure 3b). Saltation precedes and drives dust emissions, but saltation and emissions both decline following peaks in both. This type of response is also interpreted as supply limited.

  3. [21] Saltation, sustained (Figure 3c). Saltation is sustained. Dust emissions rise rapidly following the beginning of saltation and maintain high concentrations of dust at a constant friction velocity. In this case, there is a sustaining supply of sand-sized grains for saltation, as well as an abundant supply of dust for suspension.

  4. [22] Direct suspension and saltation (Figure 3d). Dust emissions have a lower threshold than saltation; at steady, high friction velocity, dust is emitted without saltation. Following a 30 to 60 s time lag, saltation begins and increases. In some cases, as saltation increases, dust emission decreases. A possible explanation for this unexpected response is that dust concentrations are high enough not to be affected by an introduction of saltating particles. The time lag prior to initiation of saltation at a constant friction velocity could be due to the initial immobility of potential saltators that are embedded in fine material, but become mobile as that fine material is eroded.

Figure 3.

PI-SWERL step tests depicting the four main types of emissions responses on loess. (a) Direct suspension, supply limited (Orr); (b) Saltation, supply limited (Orr); (c) Saltation, sustained (Orr); (d) Direct suspension and saltation (Judkins). Horizontal black line differentiates noise (below) from true saltation (above).

[23] Weakly cemented silt-clay aggregates could supply directly entrainable silt. Based on particle size analysis, however, most aggregates prove to be relatively resistant in that they withstand several minutes of mechanical force from stirring and physiochemical dispersion in water. Therefore, it is unlikely that much silt would contribute to dust without the mechanical force of impacts.

3.3 Threshold Friction Velocities for Sand and Dust

[24] Two types of u*t were estimated with PI-SWERL ramp tests: u*t,sand for sand movement, recorded by sand movement sensors underneath the PI-SWERL chamber and interpreted as saltation, and u*t,dust for dust, noted as the point where dust concentrations measured by the DustTrak begin to rise from background levels (Figure 4). The average ambient levels of dust preceding PI-SWERL tests were typically about 0.04 mg m−3. Our PI-SWERL tests on Nebraska Sand Hills dune sand reveal the typical dust emissions response found in a great deal of previous research using field measurements or wind tunnel experiments, where the threshold for sand is met first, followed by dust emissions (Figure 4a and Table 2). In most cases, the thresholds for sand and dust are nearly identical because the emission of dust is dependent upon the movement of sand grains. At the least, the dust threshold is slightly greater than, but very near, the saltation threshold.

Figure 4.

PI-SWERL ramp tests measuring the potential of field and control plots to emit dust with or without saltation. Data are collected every second. The 20-point (or 20 s) moving average curve smooths the sand movement data and clarifies when the threshold velocity for sand movement (saltation) begins. (a) Barta dune sand depicts saltation prior to dust emissions; (b) Prairie Lake fine loess, where saltation enables dust emissions; (c) Wauneta coarse loess, where direct suspension of dust without saltation occurs; (d) Fanning coarse loess; (e) Broken Bow crusted fine loess with no saltation and minimal dust; (f) Wauneta crusted coarse loess with no to intermittent saltation and direct suspension of dust.

Table 2. Threshold Friction Velocities for Sand and Dust in Loess
SiteThreshold Dusta (m s−1)Threshold Sanda (m s−1)Notes
  1. aI = possible intermittent saltation, ND = no saltation detected, B = background level.
 
Field Sites: Benches
Foran 60.55ICoarse grained
Fanning 10.320.45Coarse grained
Fanning 20.290.41Coarse grained
Fanning 30.360.44Coarse grained
Fanning 50.42NDCoarse grained, crust
Orr 10.390.50Coarse grained
Murphy 10.550.42Coarse grained, crust
 
Plots: Loose, No Crust
Wauneta 10.26ICoarse grained
Wauneta 20.26ICoarse grained
Wauneta 30.26ICoarse grained
Judkins 10.26ICoarse grained
Judkins 20.26ICoarse grained
Judkins 30.27ICoarse grained
Broken Bow 10.260.41Fine grained
Broken Bow 20.290.39Fine grained
Broken Bow 30.290.44Fine grained
Prairie Lake 10.300.38Fine grained
Prairie Lake 20.300.35Fine grained
Prairie Lake 30.300.35Fine grained
Obert 10.270.33Fine grained
Obert 20.270.42Fine grained
Obert 30.270.38Fine grained
 
Plots: Crusted
Wauneta 1BNDCoarse grained
Wauneta 30.260.50Coarse grained
Judkins 10.30NDCoarse grained
Prairie Lake 1BNDFine grained
Broken Bow 1BNDFine grained
Obert 1BNDFine grained
 
Dune Sand
Barta 10.410.38Blowout
Barta 20.400.40Blowout
Barta G10.450.45Reactivated dune
Barta G20.420.42Reactivated dune

[25] Our experiments on loess produced quite different results. In almost all cases, u*t,dust was less than u*t,sand (Table 2). For plots of loose, disturbed loess, u*t,dust values fall within a narrow range (0.26 to 0.30 m s−1). The thresholds were more variable and generally greater for the in situ loess benches than for plots of loose, disturbed loess, likely due to variable compaction and weak cementation by carbonates or clay bridges in the in situ loess. For some loess, the saltation threshold was apparently never met, since saltation was not observed, yet significant dust emissions occurred (Figure 4c).

3.4 Dust Emission Potential: Coarse Loess

[26] For indoor tests on plots of loose, coarse loess (Wauneta, Judkins), relatively high concentrations of dust are emitted through direct suspension beginning at u*t,dust = 0.26 m s−1 (Figure 4c). Intermittent saltation does not begin until u* > 0.40 m s−1. Sustained saltation does not occur until after a 30 to 60 s lag when u* is held constant at 0.69 m s−1 (Figure 3d).

[27] Field benches of coarse loess (Foran, Fanning, Orr) are more variable in their responses, likely due to variable compaction and weak cementation, but the overall behavior is similar to that observed in loose coarse plots. Relatively high concentrations of dust are emitted through direct suspension prior to reaching the saltation threshold. At greater velocities, dust emissions at some sites are supply limited, with peaked emissions that drop rapidly to lower background levels or peaked emissions that drop to moderately high sustained emissions (Figure 3b). Other sites have intermittent or sustained saltation that probably assists in maintaining relatively high dust emissions (Figures 3c and 4d).

3.5 Dust Emission Potential: Fine Loess

[28] For indoor tests on plots of loose, finer loess (Broken Bow, Obert, and Prairie Lake), direct suspension of dust is again evident at velocities (u*t,dust = 0.26 to 0.30 m s−1) less than the saltation threshold (u*t,sand = 0.33 to 0.44 m s−1). The Broken Bow plot emitted the lowest concentrations of all testing sites, even when significant saltation occurred at greater velocities. In this case, there is no direct relation between saltation and increases in dust emissions. The Broken Bow tests may be considered an outlier. Prairie Lake, on the other hand, was the largest emitter of dust among all tests on plots and benches (Figure 4b). Saltation thresholds were relatively low, which may be related to greater abundance of saltating aggregates relative to primary sand grains. Some dust emission may be related to breakdown of saltating aggregates. Obert loess behaves similarly to Prairie Lake loess, although at the highest velocities, significant saltation occurs as dust emissions decline. Field benches of finer loess (Murphy and Ansley) had significant direct suspension of dust at velocities less than the saltation threshold. Saltation is more significant at greater velocities and likely contributed to dust emissions.

3.6 Dust Emission Potential: Crusted Loess

[29] For natural crusts tested in the field and crusts produced artificially by watering the indoor plots, dust emissions were of low magnitude, with short-lived peaks in dust that declined dramatically and in most cases saltation was not evident (Figures 3a and 4e). On finer-grained crusts, dust was a product of direct suspension in a supply-limited situation. Crusts remained intact for the duration of testing. For the coarser loess crusts, emissions were also peaked, but some tests on crusted indoor plots recorded concentrations of dust similar to uncrusted loess. Observations of the crusts following these tests revealed that crusts deteriorated at greater velocities resulting in high dust concentrations and intermittent saltation, especially evident for Wauneta (Figure 4f ).

[30] Shear vane tests (Figure 5) on artificially crusted loess revealed that crusts on coarser loess are significantly weaker than crusts on finer-grained loess, likely because of lower clay contents (Table 1). Coarser crusts are less cohesive and more susceptible to being broken apart under greater friction velocities, as we observed (Figure 4f). The finer loess with the higher clay content had stronger crusts that remained intact throughout the testing.

Figure 5.

Box plots of crust strength of loess in controlled plots. Boxes represent the 25th to 75th percentiles, and the whiskers capture the 10th and 90th percentiles. Top and bottom dots represent the maximum and minimum values, respectively. The finer-grained crusts in loess (Broken Bow, n = 18; Prairie Lake, n = 15; and Obert, n = 18) have stronger crusts that prohibit dust emissions compared to crusts in coarse loess (Wauneta, n = 14; Judkins, n = 18) that are susceptible to breaking apart during testing. A one-way analysis of variance multiple pairwise comparison using Dunn's method reveals statistical differences at the 99% confidence interval between weak crusts (Wauneta, Judkins) and strong crusts (Prairie Lake, Obert).

3.7 Dust Flux From Loess

[31] Dust fluxes are calculated from step test results. Dust emissions from plot and field tests overlap, indicating that our controlled tests adequately replicate conditions in the field (Figure 6a). Crusting tended to reduce dust emissions considerably (Figure 6b); the exception was for the coarsest loess (Wauneta and Judkins) that had weaker crusts. For uncrusted field benches or indoor plots, dust emissions were generally greater from finer loess. The best explanation for this relationship is simply that the sensor on the PI-SWERL measures PM10 dust and the finer loess has more material < 10 µm available to emit. There is a fairly strong positive correlation between PM10 dust emission and < 10 µm content of the loess, especially at greater friction velocities (Figure 7a). There is a negative correlation between sand content and PM10 emitted (Figure 7b). This is expected because the coarser loess has more sand, but it also shows that dust emissions are not related to the availability of sand for saltation as might occur where the sandblasting mechanism predominates.

Figure 6.

Dust fluxes from tests on (a) loose loess and (b) crusted loess. Field tests are graphed with solid lines; control plots are graphed with dashed lines. Results from field and control tests overlap, validating our research approach.

Figure 7.

(a) Percent PM10 measured by particle size analysis versus PM10 dust emitted, revealing that finer-grained loess has the largest potential to emit dust, and (b) percent sand in the loess versus PM10 dust emitted is negatively correlated, suggesting that dust emission is not primarily a function of saltation.

[32] Note that the relationships between loess particle size and dust emission rates do not affect our interpretation of the mechanism(s) of dust entrainment. Despite relatively low sand content in the Prairie Lake loess sample, it was clearly evident in PI-SWERL tests that saltation played a role in generating dust from the Prairie Lake plot. Similarly, although the sand content of the Wauneta sample was relatively high, saltation was not detectable above the background noise and direct entrainment was clearly the predominant mechanism of dust emission.

[33] As suggested above, the total flux of dust is not captured by the instrumentation of the PI-SWERL; only PM10 fluxes are estimated. Bacon et al. [2011] utilized a different type of dust monitor that measured total suspended particulate during PI-SWERL tests and found that the total amount of dust emitted was about 3.5 times higher than PM10 emissions on a variety of soil surfaces, including loess. In reality, dust fluxes from loess surfaces are likely at least 3.5 times higher than reported here.

4 Discussion

4.1 Mechanisms of Dust Emission From Loess

[34] Loess, if unvegetated or disturbed, has been shown to have high potential for dust emissions [Kjelgaard et al., 2004; Sharratt et al., 2007; Bacon et al., 2011]. The PI-SWERL data reported here suggest that dust emissions from bare loess surfaces can reach levels comparable to landforms that have the highest rate of emissions in the Mojave Desert [Sweeney et al., 2011]. Our data also indicate that these rates of emission are not dependent upon saltation, even for undisturbed, relatively cohesive loess deposits such as those on the benches; instead, substantial amounts of dust can be generated from loess by direct aerodynamic entrainment of silt. At higher wind velocities, saltation often does occur, mobilizing sand or sand-sized aggregates that the loess contains and facilitating higher dust emissions. A relatively high content of sand, however, is not required for high dust emissions from the loess we tested; in fact, in our data, dust emissions are negatively correlated with sand content. This relationship is quite different from what might be expected given the importance of saltation and sandblasting in other settings. For example, Sweeney et al. [2008, 2011] found that the largest natural emitters of dust in the Mojave Desert were landforms that contained appreciable sand for saltation. Wind tunnel studies have also shown a positive correlation between dust emission rates and horizontal saltation flux [Nickling and Gillies, 1989; Shao et al., 1993; Houser and Nickling, 2001]. In these studies, however, direct suspension was not a major contributor to overall dust emission.

[35] The threshold shear velocities for dust emission, as determined from PI-SWERL experiments on plots of loose, disturbed loess, are relatively low (0.26–0.30 m s−1; Table 2). The DustTrak sensor measures PM10 dust, but observations of the dust produced and the scouring of the bed under the PI-SWERL suggested that the medium-coarse silt and very fine sand particles that dominate the loess (modes of 50–85 µm, minimally dispersed, and 30–66 µm, fully dispersed) were mobilized once dust emission began. The thresholds for dust emission from these samples are actually quite close to those estimated for 30–66 µm particles by models of u*t based on the balance of forces on individual grains [Iversen and White, 1982; Marticorena and Bergametti, 1995]; therefore, it seems possible that when the loess is dry and loose, its dust emission threshold is related mainly to mobilization of the predominant coarser grains and is not greatly limited by the cohesion imparted by lesser amounts of clay and fine silt. Experiments in which dust coarser than 10 µm is measured are needed to test this interpretation. The higher thresholds for dust emission from in situ benches are not surprising given the likelihood of some compaction and weak cementation, but several tests on benches yielded u*t,dust values only slightly higher than the plots of loose sediment (Table 2).

4.2 Implications for Loess Landscapes as Dust Sinks and Sources

[36] To further evaluate the implications of these findings for loess accumulation and preservation, and the potential for loess landscapes to act as major regional dust sources, we first consider the wind speeds implied by PI-SWERL-based thresholds for dust emission and saltation. For a logarithmic wind speed profile, the relationship between u* values and wind velocities at 10 m height is expressed by the Prandtl equation [Prandtl, 1935]:

display math(2)

where uz is the wind speed (m s−1) at height z (m), κ is von Karman's constant (0.4), and zo is the aerodynamic roughness height (m). Unlike conventional wind tunnels, the PI-SWERL does not simulate the atmospheric boundary layer and therefore does not allow measurement of the wind speed profile and calculation of zo [Etyemezian et al., 2007], so we use a range of estimates. A theoretical minimum value for smooth surfaces of loess would be 1/30 of the predominant grain diameter, on the order of 1 × 10−6 m. Using zo = 1 × 10−6 m and the u* values for loess plots and benches in Table 2, equation (2) yields 10 m wind speed thresholds of 10.4 to 22.3 m s−1 for dust emission and 15.2 to 20.0 m s−1 for saltation. The range of effective zo for natural unvegetated loess surfaces is likely to extend well above this theoretical minimum, for example, Sweeney et al. [2008] reported values of < 1 × 10−6 to > 1 × 10−3 m in wind tunnel experiments on unvegetated, gravel-free desert surfaces. Using 5 × 10−5 m s−1 as a higher estimate of zo yields 10 m wind speed thresholds of 7.9 to 16.9 m s−1 for dust emission and 12 to 15.2 m s−1 for saltation.

[37] Hourly measurements of 10 m wind speed at Garden City, Kansas (1985–2005 data, NOAA National Climatic Data Center) provide a representative sample of modern winds on a nearly flat High Plains surface, comparable to loess-mantled tablelands near our study sites, though about 350 km to the south. In that data set, 20% of observations record winds > 8 m s−1, 9% > 10 m s−1, 3% > 12 m s−1 (or roughly 263 h yr−1), and 0.6% > 15 m s−1 (or roughly 53 h yr−1). Much higher wind speeds are occasionally reported from this region; for example, the National Weather Service reported winds of 26–33 m s−1 on 18 October 2012 that produced a major dust storm in the central Great Plains. Thus, wind speeds capable of producing dust from unvegetated loess surfaces do occur in the modern climate, even with the theoretical minimum of zo, and those corresponding to lower values of u*t,dust and zo = 5 × 10−5 are quite common. General circulation models representing the climate of the Last Glacial Maximum, when Peoria Loess was initially accumulating, indicate somewhat weaker mean surface winds than at present [e.g., Mason et al., 2011b, Figure 7]. On the other hand, it has been suggested that greater gustiness characterized glacial climates [McGee et al., 2010]. Winds exceeding the thresholds for dust emission and saltation could have been less frequent than at present, but it still seems likely that they would have occurred under conditions that commonly produce strong winds today, such as cold fronts associated with strong cyclones.

[38] With that background, we conclude that loose, freshly deposited Peoria Loess would have a high potential for resuspension under the modern wind regime and probably during the late Pleistocene, unless the effective u*t,dust was raised substantially by crust formation or vegetation cover. Postdepositional compaction and weak cementation would reduce the potential for resuspension, but the bench tests indicate that it could still occur on unvegetated and uncrusted surfaces under the strong winds that characterize the Great Plains. Soil development in the loess probably has significant, long-term effects on dust emission, but we did not investigate that issue.

[39] Two remaining factors, vegetation cover and crust formation, could have limited the potential for resuspension. Recent research has emphasized the importance of increased functional connectivity between bare patches in grass- or shrublands affected by land degradation, allowing greater aeolian erosion [Okin et al., 2009; Ravi et al., 2010], and there is a reason to believe that drought or other climatic effects can have a similar effect on Great Plains grasslands. In loess-mantled areas under natural grassland vegetation in the Great Plains today, fairly continuous vegetation covers much of the landscape in most years. However, during severe droughts in the 1930s, bare patches were common in these grasslands [Weaver and Albertson, 1936]. More extensive and better connected patches were probably exposed during more severe droughts of the Holocene [Miao et al., 2007]. Peoria Loess accumulated in a poorly understood late Pleistocene environment, which had not only some characteristics of open, semiarid steppes but also at least small patches of trees and large areas of unvegetated, active aeolian sand (see reviews of the paleoenvironmental evidence by Muhs et al. [1999], Bettis et al. [2003], and Mason et al. [2011b]). In both the Holocene and late Pleistocene, climatically driven shifts from grassland toward shrubland vegetation could also have enhanced wind erosion rates [Breshears et al., 2003].

[40] Rainfall quickly forms crusts on the small patches of unvegetated loess exposed today. PI-SWERL data indicate that crusts can almost completely prevent dust emission, but crusts on two plots of coarse loess broke down and allowed dust emission at relatively low u* (Table 2). Shear vane tests confirmed that crusts on those two samples had lower strength than crusts on finer samples, which has important implications if applicable to coarse loess in general.

[41] Peoria Loess accumulated at fast rates [Roberts et al., 2003] and forms a sheet of sediment that is of quite uniform local thickness on stable uplands and thins systematically away from source areas [Mason, 2001]. These observations suggest that a combination of at least patchy vegetation cover and rapid crust formation on unvegetated surfaces prevented resuspension of most Peoria Loess during and after the late Pleistocene. We propose, however, that the high dust emission potential of the loess could have been achieved under favorable conditions, especially near the upwind edge of the loess sheet. There, rapid accumulation and relatively low moisture retention in coarse loess may have led to especially sparse and/or patchy vegetation. Weak crusts on coarse loess would have broken down under wind stress more frequently than crusts on finer loess downwind. The upwind edge of thick Peoria Loess on the Great Plains often forms a high, windward-facing escarpment, which would have been present to some degree in later phases of accumulation. Wind acceleration over these scarps could have increased the potential for erosion on the windward slope. At its upwind edge, Peoria Loess also borders on extensive dunefields that were active in both the late Pleistocene and the Holocene [Mason et al., 2011b]. Encroachment of aeolian sand from these dunefields onto the loess could have inhibited vegetation cover and also more directly enhanced dust emission through sandblasting.

[42] A variety of factors make the source-proximal edge of the loess sheet a likely setting for resuspension and atmospheric dust loading. Consistent with that hypothesis, troughs and scarps that are apparently wind aligned (Figure 1b) occur mainly in this location within the loess-mantled landscapes of the Great Plains. We note that their spatial pattern clearly suggests that they record large-scale deflation, locally converting the loess sheet from a sink to a major source of dust. While this paper has focused on Peoria Loess of the Great Plains, it is likely that loess of similar grain size in northern China, central Europe, and other regions has similarly high dust emission potential. That potential would most likely have been achieved in near-source areas of especially coarse, thick loess.

5 Conclusions

[43] PI-SWERL tests on Peoria Loess reveal that in most cases, the threshold friction velocity for silt is less than that for sand, demonstrating the importance of direct entrainment of silt on bare loess surfaces. Dust emissions from bare loess surfaces via direct entrainment can be high in magnitude, and their potential importance goes beyond the contribution to ambient dust levels previously attributed to direct entrainment. In addition to the potential for direct entrainment of dust from loess, grain size analysis reveals an abundance of sand-sized, silt-clay aggregates in the loess that have the potential to saltate and contribute to or help sustain dust emissions. Surface crusts in finer loess are effective at minimizing wind erosion; however, crusts formed in coarser loess can quickly deteriorate during a sustained wind. These results demonstrate that loess landscapes can be major sources as well as sinks of dust, even without a local source of saltating sand to drive dust emission. The potential for direct entrainment of loess may explain large, wind-aligned ridges and troughs such as those observed in the thick loess of Nebraska, USA, and other loess regions. Furthermore, such large-scale remobilization of loess would have important implications for net loess accumulation rates and regional dust loading as the erosion occurred.

Acknowledgments

[44] This research was funded by National Science Foundation grants EAR-0921312 and EAR-0921915. We thank Zac Irvine for the assistance in the field; the South Dakota Geological Survey and Ann Kepler, manager of the Barta Brothers Ranch, for the logistical support; DustQuant, LLC and Vic Etyemezian for the PI-SWERL rental and technical support; and several landowners for study site access. We also thank Editor A. Densmore, the Associate Editor, reviewers Bill Johnson and Art Bettis, and an anonymous reviewer for the helpful comments.

Ancillary

Advertisement