The Loess Plateau (33°41′–41°16′ N, 100°52′–114°33′ E) is located in northwest China, approximately 640,000 km². The terrain of the Loess Plateau is dominated by hill and gully landscapes, with slope gradients mostly ranging from 24.9% to 53.2% (Lin et al., 2019). The Loess Plateau has a warm temperate continental monsoon climate, with precipitation and temperature decreasing from southeast to northwest. The annual average temperature ranges from 6°C to 14°C, and the annual average precipitation is 200–700 mm, with most rainfall occurring from July to September (Yao et al., 2023a). The soil is characterized by fine particles, soft texture, and abundant mineral nutrients, making it conducive to farming. However, its porous, loose, and poor structure, developed vertical joints, and high water permeability, combined with frequent summer rainstorms, have led to severe soil erosion. The test soil is collected from a local abandoned farmland with clay, silt, sand, and soil organic matter contents of 28.2%, 41.6%, 30.2%, and 19.17 g kg^–1, respectively.

A side-spray rainfall simulator was used in the current experiment, developed by the Institute of Soil and Water Conservation, Chinese Academy of Sciences ( Figure 1a ), which has been widely used for rainfall simulation. The rainfall system consisted of two side-spray nozzles (5.0–12.0 mm), two brackets (7.5 m), two pressure gauges (0.10–0.15 MPa), a water pump (2.2 kW), a water tank (length of 4.0 m, width of 2.5 m, and depth of 2.5 m), and a water supply pipe (inner diameter of 48 mm). The nozzles were installed at the top of the brackets with a spray height of 1 m. The rainfall height of the rainfall simulator is 7.5 m, ensuring that raindrops reached near-terminal velocity. The effective rainfall area is 7 m × 5 m, with a rainfall uniformity of over 85%. The rainfall intensity with a range of 30–140 mm h^–1 and raindrop diameter of 0.17–3.61 mm can be adjusted by regulating the water pressure control valve and replacing the nozzle gaskets. Before rainfall simulation, the rainfall intensity was calibrated to meet the experiment’s requirements.

Rainfall simulation experiments were conducted at the Hydraulic Engineering Laboratory of Northwest A&F University, Shaanxi, China. Soil boxes (length of 2.0 m, width of 1.0 m, and depth of 0.5 m) were used for alfalfa planting and rainfall simulation. The test soil was collected from the top 0–20 cm of the arable layer of a local farmland in Yangling. The collected soil was air-dried and screened by using a 10-mm screen to eliminate the effect of root and other impurities. A 4-cm-thick layer of fine sand was placed at the bottom of the soil box to simulate natural drainage conditions. A porous jute sheet was used to separate the sand and soil layers. The soil was filled into the boxes in layers to achieve a bulk density of 1.25 g cm^-3, based on field measurements in the local farmland. To ensure continuity between layers, each layer was gently scraped before adding the next. After filling, the soil was left under natural conditions for 30 days to simulate a natural soil state.

The experimental grass species was annual alfalfa, a typical forage on the Loess Plateau, which was sown at the end of April 2022. Soil boxes not planted with alfalfa were designed as bare soil control (CK). To explore the impact of vegetation spatial distribution patterns on hillslope erosion, four different vegetation pattern types with 50% vegetation cover were established: planting at the downstream hillslopes (LS), at the middle hillslopes (MS), and at the upper hillslopes (US) and equally spaced planting (SS) ( Figure 1b ). To ensure the hillslope vegetation coverage and different spatial alfalfa distribution patterns, the prepared soil boxes were divided into several soil plots (length × width, 0.2 m × 0.2 m). The grass seeds were evenly sown at specific positions on the hillslopes, and then the grass seeds were evenly mixed with the topsoil. Rainfall simulation experiments were conducted at the alfalfa podding stage. The planting area in all four vegetation spatial distribution patterns was kept the same, accounting for 50% of the entire hillslope area, ensuring consistent alfalfa coverage across all treatments. According to landform types of the Loess Plateau, the experimental runoff plots were established with four slope gradients of 8.75%, 17.63%, 26.80%, and 36.40% (Lin et al., 2019). Based on the long-term rainfall characteristics of the study area, a rainfall intensity of 90 mm h^–1 was applied, with a simulation duration of 60 min.

After adjusting the rainfall intensity and slope gradient, the rainfall simulation experiment was initiated, and the soil erosion process was observed and recorded. The initial time to runoff was recorded, and the sediment samples were collected every 3 min using plastic buckets. The collected sediment samples were weighed, left to settle for 24 h, the supernatant was dropped, and then the sediment was dried at 105°C for 24 h before recording the dry weight. The runoff volume was calculated as the total mass of sediment and runoff minus the dry mass of sediment. Assuming constant rainfall intensity throughout the experiment, infiltration was estimated as the difference between total precipitation and runoff. In this study, the variables used to characterize the soil erosion process are initial time to runoff (TR), runoff rate (RR), sediment yield rate (SR), and infiltration rate (RR). These variables were calculated as follows (Equations 1–3):

where SR, Se, s, t, RR, Rn, ST, IR, IT, RT, and Ri are the sediment yield rate (g m^–2 min^–1), dry mass of collected sediment (g), runoff plot area (m²), sampling time (min), runoff rate (L m^–2 min^–1), runoff volume (L), mass of the collected sediment samples (g), infiltration rate (mm m^–2 min^–1), total infiltration amount (mm), total precipitation (mm), and runoff depth (mm), respectively. The runoff and sediment reduction benefits for alfalfa-covered hillslopes were calculated as follows (Equations 4–5):

where RRB, R_ck , R_i , SRB, S_ck , and S_i are the runoff reduction benefits, runoff volume for CK, runoff volume for alfalfa-covered hillslopes with different spatial distribution patterns, sediment reduction benefits, sediment yield for CK, and sediment yield for alfalfa-covered hillslopes with different spatial distribution patterns, respectively.

The variation in initial runoff generation time for CK and alfalfa-covered hillslopes is shown in Figure 2 . On the slope gradient of 8.75%–36.40%, the initial runoff generation time for CK and alfalfa-covered hillslopes with US, MS, LS, and SS pattern was 2.50 to 8.08, 2.90 to 8.80, 3.25 to 9.03, 6.10 to 9.65, and 7.80 to 11.50 min, respectively. The initial runoff generation time for the US, MS, LS, and SS alfalfa-covered hillslopes was delayed in comparison to CK. The initial runoff generation time with slope gradient of 17.63% to 36.40% for CK US, MS, LS, and SS patterns was advanced at 1.00 to 5.58, 1.30 to 5.90, 1.50 to 5.78, 1.50 to 3.55, and 1.35 to 3.70 min compared with those for the slope gradient of 8.75%, respectively, indicating that the initial runoff generation time was advanced with increasing slope gradient.

Figure 6 illustrates the total runoff volume, total sediment yield, total infiltration, and sediment concentration for CK and alfalfa-covered hillslopes. The total runoff volume, total sediment yield, and sediment concentration increased with slope gradient and followed the order CK, US, MS, LS, and SS. However, the total infiltration decreased with slope gradient and followed the order SS, LS, MS, US, and CK. Meanwhile, compared with CK, total runoff, sediment yield, and sediment concentration on alfalfa-covered hillslopes decreased by 5.33 to 49.89 L m^–², 115.09 to 657.49 g m^–², and 0.26 to 2.22 g L^–1, respectively, and total infiltration increased by 6.27 to 50.71 mm^–2. These findings clearly indicated that alfalfa vegetation significantly enhanced soil infiltration while reducing surface runoff, sediment yield, and sediment concentration. Table 1 presents the runoff reduction benefit (RRB) and sediment reduction benefit (SRB) of alfalfa-covered hillslopes. Both RRB and SRB decreased with increasing slope gradient and varied with the spatial distribution pattern of vegetation. For alfalfa-covered hillslopes, the RRB with slope gradient of 8.75% was 1.92–3.19 times than that for slope gradient of 36.40%, while the SRB with slope gradient of 8.75% was 1.68–2.34 times than that at 36.40%. Among the different distribution patterns, the SS configuration consistently showed the highest RRB and SRB, whereas the US pattern was relatively poor.

In the current study, alfalfa-covered hillslopes exhibited good benefits in reducing soil erosion compared with CK. This indicated that plant-covered hillslopes could enhance soil infiltration, thereby reducing runoff volume and concentrated flow scouring, ultimately achieving the objective of mitigating soil erosion (Lamm and Manges, 2000; Levia and Frost, 2003; Van Dijk and Bruijnzeel, 2001). The anti-erosion effect of alfalfa-covered hillslopes was primarily achieved through the canopy, root system, and variation in soil properties (Ghestem et al., 2014; Levia and Frost, 2003; Nardi et al., 2000). The alfalfa canopy decreased the rainfall intensity by redistributing the precipitation to increase stem flow and canopy interception (Van Dijk and Bruijnzeel, 2001), thus reducing the direct raindrops’ impact on the soil surface and minimizing soil aggregation disruption (Hu et al., 2023; Zi et al., 2024). Additionally, the presence of vegetation increased the surface roughness, which further reduced the overland flow velocity and mitigated the soil detachment and sediment transport by flowing water (Li et al., 2024a; Yao et al., 2023b). In this study, compared with CK, the total runoff volume and sediment yield on alfalfa-covered hillslopes decreased by 7.19% to 84.67% and 12.62% to 86.48%, respectively. Zi et al. (2024) reported that the runoff depth and sediment yields for bare soil were 1.19 to 1.64 times and 1.92 to 3.16 times those of hillslopes covered with 50% Nostoc commune, respectively. The difference in benefits in reducing runoff and sediment might be attributed to variations among plant species and herbage spatial distribution patterns (De Baets et al., 2006; Yao et al., 2023a).

The preventive and mitigating efficiency of root systems on soil erosion was primarily manifested through physical consolidation and chemical bonding (Gyssels et al., 2005; Wang and Zhang, 2017; Yao et al., 2023a). The chemical bonding by root systems primarily involved sugars, enzymes, and other organic compounds, which promoted the formation and stabilization of soil organic matter and soil aggregates, thereby enhancing soil stability (Atere et al., 2020; Le Bissonnais et al., 2018; Nardi et al., 2000). Meanwhile, soil particles were adsorbed by the roots, forming root–soil composites and enhancing soil erosion resistance (Entezami et al., 2024; Ghestem et al., 2014; Yang et al., 2024; Zhao et al., 2024). Additionally, the roots physically intertwined and compressed the soil, forming a network structure that improved soil stability, shear strength, and infiltration capacity (Song et al., 2025; Wang et al., 2022a, 2024a). In this study, compared with CK, the infiltration on alfalfa-covered hillslopes increased by 40.49% to 355.86%, demonstrating the role of root systems in improving soil permeability and infiltration. The finding of Liu et al. (2022b) indicated that the presence of Elymus tangutorum clearly increased the rainfall infiltration rate by 73% compared with CK. The difference in rainfall infiltration for alfalfa and Elymus tangutorum might be induced by different root types (De Baets et al., 2006; Entezami et al., 2024; Futerman et al., 2025; Gyssels et al., 2005). In general, the taproot system and the fibrous root system exhibited different soil anti-erosion functions (Le Bissonnais et al., 2018; Wang and Zhang, 2017).

The results indicated that total runoff and total sediment followed the order US, MS, LS, and SS, underscoring the complexity of vegetation spatial distribution in affecting the hillslope erosion dynamics (Feng et al., 2018; Hu et al., 2023). Vegetation located upstream of hillslopes (US) implied that the downstream areas lacked a direct shielding effect from vegetation, thus increasing the susceptibility of soil detachment and sediment transport by runoff (Li et al., 2024b; Niu et al., 2024). The absence of vegetation in downstream areas allowed the sediment to flow unobstructed downslope, increasing the concentrated flow erosion (Hu et al., 2023; Li et al., 2024b). In contrast, vegetation located downstream of hillslopes effectively intercepted sediments from upstream, reduced its downward transport, and facilitated sediment deposition via its dense root network and surface cover, forming a natural deposition zone (Futerman et al., 2025; Lin et al., 2024). This “sediment dam” effect greatly alleviated erosion in downstream areas, highlighting the critical role of downstream vegetation in mitigating erosion (Mu et al., 2019; Vinatier et al., 2024). Vegetation patterns located mid-hillslopes (MS) exhibited runoff and sediment reduction effects that lie between those for US and LS, which reduced the erosive force by reducing the velocity of overland flow and enhancing the hillslope surface roughness (Bahddou et al., 2023; Hu et al., 2023). Uniformly distributed vegetation enhanced the canopy interception, facilitated the runoff dispersion, and reduced the runoff velocity, thereby minimizing concentrated flow erosion (Bahddou et al., 2023; Oliveira et al., 2024). Additionally, the extensive distribution of root network formed a multi-layered erosion protection system (Entezami et al., 2024; Song et al., 2025). Thus, the SS pattern provided optimal erosion mitigation by changing runoff paths, improving soil stability, and reducing erosive energy.

This study investigated the effect of herbaceous vegetation spatial distribution patterns on hillslope erosion process with different slope gradients (8.75%–36.40%) and a rainfall intensity of 90 mm h^–1 on the Loess Plateau. The results indicated that alfalfa-covered hillslopes altered the soil erosion process by delaying the initial runoff generation time and reducing the runoff rate and sediment yield rate. The findings also highlighted the influence of vegetation spatial distribution patterns on hillslope erosion ranging from gentle to steep. The study enhanced our understanding on how herbaceous vegetation and its distribution patterns affected soil erosion and provided more scientific guidance to innovate soil and water conservation regulations on the Loess Plateau. Despite the significant results, the current study only validated the impact of a single herbaceous vegetation spatial distribution pattern on soil erosion. The effect of composite vegetation on soil erosion required additional verification. Moreover, the applicability of these findings needs to be validated in other erosion-prone regions.