Field measurements were performed in the gobi region of Milan, Southern Xinjiang, China (39°8.062′N, 88°59.967′E) (Figures 1A, B). The gobi surface was located on a plain of alluvial-proluvial gobi in northern front of the Altun Mountains. The gobi surface has a gravel coverage rate ranging from 40 to 50%, and is composed mainly of fine gravels (<6 cm), underlain mainly by fine sand (median grain size d ₅₀ 202 μm) (Figure 1C). The main sand-driving wind directions in the study area were ENE and E, and the annual sand drift potential (DP) in 2018 reached 1,160 vector units, indicating a high wind-energy environment. Detailed information on the aeolian environment has been described in Tan et al. (2020).

Sand particles in wind-blown sand over gobi during the three transport events were collected by a vertical array of Big Spring Number Eight (BSNE) traps (Fryrear, 1986), including seven traps at different heights (Figure 1D). The seven BSNE sand traps were placed at mid-inlet heights of 0.025, 0.175, 0.285, 0.585, 1.16, 1.79, and 3 m, and the direction of the spanwise layout was NW-SE, which was almost orthogonal to the main sand-driving wind direction (ENE) in the study area (Tan et al., 2020). Wind data used in this study were obtained from a meteorological observation tower with three 2D ultrasonic anemometers at heights of 2, 6, and 10 m, and recorded as 1 min averages, located approximately 150 m west of the saltation observation site.

All grain-size data of wind-blown sand samples in this study were measured by a Malvern Mastersizer-2000 grain-size analyzer. Grain-size statistics of mean, sorting, skewness, and kurtosis were calculated using the method of Folk and Ward (1957). Friction velocity u _* was computed from the logarithmic slope between the two instruments at the heights of 2 and 10 m using the following equation (Martin et al., 2013). u ∗ = k ( u 2 − u 1 ) ln ( z 2 / z 1 ) (1) where u ₁ and u ₂ were the wind velocities at the heights of 2 m (z₁) and 10 m (z₂).

Three sand transport events (20180502, 20180504, and 20180507, all the following descriptions regarding the three transport events are in this order) were observed during the fieldwork from 1 May 2018 to 8 May 2018. The variation of wind speed and direction at 10 m above the ground with time is shown in Figure 2. The measurement duration fell in between 110 and 420 min. The values of wind speed at a height of 10 m for the three transport events ranged from 8.35 to 19.27 m s⁻¹, 8.28–19.82 m s⁻¹, and 11.90–23.55 m s⁻¹, respectively, and the corresponding average wind speeds were 13.92, 13.70, and 18.78 m s⁻¹. At the same time, the values of wind direction at a height of 10 m for the three transport events ranged from 54.2° to 91.2°, 40.5° to 97.0°, and 19.5° to 75.3°;, respectively. The corresponding average wind directions were 80.6°, 69.7°, and 54.9°, which belonged to E, ENE, and NE winds, respectively. During the three transport events, the values of friction velocity u _* ranged from 0.11 to 0.79 m s⁻¹, 0.01–0.72 m s⁻¹, and 0.10–1.28 m s⁻¹, respectively, and the corresponding average values of u _* were 0.40, 0.28, and 0.66 m s⁻¹ (Figure 2). The measured threshold friction velocity for sand entrainment u _*t was 0.30 m s⁻¹ in the study area. During the transport event 20180504, the average u _* was smaller than u _*t, and intermittent aeolian sand transport occurred, especially in the period of 13:00–16:59 (Tan et al., 2020). In contrast, during the transport event 20180507, almost all the values of u _* were over u _*t in the whole measurement period, and strong aeolian sand transport occurred (see Figure 3A in the next part).

Vertical sand flux profiles show that the sand flux density decreased exponentially with increasing height during the three transport events (Figure 3A). The proportions of the transported sand particles caught at the lowest trap (with a mid-inlet height of 0.025 m) were 40.04, 36.14, and 34.39% for the three transport events, respectively. 90% of the total sand flux for the three transport events was concentrated within layers of 0.20, 0.23, and 0.25 m above the ground, respectively. The sand particles caught within a height of 0.585 m above the surface accounted for less than 1% of the total, and thus, 99% of the sand particles in wind-blown sand over gobi were transported in the layer of 0–0.585 m (Figure 3B).

Many previous studies, mostly through wind tunnel tests, show that the sand flux density varies with height in a non-monotonic curve, with a maximum occurring at a certain height above the ground (e.g., Dong et al., 2004; Qu et al., 2005; Tan et al., 2013; Zhang et al., 2014). In contrast, this study shows that sand flux density decreases exponentially with increasing height among all the three measured transport events. Furthermore, the field study of Tan et al. (2016) also reveals that the sand flux density decreases exponentially with increasing height, which is different from the results of previous wind tunnel studies. Meanwhile, the recent field study of Tan et al. (2021) shows that the particle flux density of wind-blown sand over gobi decreases exponentially with height, which is consistent with the exponential decay model of the sand flux density against height determined using BSNE traps. The difference in results of wind tunnel tests and field observation can be attributed to the fact that the blocking effect of gravels on sand transport near the gobi surface is not as pronounced in the field as that in the wind tunnel. The scale of gravels cannot be reduced in wind tunnel tests, which, in turn, increases their blocking effect, while in the field, the relief of the terrain makes the function of gravels in inhibiting near-surface wind-blown sand less obvious than that in the wind tunnel.

This study shows that 99% of the sand transport amount over gobi is concentrated in the near-surface layer of 0.6 m. Comparatively, it shows that 99% of the sand transport amount over gobi atop the Mogao Grottoes is concentrated in the near-surface layer of 0.3 m (Figure 6), and thus the height of the saltation layer in this study is nearly two times of that atop the Mogao Grottoes. However, in the Hundred Miles windy area along the Lanzhou-Xinjiang High-Speed Railway, 99% of sand particles are transported in the layer of 3 m above the gobi surface (Wang et al., 2022). This can be attributed to the difference in the wind speed and grain size, and Zhang et al. (2017) revealed that the cumulative sand flux is controlled more by the mean grain size of sand surfaces than wind speed. For example, the maximum mean grain size of wind-blown atop the Mogao Grottoes is 150 μm, while that of in this study is 220 μm, and larger saltating particles can reach higher in the process of aeolian saltation over gobi. In contrast, sand transport on sand surfaces is mainly concentrated in a height range of 0–0.2 m. This indicates that the grain-bed collision over gobi facilitates aeolian saltation by increasing saltation heights, in which sand particles can retain a higher proportion of their impact energy. In addition, Nield and Wiggs (2011) revealed that compared with dry, rippled sand surfaces, the saltation cloud height increased over slightly damper and harder sand surfaces.

Vertical mean grain-size profiles of gobi wind-blown sand show that the mean grain size increases with height above the ground until a reversal occurs, and then the mean grain size decreases with height. It indicates that the coarsest grains are found at a finite height above the ground rather than on the surface. The field work atop the Mogao Grottoes also revealed a reversal in the variation of grain size with height, with an inflection occurring at 0.22 and 0.24 m for the two transport events of 20130415 and 20130417, respectively (Tan et al., 2016). Below the inflection, the mean grain size also increased with height; however, above the inflection, it showed a linear decrease in mean grain size with height, in which only 0.3 m above the inflection had been measured, while in this study an approximate 0.6 m transport layer had been considered (Figure 5A). The reason is that the study area of Mogao Grottoes belongs to a low wind-energy environment (Zhang et al., 2014), while the study areas in this study indicates a high wind-energy environment (Tan et al., 2020), resulting in a relatively thicker saltation layer in this study. Zhang and Dong (2014) also used BSNE traps to examine characteristics of grain-size distribution for an artificial gobi bed; however, only four traps at 0.25, 0.5, 1, and 2 m above the ground had been used. As a result, they showed that the mean grain size decreased monotonically with height, yet the change in mean grain size with height could not be examined below 0.25 m. The field study of Sharp (1964) revealed that the mean grain size for wind-blown sand over a bouldery alluvial plain in Coachella Valley, California, United States, achieved a maximum at heights ranging from 0.25 to 1 m above the ground. Wang et al. (2022) revealed that the inflection point in the vertical median grain size profile of wind-blown sand over a gobi surface in the Hundred Miles windy area, China reached 2 m above the ground surface. The heights where the inflection occurred in these two studies were larger than those of this study. This may be caused by the difference in wind power and the grain-size composition of sand sources for wind-blown sand over the rough surface.

Apparently, the vertical mean grain-size profile over gobi is significantly different from those of sand surfaces, in which the dominant profile pattern is grain size first decreases till to an inflection and then increases with elevation. This can be attributed to the difference in grain-bed collisions between gobi beds and sand surfaces. The collision between saltating particles and gravels on the gobi surface is nearly elastic, and thus, saltating sand particles will rebound and maintain a higher proportion of momentum; however, those of sand surfaces will transport momentum to new particles and the momentum has been consumed mainly on bed deformation. Liu et al. (2021) reveals that for the same gobi bed and at the same experimental particle speed, the larger the sand particle size, the higher the bounce height of sand particles after colliding with gravels. This is because gravity and the drag of the wind are the main influencing factors for the trajectory of rebounding sand particles, and as grain size increases, the ratio of inertia to drag force becomes larger. As a result, larger grains can travel in higher and farther trajectories than smaller grains. Figure 7 shows the variation of frequencies for three grain-size groups (very fine, fine, and medium sand) with height in the three transport events. The results demonstrate that below the height of the inflection point, both the frequencies of very fine and fine sand in the three transport events decreased with height (Figures 7A, B), while that of the medium sand increased with height (Figure 7C), resulting in the increase of the mean grain size with height. In contrast, above the inflection points, the frequency of very fine sand increased (Figure 7A), while that of the medium sand decreased (Figure 7C), resulting the decrease of the mean grain size with height. This indicates that larger grains like medium sand particles participate in the grain-bed collision process over gobi and the rebound medium sand particles can reach a higher layer. Comparatively, very fine sand is more notably affected by turbulence, and the frequency only increases in the upper saltation layer where wind momentum is less affected by saltating particles. However, in regard to sand surfaces, bed deformation consumes more energy of saltating grains, resulting in the reduction in the available energy for grain ejection. Larger grains are relatively more difficult than smaller grains to reach the same launch speed, and thus larger particles will move close to the bed in reptation rather than saltation (e.g., Rice et al., 1995; Farrell et al., 2012). Consequently, the coarse fraction of the transport sand population is distributed within near-bed elevations, while the finer fraction is at higher elevations. Thus, the mean grain size decreased with height in the saltation layer over sand surfaces.

Studies of quantifying the variation of sorting, skewness, and kurtosis with height are relatively few for sand surfaces. Williams (1964), Li et al. (2008), and Farrell et al. (2012) found that sorting improved with height. Zhang et al. (2017) revealed that sorting of sand surfaces composed of fine or medium sands improved with increasing height, while both profile patterns of sorting improving and declining with height presented for coarse sand surfaces. However, the sorting of gobi wind-blown sand in this study shows a decreasing trend with increasing height, which is consistent with the result of Tan et al. (2016).