Gansu Province is in northwest China at the junction of the Loess Plateau, the Qinghai–Tibet Plateau, and the Inner Mongolia Plateau, and has complex and diverse geological conditions. Wen County in Longnan City is in southeast Gansu Province (Figure 1). It is a zone of intersection between the westward extension of the western Qinling Mountains and the eastward extension of the Minshan Mountain range. The terrain is steep, the geological structure is well developed, seismic activity is frequent, the ravines are vertical and horizontal, and high mountain valleys are interspersed and distributed. Landslides and collapses are prevalent in the gullies, and loose materials are abundant. The region is located south of 33° North latitude and belongs to the marginal zone of the subtropics. The average annual precipitation in the area in the last decade was 513.2 mm/year, with abundant rainfall that provided favorable conditions for the formation of debris flows. There are a total of 256 large and small gullies for debris flows in the county, with a developmental density of 0.051 gullies/km². Debris flows mostly occur from June to September each year and are all induced by rainfall (Zhou et al., 2022; Zhu et al., 2024; Yu et al., 2016; Xie et al., 2019). They occur suddenly and with great momentum, pose a direct threat to important facilities, such as the towns, villages, and trunk lines for traffic in the area, and severely affect the normal life and order of production of the locals (Cui et al., 2015; Wang et al., 2024). To reduce the damage caused by them, interception dams and drainage ditches have been built in many basins, and have helped prevent major disasters.

We surveyed and collected data on 67 check dams installed in 24 representative debris flow gullies in Wenxian County, Gansu Province (Figure 2). The results are shown in Table 1. The forms of damage to the check dams included overall rush destruction, and damage to the auxiliary dams, their overflow holes, and their shoulders and foundation. A comparison of closed and permeable check dams showed that the rate of rush destruction of closed dams was 24% and their rate of silting (the check dams had been silted up) was 83%, while the rate of rush destruction of permeable dams was only 6% and their rate of silting was 57%. This shows that installing drainage culverts can help regulate the composition of debris flows, extend the silting time of the storage capacity of check dams, and improve their service life. We also found during our investigation that check dams with higher opening rates still had empty reservoirs after more than 10 years of operation, which means that prevention and treatment had not had a significant effect. Therefore, a higher opening rate was not necessarily better for preventing silting behind the dam. A comparison of the status of operation of silted and non-silted check dams showed that the rate of damage to the former was 48%, while that to the lat ter was less than 5%. This shows that once the storage capacity of the check dams had been silted up, the probability of damage to them increased. This is because, on the one hand, the water level in front of the dam had increased to lift debris flows and increase the area of contact between them, and the overflow holes and dam shoulders to enhance erosion and destruction. On the other hand, the increased energy of debris flows increased impact-induced damage to the foundation or the revetment downstream once the debris flows had overflowed from the dam.

The tests were carried out in the landslide and debris flow laboratory of the Institute of Geological Natural Disaster Prevention of the Gansu Academy of Sciences. The experimental setup is shown in Figure 3 and mainly included brackets, water tanks, flumes, areas of deposition, sedimentation pools, and data acquisition systems. The flume was 6,500 mm long, 500 mm wide, and 800 mm high. The bottom was made of stainless steel and the sides were made of transparent organic glass. The contact surface was smooth, because of which the influence of boundary effects on the results could be ignored.

The range of flume gradients was 0°–40°. The gradient of the gully in which the sampling point was located was approximately 9° and was used as the flume gradient in the tests. After removing particles larger than 10 cm, the gully deposit was simplified into a trapezoid with the dimensions shown in Figure 3A. The total mass of the deposit was 250 kg, and its density was 1.68 g/cm³. According to a preliminary test, 0.9 m³ of water discharge at a time could ensure that the deposit in the ditch was completely activated to form debris flows.

Figure 4 shows that the body of the dam was placed 2,500 mm from the outlet. The area between the water tank and the dam was the zone of flow of clean water to ensure that it could smoothly flush through the body of the dam. The check dam was set at 5,200 mm from the outlet to ensure that the soil and water were fully mixed to form the debris flow before encountering the check dam. Sensors to detect the pore water pressure and volumetric water content were installed behind the dam to monitor changes in them during the check of debris flow. Two sensors were arranged in each group at distances of 4,650 mm (Location 1) and 4,950 mm (Location 2) from the outlet. In addition, an HD camera was arranged on each of the upper parts of the check dam, the side face of the water tank, and the opposite side of the test platform to fully observe the phenomena.

Adjacent check dams are usually far apart in practice, and thus rarely protect one another. Most check dams in a gully operate independently. Therefore, we used a single dam form and changed only the number of drainage holes in the check dam (to change the rate of its opening) to simulate its protective effect to determine the optimal range of its opening rate. Figure 5 shows that we designed five types of dams. The height of each drainage hole was 65 mm, its width was 30 mm, and the radius of the arc was 15 mm. The rates of the opening of Dams I to V were 2.1%, 4.2%, 6.3%, 8.4%, and 10.5%, respectively.

Figure 7 shows the curve of variations in the volumetric water content at different locations behind the dam under the action of debris flow. It exhibited an overall “single-peak” pattern of variation, that is, it increased rapidly to its maximum value after the formation of debris flow and then decreased rapidly to a relatively stable value. The final volumetric water contents after the stabilization of debris flow for dams I–V were 45.5%, 45.6%, 43%, 32.9%, and 32.9%, respectively, showing an overall trend of a gradual decrease with an increase in the opening rate. When the opening rate was 6.3%, the final volumetric water content did not change.

Figure 8 shows the curves of variations in the pore water pressure over time. They can be divided into three stages: stages of stability, rapid increase, and recession. The curves exhibited an overall “single-peak” pattern. That is, after reaching the peak, the pore water pressure began to decrease slowly and then tended to stabilize. Because location 1 first encountered the debris flow, PW1 generally occurred earlier than PW2. Under the impact of the head of the debris flow, the value of PW1 fluctuated to a greater extent than that of PW2. Because the conditions of the formation of debris flow and the slope of the flume remained constant during the test, the initial energy of the debris flow was fixed. The pore water pressure attained its peak after 43 s for dams I–IV. For dam V, which had a higher permeability than the other dams, the pore water pressure reached its peak after 40 s. The peak values of PW2 for dams I and II were 2.42 kPa and 2.39 kPa, respectively, and the drop in these values was not significant. The peak value of PW2 of dam III dropped sharply to 1.52 kPa, a decrease of 36.4%. The peak values of PW2 of dams IV and V were 1.52 kPa and 1.58 kPa, respectively, nearly identical to that of dam III. This indicates that as the opening rate of the dams increased, the efficiency of debris flow passing through the check dams improved, the effective area of contact between the debris flow and the check dam decreased, and the pore water pressure therefore decreased accordingly. However, the pore water pressure behind the dam did not exhibit a simple trend of decrease. When the opening rate was 6.3%, the pore water pressure did not change significantly, which is similar to the law of change in the volumetric water content.

Once the experiment had been completed and the debris flow in the flume had completely stopped running, we considered the check dam as the x-axis and the gully bed along the longitudinal direction as the y-axis to measure the morphology of the deposits behind the check dam. We used vertical and horizontal spacings of 50 mm to insert a steel ruler into the deposits behind the dam to measure the depth of mud (Figure 9) and used this information to generate a 3D graph of the deposits (Figure 10). It clearly and intuitively showed the morphology of the deposit when debris flows passed through the different types of check dams.

Figure 10 shows that owing to the low porosity of dams I and II, the deposit behind these dams was significantly higher than those behind the other types of dams. The deposit was particularly high within 60 cm behind the dam, with maximum heights of 16.3 cm and 15.7 cm for dams I and II, respectively (Figures 10A, B). The height of the deposit then decreased rapidly. For dams III–V, which had relatively large pores, the morphology of the deposit behind the dam was relatively gentle, with maximum heights of 12.9, 10.3, and 13 cm, respectively. This indicates that when the opening rate of the check dam was 6.3%, the morphology of the deposit behind it tended to be gentle. Continually increasing the opening rate had little effect on the maximum height of the deposit behind the dam.

The volume of the sedimentation body behind the dam is obtained by 2D volume integration of the sedimentation shape surface behind the dam, as shown in Figure 11. The sedimentation volume of Dam-I and Dam-II is close to each other, while the sedimentation volume of Dam-III decreases rapidly and then slowly. When the opening rate is greater than 8%, the sedimentation volume behind the dam remains unchanged, indicating that a good effect of water-sediment separation has been achieved. Further increasing the opening rate has little effect on improving water-sediment separation. But it may affect the structural strength of the dam.

Retaining the coarse and sluicing the fine is an important function of the check dam. Intercepting large particulate material can reduce the kinetic energy of debris flows and mitigate their destructive power while draining fine particulate material to downstream reduces the burden on the storage capacity of the check dam and improves its efficiency. By comparing and analyzing the particle size distribution of the original soil samples and deposits behind the dam, we quantitatively assessed the effects of the interception of coarse particles and draining of fine particles by the check dam. We dried the original soil samples and deposits behind the dam and then used 500 g of each for a sieve analysis test to obtain the accumulated grading curves of the deposits behind different types of dams, as shown in Figure 12. Compared with those of the original soil samples, the grading curves of samples from different types of dams shifted to the right. That is, the overall particle size became coarser, and this verifies the interception of coarse particles and the draining of fine particles by the check dam.

The Lagrange interpolation method was used to determine the median particle diameter d ₅₀ of deposits to characterize their degree of coarsening (Cui et al., 2015; Wang et al., 2024), as shown in Table 3. There was no significant linear relationship between the degree of coarsening of the deposits behind the dam and its opening rate. The maximum value of d ₅₀ of dam II was 2.82 mm, about 2.9 times that of the initial soil sample, while dam V exhibited a more significant extent of reduction, with its value of d ₅₀ decreasing to 2.21 mm. This shows that simply increasing the opening rate of the dam did not maximize its interception of coarse particles. The results showed that there was an optimal interval in this regard. The average value of d ₅₀ in the test was 2.60 mm in the range of opening rates of 2.1%–10.5%. These deposits behind the dam were coarser than the initial soil sample by about 2.65 times.

In addition, the particle size distribution of the deposits behind the check dams with different opening rates was not the same. To analyze this trend for different types of dams, we plotted curves of the particle size distribution of deposits behind the dams as shown in Figure 13. All the soil samples exhibited a law of bimodal distribution. Under varying opening rate conditions, the particle size distribution of the deposits behind the dam changes from 0.5–1 mm to 2–5 mm, with more than 40% falling in this range. Dam IV had the highest concentration of 48%, followed by Dam III with 46%. Figure 12 shows that the value of d ₅₀ of the deposits behind the dam increased when the ratio of particles with sizes in the range of 2–5 mm increased.

The conditions used to trigger debris flow were the same in all our experiments. The slope of all channel beds was 9° and the locations of the check dams were fixed. The results show that it took 2 s for the debris flow to reach the front of the dam, indicating that its initial velocity was the same in each case, i.e., all the debris flows had equal initial energy. As shown in Equation 1, the initial energy was converted into dissipation energy generated during the interaction between the debris flow and the check dam, according to the principle of energy dissipation (Gao et al., 2022), and constituted the final energy that continued to move downstream: E 0 = E ′ + E 1 (1) where E ₀ is the initial energy of the debris flow (J), E′ is the dissipation energy generated during its interaction with the check dam (J), and E ₁ is the final energy of the debris flow after overflowing past the check dam (J).

Because the initial energy E ₀ was constant, the energy output by the debris flow downstream was the minimum when the dissipation energy was the maximum, that is, the check dam was able to optimally regulate the energy of the debris flow. We can thus assess the regulatory effects of different check dams by comparing the dissipation energy generated by them. However, due to constraints imposed by different types of debris flow, the shape of the check dam, and the characteristics of the material, no accurate means of measuring the dissipation energy is currently available. Some researchers have defined the ratio of the pore water pressure behind the dam to its normal stress (Hübl et al., 2003; Tian et al., 2017) during the interaction between it and the debris flow as the liquefaction coefficient. This can be indirectly used to characterize the magnitude of the dissipation energy. The liquefaction coefficient is calculated as follows in Equation 2: L = σ w σ t = σ w ρ 0 g h ⁡ cos ⁡ θ × 100 % (2) where L is the liquefaction coefficient (%); σ _w is the pore water pressure behind the dam (kPa); σ _t is the normal stress behind the dam (kPa), ρ ₀ is the initial density (g/cm³); h is the height of mud behind the dam, and is taken as its maximum accumulated height (cm); g is the gravitational acceleration (m/s²); and θ is the slope of the channel bed, which was set to 9° in this paper.

By comparing the liquefaction coefficients of check dams with different opening ratios, their regulatory effects on the energy of the debris flow can be quantitatively evaluated. The curve of the relationship between check dams with different opening ratios and their liquefaction coefficients is plotted in Figure 14. It is evident from it that the liquefaction coefficient of Dam II was the highest, 91.6%. That is, the interaction between this check dam and the internal debris flow was the most intense, generated the maximum dissipation energy and had the best effect in terms of energy dissipation on the debris flow. This may also be why the quality of the body of deposit behind dam II was the highest and its particles were the coarsest: A large number of solid particles in the debris flow collided and rubbed behind the dam, and finally the energy was completely dissipated causing the deposit to tend towards stillness. Following this, as the opening ratio increased, the liquefaction coefficient decreased rapidly, the effect of dissipation in the energy of the debris flow due to the check dam weakened, the debris flow developed better penetrability, and thus the maximum height of deposition behind the dam decreased significantly compared with that in case of dam II (Figure 10).

A core function of check dams is to retain solids in the debris flow. The capture coefficient is commonly used to quantitatively evaluate their effects on solids in the debris flow (Hübl et al., 2003): T = M i M 0 × 100 % (3) where T is the capture coefficient (%), M ₀ is the total mass of solid matter in the debris flow (kg), and M _i is the mass of solid matter in the debris flow that is retained by the check dam (kg).

In our experiments, the body of the deposit in the flume was scoured and entrained into the debris flow by the incoming water. Some of the solid matter was transported downstream over the check dam along with the debris flow, while the remainder was captured by the check dam to form the deposit behind it. Therefore, the capture coefficient under different conditions of the dam was calculated by using Equation 3, and the curve of the relationship between the capture coefficient and the opening ratio of the dam was plotted, as shown in Figure 14. It shows that the curve exhibited an overall “single-peak” trend of change. The capture coefficient was the maximum at 30.6% for dam II, in which case the effect of retaining solid materials was the best. Subsequently, the capture coefficient continued to decrease as the opening ratio increased.

The regulatory effects of check dams on debris flows are mainly related to two aspects: regulating energy and solid matter (Lin et al., 2017). Moreover, the safety of check dams is also an indispensable factor in prolonging their service life (De Haas et al., 2015). A balance must be achieved between reducing the energy of debris flow and the removal of solid matter from it, and the safety of the dam during the construction of check dams (Chen et al., 2015). Figure 14 shows that better results were obtained when the coefficient of capture of the check dam was greater than its liquefaction coefficient. Because the benefits brought about by such retention could compensate for some of the damage to the check dam caused by the impact of debris flows, it is suitable to set an opening ratio in the range of 4.2%–6.3%, as shown in Figure 14. At the same time, the opening ratio of check dams should be reasonably selected according to the specific conditions of debris flow on-site.

For example, if local debris flows have a high frequency of occurrence and there is a large number of loose deposits in the ditch, dam II (opening ratio, 4.2%) is the most appropriate choice for regulating the energy and materials of debris flows. But this imposes stringent requirements on the strength of the dam, the service life of which can be extended by building groups of check dams. If local debris flows have a low frequency of occurrence, dam III (opening ratio, 6.3%) is the best choice. While ensuring minor damage to the dam itself such that it has a longer service life, this choice can help retain solids from the debris flow.