Desertification and dust storms are crucial environmental concerns throughout the world, particularly in arid and semiarid regions, and thus finding an effective solution to control these phenomena is essential (Wu et al., 2015; Miri and Davidson-Arnott, 2021). In China, 50% of desert lands are located in the northern part of the country, which acts as the primary producer of dust emissions for the western regions (Zhao et al., 1997; Chen and Tang, 2005). Wind erosion is the main cause of land desertification in Northwest China. In this region, vegetation has been identified as an important measure for combating desertification and reducing aeolian erosion. Shrubs are the best material for stabilizing sediment and controlling aeolian erosion due to their effectiveness, persistency, and low cost (Yang et al., 2006).

Many studies conducted in wind tunnels and in the field have indicated the potential of vegetation for reducing wind erosion by reducing the wind speed and improving soil retention (Wolfe and Nickling, 1993; Leenders et al., 2011; Ma et al., 2019; Miri et al., 2019, Torshizi et al., 2020a, 2020b; Miri et al., 2021). Vegetation is planted in the form of wind strips, shelterbelts, and windbreaks which protect the soil surface against erosive winds (Dong et al., 2001; Torshizi et al., 2020b). Windbreaks are either natural or artificial barriers that are used in farmlands for reducing wind speed and in aeolian desertified lands for erosion control (Brandle et al., 2004; Torshizi et al., 2020a, 2020b). They are arranged in one or more rows in a specific pattern (Cornelis and Gabriels, 2005). The porosity, density, and height of vegetation influence the characteristics of the airflow field around the vegetation (Cornelis and Gabriels, 2005). Porosity determines the rate of airflow passing through a windbreak, so it influences the wind speed reduction rate and protected area in the lee of the windbreak (Gillies et al., 2000; Miri et al., 2017). A taller windbreak with more rows of plants has a greater ability to reduce the wind speed (Cui et al., 2012; Wu et al., 2013). Orientation, spacing of individual plants, number of rows, and the arrangement of plants in the rows determine the efficiency of a windbreak in affecting the downwind airflow and the extent of the protected zone provided by the windbreak (Gillies et al., 2002; Brandle et al., 2004; Cornelis and Gabriels, 2005; Cheng et al., 2020; Torshizi et al., 2020a). Cheng et al. (2020) examined the effects of plant spacing, the number of rows, and arrangement on the airflow field of a simulated forest belt in a wind tunnel and found that the reduction of wind speed decreased with plant spacing and increased with the number of rows. A staggered pattern presented lower downwind speeds and a greater characteristic length of airflow recovery than the rectangular arrangements. Liu et al. (2018) found greater potential of a windbreak with the double row-staggered arrangement than those with either the double-row-rectangle or single-row arrangements. Fu et al. (2021) examined different densities (low, medium, and high) and patterns (two rows per one belt pattern, one row per one belt pattern, uniform distribution pattern, and random distribution pattern) of Haloxylon ammodendron in affecting wind erosion, and found that aeolian sediment transport is influenced by either the density or the pattern of vegetation and their interaction.

Although many studies have examined wind speed changes and the sheltering effects of different plant types in the form of windbreaks (e.g. Li et al., 2018; Torshizi et al., 2020a, 2020b; Miri et al., 2021) and individual plants (Leenders et al., 2007; Miri and Davidson-Arnott, 2021; Bhutto et al., 2022), more in-depth investigations on the effectiveness of various plant species in the form of windbreaks with different patterns and designs are still needed. Understanding the airflow field around plants in different densities, designs, and arrangements is essential for assessing the dynamic mechanisms of wind erosion in the presence of vegetation. Different plant species with different morphologies, densities, and designs show different abilities in affecting airflow and reducing wind speed (Miri et al., 2017, 2018). Thus, to improve our knowledge of the effects of plants on airflow, the interactions of airflow and various plant species should be evaluated for different wind speeds, designs, densities, and scales.

Caragana korshinskii is widely used as a windbreak in northern China for various purposes, such as protecting the soil and water resources, improving soil fertility (Niu et al., 2003; Su and Zhao, 2003; An and Huang, 2006) and most importantly controlling sand movement and dust storms (Zhang et al., 2020). Caragana korshinskii has the characteristics of many branches, high density, a large crown, a well-developed root system (which reaches a depth of >9 m and 20 m horizontally) and adaptability to harsh conditions. Many studies have demonstrated the windbreak benefits of C. korshinskii, and found that the spatial structure distribution of the C. korshinskii forest has an important influence on its protective benefits (Gao et al., 2010). Another experiment showed the importance of establishing a reasonable density in the C. korshinskii protection forest (Yan et al., 2018).

This study contributes to a better understanding of the sheltering effect of C. korshinskii with respect to different densities of distribution patterns, different row numbers, and belt spacing of the forest belts. The objectives of this study were: 1) To assess the effects of different vegetation patterns on the characteristics of the downwind airflow, the minimum downwind wind speed, and its corresponding position; 2) To compare the windbreak efficiencies of different vegetation densities and forest belt patterns of C. korshinskii; and 3) To determine the effects of the number of rows and belt spacing of C. korshinskii on wind speed. We designed a series of wind tunnel experiments with artificial plants and compared the shelter efficiencies of different vegetation designs and airflow fields downwind of the vegetation.

The simulation experiments were conducted in the wind tunnel of the Jiufeng Sand Physics Laboratory, School of Soil and Water Conservation, Beijing Forestry University (Fig. 1a). It is a non-circulating blow-type wind tunnel, in which the wind speed along the central axis of the working section can be continuously controlled from 3 to 40 m s^-1. The tunnel is divided into seven sections: the power section (1.5 m), fan section (2.9 m), transition section (0.91 m), stabilization section (1.5 m), construction section (1.5 m), test section (12 m) and diffusion section (3.7 m). The total length of the wind tunnel is 24 m and the experimental section has dimensions of 0.6 m wide, 0.6 m high, and 12 m long. The location of the anemometer probe is adjustable with a precision of 1 mm in the three-dimensional displacement measurement system. To reduce large-scale eddies and fit the airflow profile, a honeycomb network was placed at the shadow position of the stabilization section. Roughness elements and spire (3 cm tall, 4 cm wide, and 5 cm long) were used to develop a deep and fully turbulent momentum boundary-layer flow. The thickness of the boundary layer was 0.2 m, which agrees with the boundary layer recommended by White (1996).

Caragana korshinskii is a perennial medium shrub with a height of 0.5-2.0 m (Jian et al., 2014). The artificial plant models were downscaled by a ratio of 1:18 based on the geometric shape and canopy porosity of the actual plants in the field. The models restored the caragana to the greatest extent using 3D printing technology and plastic materials were used which can not only better shape the plant morphology, but also fit the flexibility characteristics of the original plant to a certain extent. The simulated plants were 3 cm high representing the 60 cm height of 2-year old C. korshinskii in the field (Wang et al., 2004), and the crown was 2.5 cm×2.5 cm. The ratio of obstacle height to boundary layer was 0.15, which agrees with the requirements of downscaling as suggested by White (1996) (0.15<0.2). The flow Reynolds number ($Re=U\delta ~/\nu $, where $Re$ is the flow Reynolds number, U is the free-stream wind velocity, δ is the thickness of the boundary layer at 0.2 m, and $\nu $ is the kinematic viscosity of air, 1.5×10^-5 m² s^-1) ranged from 8×10⁴ to 18×10⁴; and the model Reynolds number ($R{{e}_{m}}=$$Uh~/\nu $, where $R{{e}_{m}}$ is the model Reynolds number, and h is the height of models) ranged from 1.2×10⁴ to 2.8×10⁴, which are both sufficiently high (Snyder, 1972; White, 1996); the meanings of U and v are the same as above.

Plastic plant models with similar morphology to C. korshinskii were used for the experiments (Table 1). Before the experiments, wind profiles were measured along the wind tunnel without plants to select the appropriate locations for simulated vegetation where the boundary layer thickness was sufficiently well developed. The planted surfaces had a length of 2 m and were positioned at 5 m downwind from the leading edge of the test section of the tunnel (Fig. 1b), which is 25 times the boundary layer thickness. This value meets the general rules of matching mean velocity profiles (10-25 boundary layer height) (White, 1996).

The density of plants was quantified using the horizontal vegetation cover (${{C}_{v}}$) and was calculated as follows:

where N is the number of plants (N had six different values in this study corresponding to the six different densities of vegetation cover), ${{A}_{PV}}$ is the average plan view area of vegetation elements and ${{A}_{T}}$ is canopy area in the wind tunnel (12000 cm²).

In the experiments for the effect of different vegetation densities and arrangements on wind speed, the artificial plants were arranged in random, uniform, and cluster distribution patterns, each designed with six densities of 20%, 30%, 40%, 50%, 60%, and 70% (Fig. 2). In these experiments, the mean wind speeds were measured every 1 H (where H is the height of the artificial plants of C. korshinskii) from the leading edge of the vegetation to a distance of 20 H.

In the experiments for the effect of the different number of rows, three shelterbelts with the same row spacing (1 H, where h is the height of the plants) but the different numbers of rows (single-row, two-row, and three-row belts) were designed (Fig. 3a). In these experiments, the mean wind speeds were measured from the leading edge of the belt to a distance of 20 H downwind of it.

In the experiments for the effect of inter-belt spacing, four multiple-belt shelterbelts were designed. Each shelterbelt included two double-row belts with the same row spacing (1 H) but with different belt spacing of 4 H, 6 H, 8 H, 10 H, and 12 H (Fig. 3b). In these experiments, the mean wind speeds were measured from the leading edge of the first double-row belt to a distance of 20 H downwind of the second double-row belt. At all positions, the wind speeds were monitored at the middle height of the plants (1.5 cm above the surface), within the logarithmic portion of the wind tunnel (White, 1996).

For uniform, random and cluster canopies, wind speeds of 6, 10 and 14 m s^-1 were used; and for the shelterbelts, a wind speed of 10 m s^-1 was used. Wind speeds were monitored for 30 s using a KIMO hotline anemometer (Zhang et al., 2018). A reasonable evaluation of the shelter effect was essential in order to assess the efficiencies of the examined vegetation arrangements and forest belts and find the optimal design for the windbreaks. The sheltering effect is mainly determined by the shelter distance and velocity reduction (Ma et al., 2019; Torshizi et al., 2020a). The proportional reduction of wind speed (U/U₀, where U and U₀ are the leeward and windward wind speeds, respectively) was used to define the sheltering distance of the vegetation. The wind speed reduction ratio was used to assess the windbreak efficiency of the vegetation (E_v) and was calculated as follows (Cornelis and Gabriels, 2005):

where ${{U}_{a}}$ is the wind speed in the absence of vegetation and U_plant is the wind speed with the presence of vegetation.

With increasing vegetation densities, the wind speed reduction ratio increased due to the increasing number of roughness elements that reduce the force acting on the surface (Crawley and Nickling, 2003). The similarity in the efficiencies of vegetation at densities of 50%, 60%, and 70% indicates that C. korshinskii would not be very effective at reducing wind speed at a density >50%. With very dense obstacles, an area of recirculating eddies in the immediate lee is created which results in the enhancement of turbulence (Cleugh, 1998). Thus, the density of 50% can be considered as the optimal density of C. korshinskii where it shows the greatest sheltering effect. Consistent with the observations of Cheng et al. (2019), Ma et al. (2019) and Guo et al. (2021), the efficiency of shelterbelts decreased with decreasing numbers of rows and increasing belt spacing.

Inter-row/belt spacing is a key parameter that influences the sheltering effect and optimal design of windbreaks. Consistent with the findings of Cornelis and Gabriels (2005) and Pan et al. (2020), the multiple-row belts were more effective than the single-row belt. The greatest sheltering effect of multiple-belt shelterbelts with a spacing of 4 H (72%) revealed the optimal spacing for inter-belt shelterbelts of C. korshinskii. Fang et al. (2018) reported that an optimal inter-row spacing should be within the range of 5-7 H. The recovery rates of wind speed (U/U₀, where U and U₀ are the leeward and windward wind speeds, respectively) between the two double-belt shelterbelts with spacing of 4 H, 6 H, 8 H, 10 H, and 12 H were <0.36, <0.40, <0.41, <0.46 and <0.48, respectively. The recovery rate between the double-belt shelterbelts was much lower than that of a multiple-row windbreak with a row spacing of 6 H (0.80) (Liu et al., 2018). This difference is attributable to the double-belt structure of the C. korshinskii shelterbelts.

The sheltered area of a windbreak is an important index for evaluating windbreak efficiency. This area extends from the leading edge of the windbreak to a downwind distance at which U/U₀ is less than 0.7-0.9 (U/U₀, where U and U₀ are the leeward and windward wind speeds, respectively) (Seginer, 1975; Cleugh, 1998; Torita and Satou, 2007). By applying a criterion of U/U₀< 0.7, the results showed that the sheltered area was 14 H in low and high densities, 20 H in the medium density of the vegetation canopies and 20 H in multiple-row belt and multiple-belt shelterbelts. This sheltered area is comparable with those reported in previous studies, such as 0-2 H (Gao, 2010), 5 H (Bhutto et al., 2022), 7 H (Leenders et al., 2007), 10 H (He et al., 2017), 12 H (Mayaud et al., 2016), 114 H (Wu et al., 2015). Ma et al. (2019) observed a protection distance well beyond 225 H for the inter-row and inter-plant architecture of mixed windbreaks of Haloxylon ammodendron and Caragana korshinskii.

The minimum wind speed and its location (X_min) are other criteria for assessing the shelter effect of a windbreak (Ma et al., 2019; Torshizi et al., 2020a). In this study, the lowest minimum wind speed was observed for a uniform arrangement and it decreased with an increasing density of vegetation. The minimum wind speed was lowest for the doubled shelterbelt with a spacing of 4 H, and it increased with increasing belt spacing. X_min decreased with increases in the number of rows, and similar observations have been reported by Cheng et al. (2020). The locations of the minimum wind speeds for the different vegetation patterns are in reasonable agreement with the range of X_min, which is recommended as 1-5 H (Wang et al., 2001; Rosenfeld et al., 2010). The location of X_min was similar for the two-row and three-row belts (10 H) and tended to be closer to the belt for the single-row belts (6 H). This result demonstrates that by increasing the number of rows, the required distance for recovering wind speed increases (Cheng et al., 2020). The locations of X_min were similar for the multiple-belt shelterbelts with spacings of 4 H to 10 H (6 H) and tended to be closer to the belt for the belt spacing of 12 H (7 H). Cheng et al. (2020) found that X_min tended to be closer to the belt with increasing plant spacing. In the current study, the location of X_min did not show fluctuations with increases in the number of rows or belt spacing. Decreasing wind speed and reaching a minimum at a distance downwind of the multiple-belt shelterbelts has been observed downwind of the inter-plant and inter-row windbreak models of H. ammodendron and C. korshinskii (Ma et al., 2019). The location of the minimum wind speed marks the wind speed recovery zone (Li et al., 2018) and is important in sediment movement (Ma et al., 2019). The multiple-row belts and the multiple-belt shelterbelts reduced the wind speed over a long distance, in which sediment blasting did not occur. Uniform, random, and cluster patterns and the single-row belt reduced the wind speed to below 5 m s^-1 at a lower distance of 12 H. However, sediment movement could occur at a leeward distance greater than 12 H. The multiple-row belts and multiple-belt shelterbelts reduced the wind speed to below the threshold at a further distance (20 H) than those of the other patterns. Downwind of the belt, the wind speed could not recover to the threshold, an observation that was also reported by Guo et al. (2021).

Although the numbers of plants used in the single-row belt, and multiple-row belts are less than those of the low, medium, and high densities of vegetation patterns, the forest belts showed the larger sheltering distance with fewer plants, which is similar to the observations of Cui et al. (2012), Yang et al. (2016) and Cheng et al. (2020). The greater shelter effect of vegetation patterns at a distance lower than 5 H and the effectiveness of multiple-row belts and multiple-belt shelterbelts at a distance greater than 5 H indicate that these vegetation patterns can rapidly reduce the wind velocity and efficiently prevent blowing sediment, and that multiple-row belts and multiple-belt shelterbelts are optimal for providing a greater shelter distance. In arid and semi areas, soil moisture is scarce due to limited precipitation and a low groundwater table, so less vegetation density with high potential is required to reduce the wind speed and control aeolian erosion. Thus, the results of the current study suggest that designing multiple-belt shelterbelts is more efficient than the other vegetation patterns for protecting a large area. This conclusion is supported by Gao et al. (2010), who indicated that multiple-row belts cannot always decrease wind speed enough and multiple-belt windbreaks are often more effective than multiple-row windbreaks.

There are some limitations to this study. First, wind speed measurements upwind of the vegetation and at different heights below and above the canopies and shelterbelts were not taken, although they would have enabled the identification of airflow patterns at different elevations and upwind of the vegetation. Wind speed measurements in wind speeds of 6 and 14 m s^-1 for the shelterbelts were also not taken, but they would have enabled the characterization of the shelterbelt’s efficiency in various wind speeds. Second, real vegetation in the field is actually flexible but it was represented by solid objects in the current wind tunnel experiments. Thus, the effect of kinetic energy loss due to oscillations of the plant’s body on windbreak efficiency was not considered. However, the present study focused on the different vegetation distribution patterns, and the influence of the real plant flexibility on the sheltering effect will be investigated in future studies.

In this study, the effects of several different distribution patterns of vegetation—including uniform, random, and cluster canopy arrangements, forest belts with the different numbers of rows and double-belt shelterbelts with various spacing—on the characteristics of the downwind airflow, the minimum downwind wind speed, and its corresponding position were studied. The uniform pattern was more effective than random or cluster arrangements at all examined densities and wind speeds. All arrangements were more effective with higher densities than lower densities, but the density of 50% was optimal for the vegetation canopies. For forest belts with different numbers of rows but the same plant spacing of 1 H, the windbreak efficiency and sheltered area decreased and the downwind minimum wind speed increased as the number of rows decreased from multiple rows to a single row. For multiple-belt shelterbelts with different spacings but the same row spacing of 1 H, the windbreak efficiency decreased and the downwind minimum wind speed increased with increases in the spacing. The sheltered area and the location of the downwind minimum wind speed were similar for all patterns. Among all vegetation arrangements and forest belts, the single-row belt pattern was less effective and double-belt shelterbelts with the spacing of 4 H were the most effective pattern. The vegetation patterns tested (uniform, random, and cluster) rapidly reduced wind speed and efficiently prevented blowing sediment, and multiple-row belts and multiple-belt shelterbelts were optimal for shelter distance. Although the numbers of plants in the vegetation patterns with high densities (50%-70%) are greater than those in multiple-belt shelterbelts, the sheltering efficiency of multiple-belt shelterbelts is similar to the vegetation patterns at a distance close to the belt but greater at a further distance. Thus, multiple-belt shelterbelts can be recommended for windbreaks designed to reduce wind speed and control aeolian erosion in arid and semi areas where soil moisture is scarce and where windbreaks with fewer plants but high effectiveness are needed.