The experiment was conducted in the summer of 2023 on farmlands near Washington in the Columbia Plateau, where semi-arid climate dominates and soils are composed of loess and sand deposits. In this study, we collected soil samples of two contrasting soil types representative of the Columbia Plateau. The sampling plot of Warden sandy loam (Mesic Xeric Haplocambids) was near Paterson, Washington (46°10′N, 119°37′W) with annual precipitation of 200 mm. The sampling plot of Ritzville silt loam (Andic Aridic Haplustoll) was near Ritzville, Washington (47°80′N, 118°28′W) with annual precipitation of 280 mm. Both sampling plots were subject to a winter wheat-summer fallow rotation. During the fallow phase of the winter wheat-summer fallow rotation, each plot was managed by wheat growers employing different tillage practices to control weeds and preserve soil moisture. Warden sandy loam has a higher sand content (67.52%) and lower clay content (9.21%) than Ritzville silt loam (with sand content of 30.53% and clay content of 11.42%). Dispersed particle size analysis using a Malvern Mastersizer laser diffractometer (Mastersizer, Malvern Instruments, Malvern, the United Kingdom) indicated that Warden sandy loam had a geometric mean diameter of 36 mm while Ritzville silt loam had a mean diameter of 20-29 mm (Sharratt and Vaddella, 2014).

Five sampling points were set up using the 'S' soil sampling method in each plot. Soil samples collected from the top 1 cm of soil from multiple sampling points in each plot were processed to obtain the erodible portion. Soil samples were air-dried at 70.00°C and then hand-sieved to collect particle size fractions of <45, 45-63, 63-90, 90-125, 125-150, 150-250, and 250-500 μm. This study used a portable wind tunnel (working section: 1.00 m width, 1.20 m height, and 7.30 m length) to produce free-stream wind speeds of 2-20 m/s to evaluate the TFV (Fig. 1). The test section featured plexiglass windows and removable metal trays. The bottom of wind tunnel was lined with sandpaper, mimicking the roughness of the soil surface in the trays. The wind tunnel was located inside a nonregulated climate facility, where relative humidity and temperature varied from 37.42% to 45.23% and 2.32°C to 13.11°C, respectively, during the experiment. The wind tunnel conditioned winds to resemble naturally occurring shear flow characteristics in the field. Wind was conditioned by flowing through a perforated plate to control vortex and velocity fluctuations, a honeycomb frame and wire mesh to reduce turbulence, and then non-uniform mesh components to create shear flow (Pietersma et al., 1996). The wind tunnel floor was built from wood coated with fine sand to simulate the boundary layer features of the natural bare soil surface. A 20 mm deep soil tray was recessed into the wooden floor 5.00 m downwind from the grill component so that the tray top remained level with the wooden floor. The soil trays and wooden floor were constructed with precision to minimize any gap between the soil trays and wooden floor. Soil composed of a specific size fraction of particles was placed in the 1.00 m×0.20 m×0.02 m metal soil tray and then the top of the tray was smoothed with a metal screed.

Wind speed was measured utilizing a pressure transducer connected to a pitot tube, which was mounted at six different heights (5, 10, 20, 30, 50, and 100 mm) above the soil surface at the leeward edge of the soil tray. Wind speed was corrected for atmospheric pressure and temperature that varied during the experiment. Sensor technology was used instead of visual observation to measure particle emissions from the soil tray because the PM emitted from the soil surface was too small and not visible to the naked eye. Simultaneous observations of PM10 and total suspended particulate (TSP) concentrations (mg/m³) were made at the leeward edge of the soil tray using the DUSTTRAK Aerosol Monitor (Model 8520, TSI Incorporated, Shoreview, USA) and E-sampler (Met One Instruments, Inc., Grants Pass, USA), respectively. Particulate sensors were mounted at the same height above the soil surface as wind speed sensors similar to previous studies (Copeland et al., 2009). The emission of soil particles from the tray as a function of wind speed was also monitored using a Sensit (Model H11-LIN, Sensit Company, Portland, USA). The Sensit was installed at a height of 50 mm above the soil surface. The entrance of the tunnel construction section was also equipped with particulate monitors for measuring background concentrations of PM10 and TSP. It was maintained that ambient dust concentrations in the wind tunnel facility were consistently low throughout the wind tunnel experiment. All sensors were programmed to record experimental parameters every second.

The TFV for each soil particle size fraction was determined by detecting the rise in PM10 and TSP concentrations downwind of the soil tray over background concentrations within a period of several seconds. In the wind tunnel, the TSP concentration remained essentially constant as wind speed increased. When wind speed reached the TFV, PM10 and TSP concentrations exceeded background concentrations at several heights. The TFV was measured by systematically raising the wind speed every 15 s until we observed an increase in PM10 or TSP concentration above the background for multiple seconds. This procedure ensures that a critical shear stress is reached, which allows particles or agglomerates to be released from the soil surface. This method of TFV determination (Sharratt et al., 2013) differs from visual inspection, which is probably more appropriate for detecting apparent movement of larger particles on the soil surface (Gillette and Passi, 1988). In some instances, saltation activity can facilitate assessing the TFV. We observed no saltation activity using the Sensit in this study. Once we observed that PM10 or TSP concentration rose above background concentrations during a 15 s period, we continued the experiment for 15 s to validate our observations. The experiment was replicated four times with each soil type and particle size fraction. The water potential of the soil surface in the tray was measured immediately before the soil was exposed in the wind tunnel and the wind speed reduced after each experiment. Although the duration of each measurement was short (about an average of 300 s), this method obtained the range of water potential for each soil type and particle size fraction during the experiment. Soil water potential was measured using either a dew point meter (WP4T, Decagon Devices, Pullman, Washington, USA) or filter paper technique (ASTM D5298-10, 2010). Fawcett and Collis-George (1967) verified that the filter paper technique is reliable and accurate. The soil in the soil tray was discarded after each replication. Wind speed profile parameters were calculated using the wind speed data collected at the six heights. Wind speed profile was fitted according to Prandtl-von Karman's logarithmic law (Roney and White, 2006), and friction velocity and aerodynamic roughness length were experimentally determined from the wind speed profile based on the following equation:

where u_z is wind speed (m/s) at height z (m); u_* is the friction velocity (or shear velocity) (m/s); k is the von Karman's constant (k=0.4); and z₀ is the aerodynamic roughness length (m). Aerodynamic roughness lengths for different particle sizes were determined by plotting the natural logarithm of height against wind speed (Dong et al., 2002). The optimal fit was determined through regression analysis using the least squares method and expressed mathematically as:

where a and b are the regression coefficients, which are used to define the aerodynamic roughness length according to:

Combining Equations 1-3, friction velocity can be calculated as:

PM10 and TSP concentrations above background concentrations were typically found within 180-300 s of generating winds inside the wind tunnel. Our procedure attempted to eliminate errors introduced by contamination or singular or 1 s transient increases or peaks in PM10 or TSP concentration due to the emission of perched particles on the soil surface. Approximately 60 s was required to install the soil tray in the wind tunnel floor and to ready all instrumentation before generating winds to determine the TFV in a wind tunnel. The TFV (m/s) was determined from the friction velocity within 15 s after the PM10 or TSP concentration exceeded the above-background value.

Figures 2 and 3 show the variations in PM10 and TSP concentrations with particle size fraction for Warden sandy loam and Ritzville silt loam. During this experiment, PM10 and TSP concentrations changed dramatically with wind speed as the TFV was attained. Specifically, when wind speed exceeded TFV, Warden sandy loam showed the highest TSP and PM10 concentrations for particle size fractions of 63-90, 90-125, and 125-150 μm. The highest TSP and PM10 concentrations for Ritzville silt loam were observed in particle size fractions of 125-150 and 150-250 μm, respectively. This suggested that particles in size fraction of 63-150 μm for Warden sandy loam and in size fraction of 125-250 μm for Ritzville silt loam were most vulnerable generating PM10 emissions. Notably, PM10 and TSP emissions from Warden sandy loam were significantly higher than those from Ritzville silt loam when particle size was below 150 μm. This suggested that PM10 attached to Warden sandy loam tends to be released more rapidly than that attached to Ritzville silt loam under high shear stress or as a result of saltation bombardment. This difference may stem from the varying soil textures, as Warden sandy loam has a higher sand particle content (67.52%) than Ritzville silt loam (with sand content of 30.53%). Soils rich in sand content often lack aggregates, making it easy for fine particles to detach from the sand grains, thus becoming a significant source of PM10 emissions. Indeed, sandy soils with weak cohesive characteristic are more fragile and likely susceptible to bombardment as compared to the silty soils with stronger bonding characteristic. Additionally, soils with higher sand content show excellent water permeability and aeration. The absolute values of the rate of change in water potential of Warden sandy loam were significantly higher than those of Ritzville silt loam (Table 1). This indicates that sandy loam has stronger hydrophobicity and weaker water film binding force. Therefore, we assumed that both low particle forces and low aggregation can increase TSP emissions significantly.

Soils of different textures exhibited significant differences in PM10 and TSP emissions across varying particle size fractions. The PM10 and TSP emissions from Warden sandy loam were significantly higher than those from Ritzville silt loam in most particle size fractions. However, in size fraction of 150-250 μm, the PM10 and TSP emissions from Ritzville silt loam were higher than Warden sandy loam. Notably, unlike Warden sandy loam, which has a higher sand content, aggregate fragmentation became the dominant process in the clay-rich soil of Ritzville silt loam. Interestingly, the highest PM10 and TSP emissions were found in size fraction of 125-150 μm for both soils. This may indicate that, compared to soils with smaller or larger particle sizes, for soils with medium texture, the detachment of dust attached to the soil surface, the fragmentation of aggregates, and the bombardment of saltating particles may collectively lead to continuous emissions of PM10 and TSP. Therefore, we hypothesized that the soil particles in size fraction of 125-150 μm were the predominant source of high PM10 concentrations in the Columbia Plateau (Fig. 4). This provides a new perspective for further research and controlling environmental pollution in the region.

Although the Sensit was used to detect potential saltation activity, there was a lack of saltation activity in our wind tunnel experiment. This is probably due to the limited size of the soil tray and the sensitivity constraints of the sensors, rather than the lack of saltation activity. Suspension at low wind speeds is generally considered a minor component of total dust flux because surface wind speeds are probably insufficient to overcome the interaction forces between fine particles, including capillary forces. In case of PM10 emissions from finer soils, we postulated that the suspension mechanism holds greater significance. Finer soil particles are more easily lifted by wind and suspended in the air, ultimately contributing to PM10 emissions. Conversely, in coarser soils, the generation of PM10 is predominantly driven by processes such as bombardment and abrasion, marking a distinct contrast in PM10 generation mechanisms. However, we cannot confirm that direct suspension will cause an increase in PM10 and TSP emissions. Although particle size and number can affect saltation activity, Van Pelt et al. (2009) found that the Sensit was not sensitive to particles smaller than 100 μm in diameter except at very high wind speeds. Moreover, Sharratt (2011) suggested that the Sensit lacked sensitivity for saltating particles smaller than 100 μm in diameter that can be commonly found in windblown dust in the Columbia Plateau. Thus, saltation may have occurred but was not detected with our instrumentation.

Figure 5 illustrates the wind speed profiles of the two representative soil types in varying particle size fractions. The good fit (R²>0.918) of the wind speed profile data provided confidence in estimating the TFV and aerodynamic roughness length. The TFV of Ritzville silt loam and Warden sandy loam ranged from 0.304 to 0.844 and 0.249 to 0.739 m/s, respectively, across all particle size fractions. It can be found that TFV is related to the particle size, particle shape, and microstructure of soil surface (Wang et al., 2008). The highest TFV for Ritzville silt loam occurred in size fraction of <45 μm while the highest TFV for Warden sandy loam occurred in size fraction of 45-63 μm. The increase trend in TFV with soil particle size fraction was significantly different between the two soil types.

The TFV of the two soil types in different particle size fractions is shown in Figure 6. For size fraction of 45-63 μm, TFV differed between the two soil types. The TFV was significantly higher for Warden sandy loam (0.749 m/s) than for Ritzville silt loam (0.526 m/s). This difference in TFV may be related to differences in cohesive forces between soil aggregates (Greeley and Iversen, 1985; Colazo and Buschiazzo, 2010). In size fraction of 125-150 μm, Warden sandy loam had a lower TFV than Ritzville silt loam. In fact, the lowest TFV for both soil types occurred in size fraction of 250-500 μm. The TFV in size fraction of 250-500 μm was 64.01% and 66.27% lower than that in size fraction of <45 μm for Ritzville silt loam and Warden sandy loam, respectively.

In our study, exponential functions adequately described the relationship between TFV and particle size fraction (R²≥0.867), which can be used to estimate the response of TFV to particle size for the two soil types (Fig. 6). The standard errors of the fitted equation for both soil types ranged between 0.03 and 0.09 m/s. Indeed, estimating the TFV under known smooth, dry, fine-grain, loose, and bare soil surface conditions can be very effective. Accordingly, as soil particle size increased from 45 to 500 μm, TFV decreased for both soil types. Significant differences of TFV between the two soil types were observed in the same particle size fraction after saltation, because particle emission may also occur before the soil tray is placed in the wind tunnel, which corresponds to the static threshold observed during particle shaking. After the saltation process, Ritzville silt loam had the largest decrease in TFV when the particle size fraction varied from <45 to 45-63 μm. In contrast, Warden sandy loam had the largest decrease in TFV as the particle size fraction varied from 45-63 to 63-90 μm. These observations indicated that the largest differences in TFV can be found for particles smaller than 100 μm in size for the two soil types.

Aerodynamic roughness length reflects the weakening effect of the soil surface on wind speed as well as wind and sand activity (Dong et al., 2002). It is related to the size of the particles or aggregates on the soil surface. As the aggregates or particles on the soil surface get larger, the aerodynamic roughness length increases. Figure 7 shows that the aerodynamic roughness length increased with soil particle size for the two representative soil types. The aerodynamic roughness length of Ritzville silt loam and Warden sandy loam increased by 0.043 and 0.049 mm, respectively, as the particle size increased from the smallest size fraction (<45 μm) to the largest size fraction (250-500 μm). The results indicated that as the particle size increases, the destructive force of the soil surface against the airflow increases, resulting in a rougher surface that can better trap and immobilise the soil particles, thus reducing wind erosion.

Etyemezian (2019) indicated that aerodynamic roughness length can affect the TFV. As shown in Figure 8, even within a narrow range of aerodynamic roughness length, TFV exhibited noticeable variations. Specifically, the intercept of the relationship between TFV and aerodynamic roughness length was -1.151 m/s for Ritzville silt loam and -0.742 m/s for Warden sandy loam, which indicated that TFV is greater for Warden sandy loam than for Ritzville silt loam with very little aerodynamic roughness length. Additionally, Figure 8 reveals a negative correlation between aerodynamic roughness length and TFV, indicating that a decrease in aerodynamic roughness length leads to an increase in TFV. This result may seem counterintuitive, as the TFV derived in this study focused on inhalable PM emissions from isolated soil particles of different size fractions, rather than the traditional soil erosion threshold. From a physical perspective, friction velocity is determined by the momentum transferred by the surface with a certain geometric roughness after absorbing the shear stress of the boundary layer flow. Therefore, changes in wind momentum absorbed by the erodible components of the soil surface will cause a change in friction velocity. The aerodynamic roughness length generated by the surface geometric roughness is the main factor affecting the friction velocity. This study revealed that finer soil fractions correspond to smaller aerodynamic roughness length.

In arid and semi-arid areas, the impact of wind erosion on soil is a crucial factor shaping the landscape and biological potential of the surface (Shao, 2000; Chandler et al., 2004; Ebrahimi-Khusfi and Soleimani Sardoo, 2021). The size and stability of soil aggregates directly determine the sensitivity of soil to wind and water erosion (Kheirabadi et al., 2018; Yang et al., 2022; Zhu et al., 2022). This study found that the TFV of isolated particles in different size fractions was higher than previously reported results in the Columbia Plateau. For instance, Sharratt and Vaddella (2012) found that the TFV of Warden sandy loam was 0.139 m/s and the TFV of Ritzville silt loam was 0.180-0.239 m/s. This phenomenon may be attributed to the special properties of the soil particles in this study. The soil particles obtained through fine screening are smaller and have a uniform particle size distribution. This significantly reduces the gaps between particles, resulting in a compact structure. Under the same wind speed conditions, this compact structure makes it more difficult for the soil particles to be blown away by the wind, leading to higher TFV. Furthermore, experiments of Bagnold (1941) provided valuable insights, that is, even if the wind speed is sufficient to move fine stones with a diameter of 4600 μm, it is still difficult to blow finer cement sand in a stable airflow. The relationship between TFV and particle size is complex and not directly proportional. Larger particles generally require higher wind speeds to be blown away due to their increased mass and inertia, but other factors also play a role. For example, particle shape, soil moisture, and the presence of other particles can all affect the TFV (Tanner et al., 2018; Pi et al., 2020b). In our study, the particle size on the soil surface is relatively uniform, and a very smooth surface is formed after ironing treatment. This treatment results in a more uniform distribution of resistance across the entire soil bed, which has a profound impact on TFV measurement results. Additionally, the lack of particles susceptible to direct wind influence on the soil surface may also contribute to the higher TFV observed in this study.

Meanwhile, soil wind erosion is a complex phenomenon since it involves numerous interacting factors (Kjelgaard et al., 2004; Sundram et al., 2004; Zamani and Mathmoodabadi, 2013). Shao (2008) indicated that particle size distribution, soil wetness, soil surface roughness, aboveground biomass, and crust cover influence the TFV. The interaction of these factors determines the sensitivity of soil to wind erosion. Especially on dry and loosely sand surfaces, TFV is mainly influenced by the particle size of sand grains. If the particles are small enough, they will adhere well upon contact even when completely dry, and particularly in a vacuum. The TFV of Ritzville silt loam and Warden sandy loam ranged from 0.304 to 0.844 and 0.249 to 0.739 m/s, respectively, across particle size fractions in this study. We speculate that the difference in TFV between different particle size fractions of the two soil types may be partially attributed to variations in soil particle cohesion. Especially in fine-grained soils, cohesion is deeply influenced by intermolecular forces, including van der Waals forces, electrostatic forces, and capillary forces. Due to the large specific surface area of fine-grained soils, these intermolecular forces become more significant. In addition, different soil particle sizes have variable degrees of water repellency or hydrophobicity, with finer-grained materials requiring more suitable soil moisture (Shattatt, 2007; Ishizuka et al., 2008; Ravi et al., 2016). Interparticle capillary force is the primary factor resulting in increased TFV when soil moisture increases, and fine-grained soils retained more water than large-particle soil aggregates. We observed an average decrease of 8.99% in water potential for the two soil types in size fraction of <45 μm, while an average decrease of 33.32% was observed in size fraction of 250-500 μm. Interparticle cohesion plays a crucial role in the wind erosion of fine-grained materials (Colazo and Buschiazzo, 2010). As a result, moisture bonding and soil cohesion between soil particles in size fraction of <45 μm is much greater than that in size fraction of 250-500 μm, resulting in higher TFV.

Wind erosion generates dust, which serves as a significant source of atmospheric aerosols (Zhang et al., 2022; Gao et al., 2023). Agricultural practices that denude or disturb the soil in arid areas can therefore lead to environmental degradation. Potential dust emission is a complex yet crucial topic. According to previous studies, PM10 emissions increase with silt and clay content and decrease with sand content, and the difference in PM10 emissions may be related to variations in soil organic matter content, aggregate size distribution, and aggregate stability (Funk et al., 2008; Avecilla et al., 2016; Zhang et al., 2021a). This paper examined two representative soil types, unveiling a profound correlation between distinct soil particle size fractions and PM10 and TSP emissions. Our findings revealed that both PM10 and TSP emissions attain their apex for the two soil types when soil particle size falls between 125 and 150 μm. This suggested that a substantial amount of fine PM becomes suspended during the wind erosion process. It is widely accepted that the interaction between saltation components and soil surface largely depends on their composition, aggregation state, or strength during the wind erosion and dust emission processes. Our results align with this consensus, emphasizing the crucial role of high PM10 concentration and low aggregation in determining dust emission rates, especially highlighting the importance of soil particles in size fraction of 125-150 μm in dust generation. Once these complex particles or aggregates are set in motion by wind force, they can exhibit distinct behaviors in terms of PM10 emissions.