Assessment of Threshold Wind Velocities of Industrial Granular Materials: A Comparative Evaluation of Experimental Methods

Abstract

To maintain a high standard of environmental quality, industrial plants must be able to foresee and control the impacts resulting from their activities. One of the most challenging issues for the metallurgical and mining industry when it comes to protecting the environment is the measurement of particulate matter emissions generated by the wind action over the erodible surfaces of stockpiles of granular materials. It is known that the emissive phenomenon starts from a specific threshold friction velocity, which is an inherent characteristic of each material. This parameter can be derived from relationships available in the scientific and technical literature, which, however, only provide qualitative estimations. Therefore, the threshold friction velocity of the specific materials under investigation must be assessed through laboratory tests. This article discusses the results obtained for nine raw materials sampled in a metallurgical plant by applying three different procedures, (1) the sieve-based analysis suggested by U.S. EPA; (2) the laboratory tests performed with an Environmental Wind Tunnel; and (3) the PI-SWERL tests (i.e., tests performed with a Portable In-Situ Wind ERosion Lab), and presents a comparative analysis of the three methods. Findings indicate that the EPA methodology tends to be less accurate than the wind tunnel and PI-SWERL tests, though its accuracy can be slightly improved by adding an additional sieve size for materials with finer aggregates. The wind tunnel and PI-SWERL provided comparable results, with PI-SWERL offering practical advantages due to its portability and an effective synchronization between its data acquisition systems.

1. Introduction

One of the main issues associated with mining, quarrying, and industrial activities is the emission of particulate matter (PM), which may cause significant impacts on human health and the environment [1,2,3,4]. PM emissions may originate from both point sources (i.e., conveyed sources typical of industrial plants) and fugitive dust sources, such as traffic of vehicles on unpaved roads, material handling and storage within the production sites, and industrial wind erosion. Particular attention should be given to industrial wind erosion, i.e., emissions generated by the wind erosive action over the erodible surfaces of granular materials, which are typically stored in open yards of industrial plants, such as raw materials, waste, intermediate and finished products.

The emissions of PM by wind erosion on flat surfaces depend on the speed of the incident wind, which acts on the exposed surfaces of the material piles (specifically, the shear stress or friction velocity, u*), and on the physicochemical properties of the material under consideration. These include: the particle size distribution and the specific weight of both individual particles and aggregates, the interparticle cohesive forces, and the surface roughness [5]. These variables determine the material threshold friction velocity (u*_t), which represents the resistance offered by the erodible surfaces to the erosive action: when the threshold friction velocity is exceeded, the lifting and transport of particles in the air is triggered [6,7,8].

Particles can be transported into the atmosphere through direct entrainment (i.e., direct lifting into the air) or saltation [9,10,11]. However, the direct entrainment is generally considered negligible, as particles with diameters smaller than 20 microns exhibit strong cohesive forces, making them hard to find in a non-coherent state on wind-exposed surfaces. In contrast, PM₂₀ particles are often found adhered to larger particles as coatings [12] or embedded within aggregates ranging from 20 to 300 microns in size [13,14,15]. In this aggregated state, the particles are involved in the saltation phenomenon, and, particularly those between 70 and 500 microns, are lifted by the wind and transported for short distances to fall back to the surface. Upon impact, the bouncing motion causes the disintegration of both the hopping and surface aggregates, releasing fine particles and leading to the emission of PM into the atmosphere [5,13,15,16,17].

Taking into account these dynamics, the wind speed determining the onset of dust emissions corresponds to the threshold friction velocity required to initiate saltation [16]. Therefore, determining this parameter is particularly important for proper implementation of effective containment and mitigation measures.

The scientific literature offers several empirical formulations that relate the threshold friction velocity to the physical properties of the exposed material, such as the particle size distribution of the elementary particles, the aggregate size distribution, the moisture content, the roughness of the exposed surfaces, and other related factors [6,7,8,18,19,20,21,22]. Based on these relationships, and in particular on the correlation between the Aggregate Size Distribution and u*_t [8], the U.S. Environmental Protection Agency (EPA) has developed an experimental procedure that allows for the estimation of the threshold friction velocity through manual sieving of the materials under examination and associating the u*_t with the mode of the aggregate size distribution [23]. Although such methods can provide useful indications, the erodibility and emissivity of granular materials result from a complex set of interconnected factors that are difficult to represent through general formulations or procedures.

To overcome these limitations, experimental tests can be performed by using facilities that simulate the wind action over the surfaces of the granular materials. The Environmental Wind Tunnels (EWT) represent the most established solution for this type of test and have been used over the years both for the development of the empirical models and for the characterization of specific materials [24,25,26,27,28,29,30,31,32].

An alternative device is the Portable In-Situ Wind ERosion Lab (PI-SWERL), which was developed by the Desert Research Institute [33,34] to facilitate in situ emission measurements [35,36]. The PI-SWERL simulates the shear stress acting on the surface of granular materials through the rotation of a blade positioned approximately 5 cm above the ground. In recent years, this instrument has also been used to carry out laboratory tests on samples previously collected on site, similarly to the EWT [37,38,39,40,41].

This study aims to compare the above mentioned different experimental methodologies for determining the threshold friction velocity on the samples of granular materials collected from a major industrial site that operates in Sardinia (Italy). This paper describes the testing facilities, specifically the EWT designed and built in the laboratories of DICAAR (Department of Civil, Environmental and Architectural Engineering, University of Cagliari, Italy) and the PI-SWERL provided by the Physical Geography and Environmental Change Group at the University of Basel (Switzerland), as well as the procedure for the indirect estimation of the threshold friction velocity suggested by U.S. EPA. The results obtained through the application of the three different methodologies are hereby presented and discussed, aiming at assessing their consistency, applicability, and limitations through a critical and comparative evaluation.

2. Materials and Methods

2.1. Materials

For the experimental tests, nine materials stored in a metallurgical plant located In Sardinia (Italy) were sampled. The samples include three different types of lead sulphide or galena (MS, PN, and PR), two types of zinc sulphide or zinc blende (PR and RD), two types of Pb/Ag sulphates (AD and BD), mixed oxides (WL), and gypsum (GP). Prior to the experimental testing procedure, each material was characterized by determining the water content, the specific weight of the aggregates, the particle size distribution (PSD), and the aggregate size distribution (ASD).

The water content was estimated by drying the samples for 24 h, at 45 °C for gypsum GP (as per ASTM C471M -13 [42]) and at 105 °C for all other materials, and then calculating the percentage ratio between the mass of water evaporated in the oven (m_w) and the mass of dried material (m_d), according to Equation (1):

$$w={\displaystyle \frac{m_w}{m_d}}.$$

(1)

The measurement of the specific gravity of aggregates was carried out using the pycnometer method on oven-dried material with a size between 100 and 75 µm, in order to relate the indication of the specific gravity of different materials to aggregates of comparable size. The specific gravity of the aggregates ρ_s [kN/m³] was determined using Equation (2):

$${\rho }_s={\displaystyle \frac{M_s·{\rho }_L}{\left(M_1-M_2\right)}} ,$$

(2)

where M_s is the weight of the granular material used in the test; ρ_L is the specific gravity of the immersion liquid (water) at 23 °C; M₁ is the mass of the pycnometer filled with a known volume V_p of the dispersion consisting of the granular material and the immersion liquid; M₂ is the mass of the pycnometer filled with a known volume V_p of the immersion liquid only.

The Particle Size Distribution (PSD) analysis was conducted with a Mastersizer 3000E laser granulometer (Malvern Panalytical, Malvern, UK) equipped with a Hydro EV wet dispersion unit [43]. The analysis was conducted with the aid of ultrasound (40 W) for a duration of 15 min during the sample preparation stage. The ultrasound facilitates the breakage of the interparticle chemical bonds in the aggregates, thereby releasing the elementary particles. During the analysis, the sample was agitated at a speed of 1500 RPM by the stirrer to prevent sedimentation. The PSD analyses were conducted on all the materials (Figure 1) by using water as dispersant except for gypsum GP, in this case ethanol was used as a dispersant to prevent dissolution or hydration of the sample.

The Aggregate Size Distribution (ASD) analysis was performed by dry sieving, with a stack of sieves with decreasing mesh sizes (10, 4, 2, 1, 0.50, 0.25, 0.125, 0.100, and 0.075 mm). Screening was carried out mechanically using a vibrating screen (Retsch, Haan, Germany) for a minimum duration of three minutes (Figure 2).

Each parameter was calculated as the average of three repeated measurements; the mean values are reported in

Table 1. Summary of the results of the physical characterization of granular materials.

Material	Gypsum GP	Oxides WL	Sulphates BD	Sulphates AD	Galena MS	Galena PN	Galena PR	Zinc Blende PR	Zinc Blende RD
Moisture content [%]	38	7	21	17	6	3	2	4	3
Specific weight [kN/m3]	16	44	27	31	54	52	51	34	34
PSD-d50 [µm]	17	4	6	6	11	14	13	17	13
ASD-d50 [mm]	1.500	0.375	0.750	0.875	0.110	0.125	0.110	0.100	0.105

.

2.2. Experimental Devices

2.2.1. DICAAR Environmental Wind Tunnel

To analyse the emission characteristics of each source type, an experimental study was carried out using an Environmental Wind Tunnel, designed and assembled in the DICAAR laboratories, Cagliari University (Figure 3), based on an in-depth review of relevant case studies discussed in the scientific literature [25,26,27,29,30,44]. The DICAAR EWT is an open-circuit, suction-type tunnel. Its layout is depicted in Figure 3, where the four main sections can be distinguished: Convergence Section (CS), Flow Development Section (FDS), Test Section (TS), and Drive Section (DS). The CS promotes the acceleration and alignment of the airflow and reduces the air turbulence. The FDS allows the full development of the required Atmospheric Boundary Layer Profile (ABL) before the TS. The TS houses the erodible samples to be tested and two air sampling tubes, placed upstream and downstream of the sample niche at 2 cm above the floor, which are connected to the instruments for measuring wind speed (a Pitot tube anemometer) and PM concentrations (two DustTrak DRX model 8533 [45] (TSI Incorporated, Shoreview, MN, USA), which measure the instantaneous concentrations of different PM fractions, i.e., PM₁, PM₂.₅, PM₄, PM₁₀, and total PM). A suction fan is located at the end of the DS. A more detailed description of the tunnel design and measuring devices can be found in Pinna et al. [46].

During the tests, variable flow speeds were simulated using the suction fan installed at the tunnel end.

For each fan operating regime, vertical wind speed profiles were measured at eleven points along the tunnel centreline, located upstream of the erodible samples. These measurements allowed the characteristic wind velocities associated with each fan setting to be defined (

Table 2. Correspondence between fan operating regimes and investigated wind velocities.

Fan Rotation Frequency [RPM]	Wind Velocity at 2 cm Above Floor (u2cm) [m/s]	Wind Friction Velocity (u*) [m/s]	Wind Velocity Reported at 10 m (u10m) [m/s]
1650	3.98	0.34	8.82
1800	4.32	0.37	9.68
1950	4.74	0.40	10.39
2100	4.99	0.44	11.32
2255	5.39	0.47	12.08
2400	5.68	0.50	12.85
2550	6.08	0.53	13.66
2700	6.42	0.57	14.39

).

Owing to the flat configuration of the tunnel floor, the zero-plane displacement d was set to zero. Vertical wind profiles measured were fitted with the log-law wind profile equation (i.e., Law of the Wall). From this fit, we obtained both the friction velocity u* and the aerodynamic roughness length z₀. The fitted z₀ values ranged between 2.2 × 10⁻⁴ m and 3.0 × 10⁻⁴ m. Because all profiles were recorded upstream of the erodible samples, the resulting parameters (i.e., u* and z₀) are representative of the flow acting on the granular surfaces and are not biased by roughness elements generated by the tested material itself. Wind speeds at the 10 m reference height were computed by extrapolating the fitted logarithmic profile to that height.

2.2.2. PI-SWERL

The PI-SWERL is a portable device to test and measure the PM emissions generated by a standard shear stress. The model used for the tests was the PI-SWERL MPS-2p (November 2018 version). The device is composed of an open-bottomed, cylindrical chamber (test chamber) that houses a flat, annular blade (25 cm in diameter) to be positioned parallel to the ground surface at approximately 5 cm above it (Figure 4). The cylindrical chamber has a diameter of 30 cm and a height of 20 cm. A foam seal along the edge of the test chamber conforms to the natural roughness of the surface, allowing it to be perfectly closed. A computer-controlled 24-volt DC motor drives the rotation of the annular blade, thus creating a velocity gradient between the blade and the underlying surface that simulates the shear stress acting on the ground during a wind event [47]. The magnitude of the friction velocity depends on the blade rotational speed (RPM) according to the following equation, developed by Etyemezian et al. [48]:

$$u*=C_1{ \alpha }^4 RP{M}^{{\displaystyle \frac{C_2}{\alpha }}} ,$$

(3)

where u* is the friction velocity (m s⁻¹), C₁ is a constant of 0.000683, C₂ is a constant of 0.832 and α (−) is a parameter that depends on the surface roughness. The values assumed by α are based on four roughness categories described in detail by Etyemezian et al. [48].

The PM concentration is monitored at one-second intervals by a DustTrak II aerosol monitor (TSI Incorporated, Shoreview, MN, USA), model 8530 [49], connected to the test chamber by active sampling at a flow rate of 3 L min⁻¹. The SwerlView application, developed by the Desert Research Institute [33], was used to track the PM₁₀ concentration during each test and allow the synchronization between the data acquisition systems (e.g., blade rotational speed and PM₁₀ concentrations).

2.3. Procedures for the Estimation of the Threshold Friction Velocity

2.3.1. EPA Procedure

The EPA particle size analysis proposed in Document AP-42 CH 13.2.5 (Industrial Wind Erosion) [23] enables the u*_t to be identified through its association with the mode of the aggregate size distribution. The analysis consists of passing the samples through a 1 cm sieve to remove the coarse material. Subsequently, the material is passed on a stack of sieves with decreasing mesh size openings (4, 2, 1, 0.50, and 0.25 mm). The material, firstly placed at the head of the sieve stack, is manually sieved by moving the stack in a limited number of circular movements (20).

During the experimental test, it was observed that for several samples a significant quantity of material exceeded the final sieve size specified by the EPA (0.250 mm). Consequently, the EPA procedure was supplemented by incorporating a 0.125 mm sieve at the end.

After the sieving process, the material retained on each sieve surface was weighed, and the mode of the ASD was determined. For each material, a threshold friction velocity corresponding to the mode value was identified, as indicated by

Table 3. Relationship between mode of the ASD and the threshold friction velocity proposed in Document AP-42 CH 13.2.5 Industrial Wind Erosion [23]. The mode of 0.188 mm was added by the authors.

Mode [mm]	u*t [m/s]
3	1
1.5	0.76
0.75	0.58
0.375	0.43
0.188	<0.43

.

2.3.2. Surface Threshold Friction Velocity Analysis Through DICAAR EWT

The samples to be tested in the wind tunnel were preliminarily sieved using a 1 cm mesh sieve. The material passing through the 1 cm sieve was placed inside aluminium sample trays measuring 0.50 m in length, 0.20 m in width, and 0.02 m in depth (Figure 5). Inside the tray, the surface of the samples was smoothed (without compaction), as to result at the same level as the tunnel floor once inserted into the designated housing.

The emission test was performed three times for each material and friction velocity tested, with each experiment lasting for 120 s. After each experiment, the residual material was mixed with freshly sieved material from the same sample to create a consistent and representative surface for the next test.

Testing began at the minimum fan speed reported in Table 2, which was gradually increased across subsequent trials until the onset of measurable particle emission was observed. The threshold velocity was considered to be reached when, in each of the three tests, the downstream DustTrak detected an average PM₁₀ concentration 20% higher than the average concentration detected upstream of the samples. The upwind average concentration was maintained below 0.03 mg/m³ in order to ensure the consistency of the values between the various tests.

2.3.3. Surface Threshold Friction Velocity Analysis Through PI-SWERL

A circular tray (⌀ 40.5 cm) was filled with 5 litres of material before each test, and an aluminium bar was used to level the surface, as was done for the wind tunnel tests (Figure 6).

A hybrid test was performed and repeated five times for each material, according to the following procedure:

• Start-up phase: 10 s at 0 RPM with the ventilation system active (100 L min⁻¹) to clean the air in the chamber;
• Ramp phase: a gradual increase in RPM from 0 to a target value of 1500 RPM over 60 s;
• Constant phase: RPM constant for 60 s;
• Ramp phase: RPM increased by 250 over 20 s;
• Repetition of phases 3 and 4 until reaching a constant RPM of 2750;
• Final phase: RPM decreased to 0 within 30 s.

The total duration of each test was 560 s.

To strengthen the robustness of the collected experimental data and investigate higher friction velocities on less emissive materials, the dataset for all materials was supplemented with three additional hybrid tests structured as follows:

• Start-up phase: 10 s at 0 RPM with active ventilation (100 L min⁻¹) to clean the air;
• Ramp phase: slow increase in RPM from 0 to a target of 2750 RPM over 60 s;
• Constant phase: RPM maintained at 2750 for 60 s;
• Ramp phase: RPM increased from 2750 to 3000 over 20 s;
• Constant phase: RPM held at 3000 for 60 s;
• Final phase: RPM decreased to 0 within 30 s.

The total duration of these additional tests was 240 s.

A value of α equal to 0.94, corresponding to roughness category B [48], was selected for most materials for the determination of the friction velocities as a function of blade rotational speed. For the sulphates AD and BD, which exhibited visually coarser surface characteristics, a lower value (α = 0.90, Category C) was adopted.

Before conducting the next test, the leftover material from the previous experiment was recycled and mixed with fresh material to regenerate the wind-exposed surface and ensure a homogeneous sample. To prevent cross-contamination between experiments, a cleaning cycle was carried out after each test, consisting of a ramp test on an empty sample tray at 5000 RPM for a duration of 300 s.

The threshold friction velocity for each material was determined by analysing the PM₁₀ concentration signals recorded during the PI-SWERL tests. When the friction velocity reaches the value at which particle entrainment begins, the PM₁₀ concentration exhibits a characteristic sustained increase. Following the procedure proposed by Van Leeuwen [39], a MATLAB (R2024b version) code was developed to automatically detect the threshold by identifying the first sequence of consecutive increases in PM₁₀ concentration. While Van Leeuwen originally required 10 consecutive increases, in our implementation, this number was reduced to 8, which provided a more stable criterion for our datasets while maintaining the same detection logic. The friction velocity corresponding to this sequence was taken as the threshold u*_t.

3. Results

3.1. EPA Procedure

The results of the EPA tests are shown in Figure 7 and highlight a significant presence of macro-aggregates (d > 1 mm) in the gypsum GP (87%) and, to a lesser extent, in the sulphates BD (70%). A high quantity of macro-aggregates was also observed in the oxides WL samples (55% > 1 mm), while sulphates AD, zinc, and lead sulphide samples predominantly consisted of aggregates smaller than 1 mm. In particular, a notable amount of fine material was found in the galena PR and zinc blende PR samples, where approximately 40% of the material fell within the 125–250 µm size range.

The mode of the experimentally determined aggregate size distributions is reported in

Table 4. Mode of aggregate size distribution and threshold friction velocities determined according to EPA procedure.

Material	Gypsum GP	Oxides WL	Sulphates BD	Sulphates AD	Galena MS	Galena PN	Galena PR	Zinc Blende PR	Zinc Blende RD
Mode [mm]	3.00	3.00	1.50	0.375	0.375	0.375	0.188	0.188	0.188
EPA u*t [m/s]	1	1	0.76	0.43	0.43	0.43	<0.43	<0.43	<0.43

. It should be noted that, as the procedure employed in this study incorporated the employment of an additional 0.125 mm sieve (not included in the official EPA procedure), a mode of 0.188 mm was obtained for the PR galena and PR and RD blende zinc samples. Table 4 also reports the u*_t values estimated through the correlations provided by the EPA (see Table 3). The results show that the materials offering the highest wind resistance are those with a predominance of macro-aggregates (gypsum GP, oxides WL, and sulphates BD). The remaining materials exhibit a u*_t of 0.43 m/s (sulphates AD and galena PN and MS) or lower (galena PR and zinc blende PR and RD).

3.2. DICAAR EWT

The results obtained with the EWT are reported in

Table 5. Percentage difference in PM₁₀ concentrations (downwind-upwind) determined varying friction velocity.

	Variation in PM10 Concentrations (Downwind-Upwind) [%]
u* [m/s]	Gypsum GP	Oxides WL	Sulphates BD	Sulphates AD	Galena MS	Galena PN	Galena PR	Zinc Blende PR	Zinc Blende RD
0.34	0	0	0	92	0	0	5	22	0
0.37	0	0	16	-	8	4	5	68	19
0.40	0	0	0	-	0	107	49	-	125
0.44	0	0	34	-	36	-	-	-	-
0.47	0	29	-	-	-	-	-	-	-
0.50	0		-	-	-	-	-	-	-
0.53	39		-	-	-	-	-	-	-
0.56	-		-	-	-	-	-	-	-

as the mean percentage difference between the PM₁₀ concentrations measured downwind and upwind of the samples. According to the adopted criterion, the threshold friction velocity for each material (bold values) corresponds to the fan setting at which all three replicate tests exceeded the 20% difference. The results of the EWT tests indicate that sulphates AD and zinc blende PR are the materials offering the least resistance to wind erosion, as they begin to emit significant quantities of PM at a threshold friction velocity of 0.34 m/s. Higher threshold velocities were determined for the surfaces of zinc blende RD, galena PR and PN (0.40 m/s), and sulphates BD and galena MS (0.44 m/s). The materials that exhibited the highest resistance to wind erosion were the oxides WL and gypsum GP, whose threshold friction velocities were found to be 0.47 m/s and 0.53 m/s, respectively.

Table 6. Threshold speeds determined by EWT experiments.

Material	Gypsum GP	Oxides WL	Sulphates BD	Sulphates AD	Galena MS	Galena PN	Galena PR	Zinc Blende PR	Zinc Blende RD
EWT u*t [m/s]	0.53	0.47	0.44	0.34	0.44	0.40	0.40	0.34	0.40
ut at 2 cm [m/s]	6.08	5.39	4.99	3.98	4.99	4.74	4.74	3.98	4.32
ut at 10 m [m/s]	14.39	12.08	11.32	8.82	11.32	10.39	10.39	8.82	9.68

summarizes the results by listing, for each material, the threshold friction velocity, the threshold wind velocity measured 2 cm above the wind tunnel floor, and the equivalent value extrapolated at 10 m above ground level.

The experimental results showed that the 20% difference threshold (see Section 2.3.2) was reached by the materials under consideration in two distinct ways. Figure 8 present the PM₁₀ concentration curves measured at the threshold friction velocity for both the upwind (blue) and downwind (red) sample locations. It was observed that for certain materials (sulphates AD and galena PN, Figure 8a,b), the achievement of the threshold friction velocity was coincident with the onset of a sustained emission condition, ascribable to the saltation phenomenon. In other cases (sulphates BD and gypsum GP, Figure 8c,d), the increase in the average concentration downstream of the samples was mainly due to instantaneous peaks in PM concentration, rather than to a sustained emission condition, typically associated with direct entrainment processes [30,50].

3.3. PI-SWERL

The results of the PI-SWERL tests (see Section 2.3.3) are shown in

Table 7. Threshold friction velocity of surfaces and standard deviation determined through PI-SWERL.

Material	Gypsum GP	Oxides WL	Sulphates BD	Sulphates AD	Galena MS	Galena PN	Galena PR	Zinc Blende PR	Zinc Blende RD
PI-SWERL u*t [m/s]	/	0.60	0.61	0.34	0.41	0.31	0.31	0.29	0.32
Standard deviation [m/s]	/	0.065	0.078	0.008	0.087	0.011	0.018	0.017	0.025

. Threshold friction velocity values less than 0.30 m/s were found for zinc blende PR (0.29 m/s), with slightly higher velocities observed for sulphates AD, galena PN, galena PR and RD (all between 0.31 and 0.32 m/s). Oxides WL proved to be the materials offering the greatest resistance to wind action (u*_t = 0.60 m/s), slightly higher than that of sulphates BD surfaces (u*_t = 0.54 m/s). An intermediate value was found for galena MS (u*_t = 0.41 m/s). The tests were not performed on Gypsum GP, due to the deterioration of the samples during transportation from Cagliari University to Basel University. The results are summarized in Table 7. Figure 9 presents, as examples, the determination of u*_t from some tests carried out with the PI-SWERL. It is observed that, following a puff of dust induced during each ramp phase, a rapid decrease in PM concentration values occurs during the constant phase. This pulsed response is typically associated with direct entrainment of dust.

4. Discussion

4.1. Comparison of the Materials’ Erosion Resistance

The implementation of the three different procedures led to the results summarized in

Table 8. Threshold friction velocity of surfaces determined according to EPA procedure and through EWT and PI-SWERL experiments.

Material	Gypsum GP	Oxides WL	Sulphates BD	Sulphates AD	Galena MS	Galena PN	Galena PR	Zinc Blende PR	Zinc Blende RD
EPA u*t [m/s]	1	1	0.76	0.43	0.43	0.43	<0.43	<0.43	<0.43
EWT u*t [m/s]	0.53	0.47	0.44	0.34	0.44	0.40	0.40	0.34	0.40
PI-SWERL u*t [m/s]	/	0.60	0.61	0.34	0.41	0.31	0.31	0.29	0.32

. The following observations emerge:

• EPA Procedure: low threshold friction velocity values were determined for zinc blende PR and RD, and galena PR (u*_t < 0.43 m/s), sulphates AD, and galena PN and PN (u*_t = 0.43 m/s). In contrast, higher values were recorded for sulphates BD (u*_t = 0.76 m/s), and especially for gypsum GP and oxides WL (u*_t = 1 m/s).
• Wind Tunnel Tests: the threshold friction velocity was found to be 0.34 m/s for sulphates AD and zinc blende PR. Higher values were measured for zinc blende RD, galena PR and PN (u*_t = 0.40 m/s), sulphates BD and galena MS (u*_t = 0.44 m/s). The materials that showed the highest resistance to wind erosion were oxides WL (u*_t = 0.47 m/s) and gypsum GP (u*_t = 0.53 m/s).
• PI-SWERL Tests: Threshold friction velocity values ranged between 0.29 and 0.34 m/s for the two zinc blendes (PR and RD), galena PR and PN, and sulphates AD. Higher values were found for galena MS (u*_t = 0.41 m/s), sulphates BD (u*_t = 0.61 m/s), and oxides WL (u*_t = 0.60 m/s). Gypsum GP was not tested with this method.

In all tests, it was found that sulphates BD, and even more so oxides WL and gypsum GP, exhibited greater resistance to wind erosion compared to the other materials. Despite their similarity to BD, sulphates AD proved to be the most sensitive to wind action, along with zinc blende PR. The zinc blende RD and the galena PR and PN showed low resistance to wind erosion in the EPA and PI-SWERL tests; however, in the wind tunnel, higher wind speeds, compared to the zinc blende and sulphates AD, were required to initiate the erosive phenomenon. Finally, with regard to galena MS, the EPA procedure indicated a high vulnerability to erosion, contrary to the results obtained through EWT and PI-SWERL testing.

The results were cross-checked with the physical characteristics of the materials in order to identify possible dependency relationships:

• No dependency was found between the Particles Size Distribution and the threshold velocity. Materials with overlapping PSD exhibited very different threshold friction velocities, such as gypsum GP and zinc blende PR. The lack of a clear relationship between PSD and the emissivity of materials under wind erosion is consistent with the findings of Alfaro et al. [51].
• The specific weight also did not show a clear influence on the threshold velocity of the materials. For instance, gypsum GP and sulphates BD (both characterized by very high threshold velocities) have lower specific weights than all the zinc blende and galena samples, which instead displayed lower thresholds across all methods.
• A clear dependency was observed between the threshold velocity and the Aggregates Size Distribution of materials, whereby higher friction threshold velocities were consistently associated with larger median aggregate sizes. The only anomalous behavior was observed for sulphates AD, which exhibited a very low threshold velocity despite a high ASD. This outcome may be attributed either to the high propensity of the aggregates to break apart during saltation or to the greater availability of loose fine material on the surface compared to the other materials. The dependence of PM emissions on ASD has already been reported in previous research [51,52].

The effect of water content on PM₁₀ emissions from wind erosion is widely addressed in the literature for materials with homogeneous characteristics and is primarily attributed to the enhancement of interparticle bonding forces with increasing moisture [53,54]. In the present study, the determination of threshold friction velocity for each material was conducted exclusively at the moisture content values reported in Table 1. Since the bonding forces induced by water is affected by the chemical composition of each material [55], and each material was tested at a single moisture content, the effect of this parameter on u*_t was not investigated.

4.2. Comparison of Procedures

By comparing the results of the three procedures, it can be observed that the u*_t values determined with the U.S. EPA methodology are generally higher than those derived from laboratory tests performed with the EWT and PI-SWERL. This discrepancy can be explained by considering that the u*_t values provided by the EPA procedure refer to a saltation phenomenon [23]. Indeed, the EPA procedure is derived from experimental studies that correlate the mode of the material’s ASD with the threshold velocity that activates the saltation phenomenon, which is identified through visual observation [8]. In contrast, the procedures implemented with the EWT and PI-SWERL identify the threshold friction velocity also for direct lifting of particles (direct entrainment). In particular, the analysis of PI-SWERL data indicates that all threshold velocities were achieved through direct entrainment of dust particles. By contrast, in the EWT tests, some materials attained the threshold via direct entrainment—specifically those characterized by u*_t values markedly different from the EPA estimate—whereas others displayed behaviour more consistent with saltation processes.

Previous experimental studies have reported that the threshold friction velocity associated with direct entrainment may be lower than the threshold values predicted by theoretical models based on the visual identification of the saltation onset. However, the emissions resulting from direct entrainment are generally limited compared to those produced by saltation [10,50,56].

On the other hand, the threshold friction velocities derived from EWT and PI-SWERL tests are generally comparable for most of the materials under examination. The maximum observed difference across all samples was 0.17 m/s, recorded for the sulphates BD. Overall, these differences are considered acceptable and consistent with findings reported in a previous study [38], where discrepancies exceeding 33% were identified (the maximum percentage difference observed in this study is 38.6%). Specifically, the analysis carried out with the PI-SWERL tended to overestimate u*_t values compared to the EWT for the materials that showed higher resistance to wind erosion (i.e., oxides WL and sulphates BD). In contrast, for some more erodible materials (such as galena PR and PN), the trend was reversed, with u*_t values lower than those obtained in the wind tunnel.

These differences can be attributed to the different methodologies used for determining the threshold friction velocity, due to the specific technical characteristics of the test facilities.

The PI-SWERL is factory-equipped with effective synchronization between its data acquisition system, such as shear stress derived from the blade rotational speed and PM₁₀ concentrations. This facilitates the implementation of the methodology adopted in this study (i.e., the occurrence of a predefined number of consecutive increases in the measured concentration). However, the subjective number of eight consecutive increases chosen for this study is not universally defined and must be adapted to the emission behaviour of the materials under investigation. An insufficient number of increments may result in an underestimation of u*_t, indicating premature emission. On the other hand, in the case of low-emissivity materials, even when emission has indeed begun, the number of particles emitted in the atmosphere may be so limited that the detected concentration increases are small and often interrupted by sporadic instantaneous decreases in concentration. Figure 10 illustrates the results of a test in which the threshold velocity was identified at the point where eight consecutive increments in concentration were reached. The temporal trend of concentrations is represented by the green line, while the attainment of the threshold condition is indicated by the red dashed arrow. This example highlights the limitation of using the number of exceedances as a criterion; in this case, selecting a number of increments greater than eight would not have led to a correct identification of u*_t, even though the emission process appeared to have already started.

In addition, the interpretation of the optimal parameter α used in the PI-SWERL introduces an inherent source of uncertainty. The relationship between α and the surface roughness is not straightforward [48], and each roughness category corresponds to a wide interval of possible textures. For example, category B spans sandpapers with grit sizes ranging from 36 to 60. Nevertheless, this necessary simplification introduces uncertainty into the derived friction velocity values, as variations in true surface roughness within the same category influence the local shear stress field and consequently the onset of particle entrainment.

In contrast, synchronizing wind speed variations with instantaneous concentration readings proved to be operationally challenging for EWT tests. This difficulty is due to the absence of automatic synchronization systems between the wind tunnel fan and the PM₁₀ concentration measurement instrument, as well as the necessity for coordination between two operators situated at different locations, which may easily lead to errors. Therefore, in EWT the u*_t value was identified based on the percentage difference between PM₁₀ concentrations measured upwind and downwind of the samples. However, this methodology is feasible only when at least two instruments for PM₁₀ measuring are available (one upwind and one downwind), and when the background concentration level within the wind tunnel remains constant. Even when these conditions are satisfied, the reliability of the resulting u*_t value is subject to a certain degree of uncertainty, which depends on the relationship between the set percentage difference, the duration of the measurements, and the availability of erodible material within the sample trays. For example, in this study the threshold velocity was determined over a 120 s measurement period when the mean PM₁₀ concentration measured downstream of the erodible samples exceeded the upstream concentration by more than 20% in all three replicate tests. Examination of the data in Table 6 indicates that a target percentage difference of 15% would generate deviations below the experimentally determined value for sulphates BD and zinc blende RD, while a target of 25% would generate deviations above the experimental value for zinc blende PR. It should also be noted that the average percentage differences reported in Table 5 refer to a 120 s time interval. For materials that reach the threshold with emissions characterized by instantaneous peaks, as illustrated in Figure 8c,d, the choice of the time interval over which the average is calculated becomes particularly important. If an instantaneous peak occurs in either the first or second half of the test, selecting a 60 s interval would produce a percentage difference higher or lower than that observed in the present study.

In conclusion, both methodologies are subject to a certain degree of uncertainty. Nevertheless, detecting a specific number of consecutive increases in concentration provides a more objective approach for identifying the threshold friction velocity. However, the possibility of implementing this methodology in EWT depends on its configuration, which can vary significantly depending on the needs and resources of different research institutions, influencing the design, construction, and potential for automatic synchronization.

5. Conclusions

This study provides a comparative analysis of three methodologies (the EPA sieving-based estimation procedure, the DICAAR wind tunnel tests, and the PI-SWERL tests) used for determining the threshold friction velocity of nine granular materials collected from a metallurgical industrial plant located in Sardinia, Italy. For each methodology, the article discusses the ability to characterize the wind erosion potential of different surfaces of granular materials.

The EPA procedure, based on the empirical correlation between the aggregate size distribution and u*_t, produced values that were consistently higher than those obtained with the EWT and PI-SWERL tests. This discrepancy can be primarily explained by the fact that the EPA method refers to a sustained saltation phenomenon, while the experimental approaches detect the actual onset of PM emissions, including cases where saltation is incipient or direct entrainment occurs.

This work also highlights a critical limitation in the EPA procedure, as the exclusion of the 0.125 mm sieve from the standard stack makes it impossible to accurately represent the finer aggregate distributions of some materials. A notable example is the application of the standard EPA method to zinc blende PR, which would have led to a modal diameter of 1.5 mm, corresponding to an estimated u*_t of 0.76 m/s, nearly twice the value measured by adding the 0.125 mm sieve. This overestimation emphasises the risk of inaccurate characterisation of the erosion potential of materials with a significant proportion of fine aggregates. It is therefore recommended to integrate the sieve stack to include finer meshes (e.g., 0.125 mm), particularly when considering materials composed predominantly of fine aggregates. The adjustment would enhance the representativeness of the mode diameter and prevent the overestimation of u*_t.

The PI-SWERL tests provided u*_t values consistent with the EWT results. Indeed, the correlation between the two experimental methods proved to be satisfactory, with discrepancies rarely exceeding 0.10 m/s. However, the PI-SWERL system offers some advantages from an operational perspective, which include the instrument portability, the integrated design, and the ease and automatic detection of the emission (e.g., using criteria such as a specific number of consecutive PM₁₀ concentration increases). On the other hand, the wind tunnel, while more complex in terms of setup and instrumentation, allows for a more comprehensive analysis of emission behaviour, including vertical wind profiles and PM₁₀ fluxes, making it an indispensable tool for in-depth studies.

It is important to note that the results achievable with the two instruments depend on the choice of the criteria used to identify the threshold velocities (i.e., the mean percentage difference between downstream and upstream concentrations in the EWT, and the number of consecutive concentration increments in the PI-SWERL). However, an important underlying limitation remains: there is no universally accepted quantitative criterion for defining the threshold friction velocity, and this represents a fundamental source of uncertainty for all experimental studies. Any methodology employed to identify the threshold friction velocity inherently carries uncertainties, due both to the physical complexity of particle entrainment and to the operational assumptions required by each technique. To mitigate these uncertainties, the selected criteria can be refined through preliminary emission tests, which offer an initial understanding of the emission behaviour of the specific material under investigation.

Overall, the EPA procedure can serve as a rapid preliminary screening tool, particularly suitable for identifying the onset velocities of sustained saltation events, which form the basis of most theoretical and mathematical models of wind erosion emissions. In contrast, experimental techniques provide more reliable, material-specific estimates of threshold friction velocity. Both the wind tunnel and the PI-SWERL allow procedures and threshold criteria to be adapted to the specific properties of the materials under study, enabling identification of threshold velocities that account for emissions arising from both saltation and direct entrainment. While the PI-SWERL is especially well-suited for rapid assessments and field-deployable measurements, the wind tunnel remains essential for a detailed analysis of PM emission dynamics due to its ability to more accurately simulate the erosive phenomenon.