Monitoring and Evaluation of Debris Flow Disaster in the Loess Plateau Area of China: A Case Study

Abstract

The Loess Plateau area, with complex geomorphological features and geological structure, is highly prone to geologic disasters such as landslides and debris flow, which cause great losses. To investigate the initiation mechanism of landslide and debris flow disasters and their spreading patterns, historical satellite images in the Laolang gully were collected and digitized to generate three-dimensional topographic and geomorphological maps. Typical landslides were selected for landslide thickness measurement using a standard penetrometer and high-density electrical method. Numerical models were established to simulate the occurrence and development of landslides under different working conditions and to evaluate the spreading range based on the propagation algorithm and friction law. The results show that the 10 m resolution DEM data are well matched with the potential hazard events observed in the field site. The smaller the critical slope threshold, the greater the extent and distance of landslide spreading. The larger the angle of arrival, the greater the energy loss, and therefore the smaller the landslide movement distance. The results can provide scientific theoretical guidance for the prevention and control of rainfall-induced landslide and debris flow disasters in the Loess Plateau area.

1. Introduction

As one of the most catastrophic geologic disasters [1,2], debris flow disasters cause serious economic losses and casualties every year, posing a great threat to socio-economic development and the safety of human life and property [3,4]. Debris flow monitoring is a prior means of the related research, which provides the basis for its theoretical study, experimental study, mechanism analysis, physical process analysis, and mathematical simulation, as well as the basis for debris flow early warning. Numerous scholars and research institutions that have conducted systematic research on the basic conditions and excitation factors of debris flow formation [5,6,7,8] have put forward the basis of discrimination of mudslide occurrence, carried out mudslide monitoring, and established the discrimination model of mudslides.

The occurrence of debris flow is closely related to the stability of the slope. The failure of a slope easily induces debris flow disasters. Therefore, slope stability analysis and debris flow disaster assessment are important research branches. Wahab et al. [1] conducted an experiment on a physical model by demonstrating the debris flow at different slope angles and provided a risk assessment approach to mitigate the impact regions of debris flow. Zhou et al. [3] selected 17 influencing factors to establish a hazard assessment model using a machine learning algorithm, which could provide support for the prevention of debris flow hazards. Rajabian and Shukla [9] developed an analytical approach based on the friction circle method to predict the safety factor of anchor-reinforced slopes using Taylor’s chart. The predicted results were in good agreement with the results of the limit equilibrium method. Dong et al. [10,11] established two innovative interval non-probabilistic reliability evaluation methods for the surrounding jointed rockmass, in which the uncertainty of the rockmass parameters were eliminated using the interval theory. Then, the established methods were successfully applied to the slope stability analysis of a tailings dam and achieved good results [12]. Charles et al. [13] proposed a new three-dimensional (3D) theoretical model to describe the effect of hydraulic mechanical reinforcement of the root system on the stability of initially unsaturated soil slopes. Ni et al. [14] investigated the effect of plant reinforcement on the stability of coarse-grained soil slopes and pointed out that the mechanical enhancement was only effective in the upper layers. Qu et al. [15] investigated the reliability of unsaturated soil slopes under rainfall infiltration using a 3D saturated water permeability rotating anisotropic random field. Subramanian et al. [16] recommended a two-dimensional numerical modeling approach for soil slope stability assessment considering the water content changes of the soil induced by seasonal climatic effects. Freeze–thaw action and snowmelt water infiltration were found to have considerable impacts on soil slope stability. Alberto et al. [17] determined the spatial relationship between landslides and glaciers and established a preliminary relative chronology. Pierson [18] summarized the kinematic, volumetric, and hydraulic characteristics of ten large historic volcanic debris flows from four different volcanoes. Fang et al. [19] introduced a self-similar percolation model and developed it from two dimensions to three dimensions, which explained why the joint on the other side of the block with the main river is easier to break than other places of the block. Gao and Zhang [20] comprehensively considered both the frequency and scale of landslides to analyze the landslide susceptibility of a small watershed. Chen et al. [21] performed a model study to debris flow and proposed the mechanisms of initiation of debris flow.

Scientific early warning of debris flow disasters is based on predictions before debris flow initiation [22,23,24] and warning in the movement [25,26,27]. The early warning model research is roughly divided into two categories; one establishes mathematical models through data analysis to formulate the critical rainfall of mudslides, and the other establishes early warning models through the analysis of the debris flow formation mechanism. Dai et al. [28] discussed the applicability of a variety of approaches to assessing landslide risk and proposed a landslide risk assessment and management framework for reducing landslide risk. For the early warning studies of critical rainfall, the research mainly focuses on determining the critical rainfall for mudslides and establishing a mudslide early warning model after statistically analyzing the rainfall characteristics of the region where the mudslides are located. After analyzing the data of mudslide occurrence in different places, many scholars have established a series of mudslide early warning models under different conditions of rainfall characteristics [29,30,31,32,33,34]. For the early warning research on the debris flow formation mechanism, it is mainly based on the critical conditions of debris flow initiation, to study the early warning indicators under different initiation critical conditions, and then establish debris flow early warning models. Timilsina et al. [35] analyzed the soil moisture patterns and evaluated the importance of soil moisture in predicting debris flow initiation sites. The results indicated that the effect of the soil moisture pattern on the initiation of debris flow is smaller than that of the vegetation and slope. Chatra et al. [36] conducted finite difference analysis of transient water flow through unsaturated–saturated soil to analyze the effects of rainfall intensity and duration on slope stability during rainfall events. The results showed that rain-induced instability is much greater in a loose soil slope than in a medium soil slope. Lourenço et al. [37] examined the relationship between soil particle wettability and slope processes in physical models and found that a continuous capping effect generated by water repellency was a necessary condition for the generation of runoff-initiated debris flow. Based on real-time monitoring and numerical simulation, Dong et al. [38] constructed a real-time pre-alarm system using the Internet of Things and cloud computing, through which the stable or dangerous warning signals of human-caused debris flow in a tailings dam can be obtained. Hu et al. [39] performed a series of flume tests to study how the initial soil moisture influences the initiation of debris flow and pointed out that the increase in pore pressure leading to soil liquefaction and the loss of the cohesive strength of the soil are two main triggering mechanisms of debris flows. He et al. [40] analyzed the disaster mechanism and evolution process of the landslide-debris-flow geohazard chain in areas with strong earthquakes through the consideration of the accumulative sliding amount and the critical rainfall. In their study, the influence of strong earthquake activities and rainfall caused the formation of the Niumian Valley geohazard chain. Niu et al. [41] applied the disaster chain theory to study preventive measures for mudslide disasters originating from source wells. They found that the type of disaster chain is a complex cyclical loop and proposed several proactive and reactive control methods.

Loess usually has high porosity, and its pore structure is more sensitive to water. Loess shows obvious plastic characteristics after absorbing water, and it becomes more susceptible to deformation and flow. In the process of drying and wetting, loess will undergo large volume changes, showing obvious expansion and contraction characteristics. As a result, it is susceptible to destabilizing landslide disasters. The Loess Plateau, as the main occurrence area of debris flow [20], has a unique combination of loess gully landforms and inter-valley landforms, such as loess, beams, and mounts. The very fragmented surface and serious soil erosion provide a rich material source base and conditions for the occurrence of debris flow, making the Loess Plateau the most serious area subject to mudslide disasters in China [42]. Therefore, systematic monitoring and early warning research on debris flow in the Loess Plateau area is particularly important and the most cost-effective means to reduce the damage caused by debris flow disasters.

2. Engineering Background and Research Methods

Gansu Province is located at the crossroads of the Qinghai–Tibet Plateau, the Inner Mongolia Plateau, and the Loess Plateau, with complex geological and environmental conditions, fragile ecology, concentrated rainfall, and frequent torrential rains. As a result of these influences, there are frequent geologic hazards in the region, and it is one of the provinces with the most serious geologic hazards in the country. In recent years, mudslide disasters have further aggravated the trend; in just five years after 2010, four consecutive large-scale mudslide disasters occurred, with a large hazardous range and high numbers of victims and economic losses. With the increase in extreme weather, the disaster situation is very serious, and the task of disaster prevention and mitigation is onerous.

Lanzhou is located in the northeastern edge of the Qinghai–Tibet activity block with strong neotectonic activity, in the composite part of the north–south seismic zone and Qilian Mountain seismic zone, with complex geological structure, strong neotectonic movement, and special topography and geomorphology, resulting in a narrow urban space. Coupled with the increasing intensity of human engineering activities and the frequent occurrence of extreme weather and climate events, the risk of geologic hazards is very serious. Lanzhou is deep inland, located in the transition zone between the monsoon climate zone and non-monsoon climate zone. According to the statistics of meteorological stations in Lanzhou City and neighboring districts, the average precipitation in Lanzhou City has been 300–600 mm for many years, the distribution of precipitation in the territory is very uneven, the regional precipitation decreases from the south to the north, the precipitation along the Qidaoliang area of the southern mountainous region is over 550 mm, and the precipitation along the Sangyuanzi area of the northeastern part of the country is over 550 mm. In the northeast, the precipitation is 250 mm in the area of Sangyuanzi, and under the influence of the monsoon, the distribution of precipitation in each month of the year is also very uneven in Lanzhou. Geological disasters in Lanzhou City are dominated by landslides, avalanches, and mudslides, with landslides and avalanches developed in a band-like and piecemeal manner, concentrated at the front of the mountains on both sides of the Yellow River and the front edge of the high terraces of the river valleys, while mudslides are distributed in a line-like manner along the riverbanks. As of the end of September 2015, the city had identified 2446 geologic hazard sites, including 346 landslides, 194 avalanches, 1639 unstable slopes (potential avalanches and landslides), 255 mudslides, and 12 ground collapses.

The Laolang gully is located in the south side of Lanzhou City, east of Gaolan Mountain, north of Minzu Village, Gaolan Mountain Township, Chengguan District, with geographic coordinates E 103°50′30″–103°51′6″, N 36°0′19″–36°1′31″. The study area is 950 m away from Lanzhou Station, and there are a large number of public facilities and residential areas near the area, and if a debris flow disaster occurs, it will seriously threaten the local transportation, economy, and the safety of residents’ lives and properties. There have been 91 loess landslides in the Laolang gully, and the loose piles generated by landslides and collapses provide sufficient material sources for the occurrence of mudslides. To scientifically prevent the occurrence of debris flow disasters in the Laolang gully, according to the spatial and temporal distribution characteristics of loess landslides and debris flows in the Laolang gully, meteorological observation, landslide observation, debris flow observation, and image observation systems are arranged in the debris flow monitoring area.

In this study, satellite images were used to circle the landslide extent and combined with DEM data to realize the three-dimensional visualization of the study area to further improve the accuracy of mudslide source investigation and analysis. The historical satellite images of the study area were first collected, and in order to explore the influence of external factors on the quality of satellite images, they were organized according to different shooting climates, month, time, and other factors, and compared and analyzed. Then, in order to realize the three-dimensional visualization of the study area, the 1:10,000 topographic maps of the study area were digitized in ArcGIS, and the DEM data were generated and combined with selected satellite images to generate three-dimensional topographic geomorphological maps, which were then interpreted, and the circled landslide extent was verified through on-site investigations. The landslides in the study area were categorized through field surveys, and there are two types of movement: sliding and collapse. The sliding movement includes rotational landslides and translational landslides. A rotational landslide refers to the movement of a sliding body along a sliding surface that curves upward, while a transitional landslide involves downward and lateral movement along a relatively flat sliding surface. Rotational landslides often occur on steep slopes with relatively uniform soil, where the sliding surface is curved, causing the landslide to rotate. Transitional landslides typically occur in rock layers, especially in the presence of weak layers, where the sliding surface is flat and the landslide mass moves along a plane. In the engineering field, the distinction can be made based on the following methods. Firstly, the shape of the two landslides are different: rotational landslides are usually curved and shaped like a semi-circle, whereas translational landslides are flatter and tend to be straight or slightly inclined. Secondly, the modes of movement of the two landslides are different: rotational landslides rotate and slide around a point, whereas translational landslides slide along a plane. Finally, the surface characteristics of the two landslides are different: rotational landslides may have depressions and bulges in their surface, whereas translational landslides generally have a more regular surface. There were 91 landslides in the monitoring area, of which the numbers of rotational landslides, translational landslides, and collapses were 49, 37, and 5, respectively. The distribution of the landslides is shown in Figure 1a. Rotational landslides and translational landslides accounted for 54% and 41% of the total number of landslides, respectively, and became the main types of landslides in the study area. Therefore, this study focused on the distribution characteristics and morphological features of rotational landslides and translational landslides.

To measure the landslide thickness, 19 representative landslides in the study area were selected for thickness measurement using a standard penetrometer, including 9 rotational and 10 translational landslides, which were uniformly distributed in different locations in the study area, as marked in Figure 1b. Use of a standard penetrometer is a common geotechnical engineering method to detect soil strength and divide strata. The hammer is controlled to hit the interceptor nut installed on the drill rod by manpower or automatic device, so that the drill rod penetrates into the soil layer at a certain depth, and then the soil strength parameter of the corresponding depth is obtained with the measured penetration hammer number N, and the soil strength parameters of the corresponding depth are calculated. To verify the reliability of the standard penetrometer method for determining the thickness of landslides, two landslides, as shown in Figure 1b, were selected for high-density electrical method measurement in this study. The measurements were compared with those of the standard penetrometer, and the reliability of the standard penetrometer method was examined on the basis of the degree of difference between the two methods. The high-density resistivity method combines the characteristics of electrical profiling and electrical sounding. Typically, it involves an A electrode and a B electrode supplying power, while an M electrode and an N electrode receive the potential difference to measure and calculate apparent resistivity and polarization. During on-site measurements, all electrodes are arranged at intervals on designated measurement points, with the main unit automatically controlling the variations in supply and receiving electrodes to complete the measurements. This paper selected the “dipole–dipole” measurement method to determine the thickness of the sliding body and utilized RES2DINV 5.0 software for data inversion, further acquiring detailed information on the sliding surface.

The standard penetration method can obtain the landslide thickness at the monitoring point with high accuracy, but it is only the result of a single point measurement, and the landslide thickness in the neighboring area needs to be inferred from the measurement result at the monitoring point. Conversely, the high-density electrical method can obtain the landslide thickness in the area, but the accuracy is lower than that of the standard penetration method. Numerical simulation methods can simulate the dynamic behavior of mudslides, predicting their propagation paths and influence range, which is of great significance for disaster risk assessment and the design of early warning systems. In this paper, the thickness of typical landslides was first measured by the standard penetration method, and then the thicknesses of two of them were verified by the high-density electrical method. Finally, after obtaining the basic data of landslides in the Laolang gully, the Flow-R numerical model was used to simulate the occurrence and development of mudslides under different working conditions and to evaluate the propagation range based on the propagation algorithm and friction law. The model performs simulation calculations by first identifying the source regions based on morphological and user-defined criteria and then simulates the movement of debris flow in these source regions based on the friction law and flow direction algorithm. The grid is generated using DEM data. To identify the source areas, three additional terrain data layers—slope, flow accumulation, and planar curvature—are integrated with lithology and land use information. Each input data set’s raster cell is categorized as (1) favorable, if there is potential for initiation; (2) excluded, if initiation is not possible; or (3) negligible, which can be disregarded in decision making. The integration of data follows these rules: if a raster cell is selected at least once as favorable but is never identified as excluded, it may be considered a source area.

3. Results and Discussions

3.1. Landslide Thickness Measurement

Figure 2 shows the relationship curve between the hit numbers and penetration depth for the 19 selected landslides. Taking the first landslide as an example, as shown in Figure 2a, in the depth range of 0–2 m, the number of hit numbers required for each 10 cm of penetration is 2–3, which indicates that the soil is relatively loose, and the hit numbers rise rapidly when the depth of penetration reaches −2 to −2.4 m. The number of hit numbers increases to more than 6 and begins to increase steadily when the depth of penetration is greater than 2.4 m. This indicates that the standard penetrometer has penetrated the sliding surface, and the degree of compactness has increased significantly. When the penetration depth is greater than 2.4 m, the hit numbers increase to more than 6 and start to increase steadily, which is 2–3 times higher than that in the range of 0–2 m. This indicates that the standard penetrometer has already passed through the sliding surface, and the compactness of the soil body has increased significantly. Figure 2b shows the corresponding curves of the hit numbers and penetration depth of the landslide using the high-density electrical method (section A–B). In the range of 0–8.2 m, the hit numbers of the standard penetrometer increase steadily, and at the depth of penetration of 8.2–8.4 m, the hit numbers increase suddenly and rapidly, and then after 8.4 m, the hit numbers tend to stabilize. So it is judged that the penetrometer has passed through the sliding surface, and the compactness of the soil has increased significantly. After 8.4 m, the hit numbers tend to stabilize, so it is judged that the penetration depth has reached the vicinity of the sliding surface.

The measurement results of the high-density electrical method are shown in Figure 3. The elevations of points A and B are 1810 and 1915 m, respectively. The A–B profile is the longitudinal profile along the main cross-section of the landslide, with the length of the measuring line of 174 m, and the spacing of the measuring points is 3 m. Analyzing the section view results of the high-density electrical method, we can obtain the following results: the resistivity of the surface layer of the slide varies between 22.0 and 64.9 Ω·m, and with deepening of the depth, the resistivity has a tendency to decrease gradually, especially at the depth range of 4–8 m. In the measuring line within the range of 5.07–38.07 m and 110.1–143.1 m, the value of the resistivity decreases in the range of 0.849–10.0 Ω·m, which is a high water content area. The resistivity increases to more than 192 Ω·m when the average depth exceeds 8.0 m, and it is inferred that the sliding surface has been reached. The length of the survey line is in the range of 86.1–101.1 m, and the high resistivity area appears when the depth exceeds 4 m. According to the field and satellite image investigation, the whole range of the electrical measurement is a huge landslide, and several small landslides have occurred within the range of the big landslide, and the measured high resistivity area is exactly the front edge of the sliding surface of the small landslide. The high water content zone appears at the depth of 4–8 m. This is because after the occurrence of landslides, many cracks appear around and inside the sliding body, and under the conditions of rainfall, water flow can quickly enter into the sliding body through these advantageous channels. The large number of fallout holes distributed in the study area is also an important factor for recharging the groundwater in the area of the sliding surface, while the loess layer below the sliding surface is not disturbed or is less disturbed; thus, the high degree of densification is unfavorable to the recharge of groundwater and presents a high resistance zone. Based on the above analysis, the average thickness of the slide was determined to be 8.0 m. The relationship curve between the number of hammer strikes and the penetration depth obtained by the standard penetrometer is shown in Figure 2b. Because the hit numbers increased abruptly to 23 times at the depth of 8.4 m, the determined penetration depth was 8.4 m, corresponding to the thickness of the slide of 7.8 m, and the difference from the results of the electrical method was 0.1 m. The error rate was 1.3% relative to the results of the electrical method.

Section C–D is a longitudinal section along the main section of the landslide, with a total length of 275 m and a spacing of 5 m. Analyzing the electrical profile, we can obtain the following results: the resistivity of the surface layer of the landslide varies from 21.5 to 56.5 Ω·m, and with deepening of the depth, the resistivity has the same tendency to decrease gradually. In the depth range of 3–9.5 m and the length of the survey line of 15.0–110.0 m, the value of resistivity generally decreases to below 10.0 Ω·m, which is the area of high water content, especially in the range of 115–165 m. The resistivity of the surface layer of the sliding body varies between 21.5 and 56.5 Ω·m, and with depth deepening, the resistivity also has a tendency to decrease gradually. In particular, in the range of 115–165 m, the resistivity value is as low as about 8.19 Ω·m, even from the surface. When the average depth exceeds 9.5 m, the resistivity increases to more than 148 Ω·m, and it is inferred that the sliding surface has been reached. Similar to the A–B profile, the high water content zone appearing at the range of depth of 3–9.5 m is mainly due to the recharge of water flow to the sliding zone from advantageous channels, such as cracks and drop holes, as a result of rainfall, and to the low resistivity feature appearing in the range of the survey line length of 115–165 m. Combined with the topographic map of this section of the region, we found that the range of the measurement line is a concave terrain, which is very conducive to the pooling of groundwater, so the landslide has the conditions for the occurrence of secondary landslides, and there should be a focus on monitoring. The loess layer below the sliding surface is unfavorable to groundwater recharge due to its high degree of densification, presenting a high resistance zone. Based on the above analysis, the average thickness of the slide was determined to be 9.5 m. The relationship curve between the number of hammer strikes and the penetration depth obtained by the standard penetrometer is shown in Figure 2o. Since the number of hammer strikes increases abruptly to 20 times at the penetration depth of 9.5 m, the penetration depth of 9.5 m corresponded to a thickness of the sliding body of 8.9 m, with a difference of 0.6 m, and the error rate relative to the results of the electrical method was 6.7%.

Combined with the characteristics of the two methods of measurement, the error was mainly due to two reasons: the first is that the electrical method of exploration of the thickness of the sliding body yields an average value, which cannot fully reflect the thickness value of a location point, while the standard penetration measurement is based on the measurement value of a point on the sliding surface to reflect the depth of the entire sliding surface, which leads to error between the two results; the second is that, in practice, the local ground surface as well as the sliding surface undulation have an impact on the results of the standard penetration test measurement. However, comparing the two measurement results, it can be seen that the error of the standard penetrometer measurement results relative to the results of the electrical method was less than 7%; therefore, the measurement results of the standard penetrometer can more accurately reflect the thickness of the sliding body. At the same time, the small size of this instrument makes it easy to carry and perform measurements, which is another advantage. So in the Loess Plateau region, the standard penetrometer can be used as an effective and fast tool to measure shallow loess slide thickness. As for the high-density electrical exploration instrument, although its accuracy is very high, it can reflect the distribution of strata in the main section of the landslide in detail, and it can analyze the detailed distribution pattern of groundwater in the landslide body, which is of great significance for the analysis of the cause mechanism of landslides, early warning, and other work, due to the large size of the instrument and high time cost and economic cost of measurements, it is not applicable to measuring the thicknesses of large numbers of landslides in watersheds of various scales, and it is difficult to carry out the work in complex and high-slope terrains. Therefore, the use of a standard penetrometer is more suitable than electrical surveying for determining the thickness of shallow loess landslides.

3.2. Risk Assessment of Debris Flow at Different Topographic Resolutions

DEM data with different resolutions were selected to analyze the ability of DEM data to identify potential source areas. DEM data with 5 m, 10 m, and 20 m resolutions at the watershed scale of the Laolang gully were selected, and the same source-area identification criteria and motion-diffusion parameters were used. Hazard evaluations under four different conditions, namely, flood, collapse-affected debris flow, landslide-affected debris flow, and debris flow under the simultaneous influence of landslide and collapse, were analyzed at different topographic resolutions.

The simulated flood source areas and motion spread ranges are shown in Figure 4 and Figure 5, respectively. It was found that large numbers of small depressions on the slope surface were misidentified as potential flood source areas in the 5 m resolution DEM data, which greatly increased the range of potential source areas, and the flood events generated by these potential source areas were quite different from the actual situation, which had a certain degree of error for correctly evaluating the risk of the disaster. The coarse resolution of the 20 m resolution DEM data was particularly unfavorable for correctly simulating the source area and the range of motion. Compared to the 5 m and 20 m resolution DEM data, the 10 m resolution DEM data were considered to be a good simulation data source for the region, and the simulated results matched well with the potential flood events observed in the field.

Considering that the collapse on both sides of the channel provides a good material source for the formation of debris flow, different precision DEM data were used to analyze the evaluation of debris flow hazard under the influence of collapse, and the simulation results are shown in Figure 6 and Figure 7. The source area and the movement process of the simulation with data resolutions of 5 m and 20 m were not consistent with the reality, but the simulation results with 10 m resolution data were reasonable for the evaluation of the risk of debris flow. However, the simulation results with 10 m resolution data were more reasonable for the evaluation of the mudflow danger. After the field investigation, it was found that, due to the natural erosion of the stream at the bottom of the gully, the lower part of the slope was hollowed out, which led to separation of the soil from the slope and then it fell down, and there were a number of small and medium-sized avalanches on both sides of the main gully extending in the N–S direction. The debris formed by avalanches affects the topography of the local area to a certain extent, and at the same time, these debris can be regarded as a good source of debris flow hazard. It is believed that the debris on both sides of the channel provides part of the energy for the movement in the mudslide channel, and with the frictional consumption of the movement process, the energy is gradually dissipated at the mouth of the channel, and the movement stops.

Analyzing the spreading results of landslide-influenced debris flow with different precision DEM data, as shown in Figure 8 and Figure 9, the comparison revealed that the simulation results at 10 × 10 m resolution were more reasonable compared with the results at other resolutions. After the field engineering geomorphological mapping and remote sensing image interpretation, it was found that large numbers of landslides were distributed on the slopes on both sides of the gully in the study area, the number of landslides at the mouth of the gully was significantly higher than that at the end of the gully, the landslides were mainly distributed in the gully entrance area where the gradient is steeper, but most of them were mainly small-sized landslides, and the tail area was slower due to the slope gradient and the vegetation cover was higher, so there were fewer landslides on the steep slopes with less water than those at the mouth of the gully, but at larger scales. The occurrence of landslides is often accompanied by the formation of large-scale joint cracks, which has a greater impact on the local hydrological conditions. At the same time, the topography and geomorphology have certain roles in changing the local gradient and catchment conditions of the slope. Therefore, affected by a large number of landslides on the slopes on both sides of the channel, the debris flow events in the watershed were re-identified and simulated. Landslides on the slopes provide large numbers of material sources for debris flow, and the slopes on both sides of the central part of the channel are flat with no obvious tributary ditch development, so that the expansion of the material source movement on both sides has a larger range.

Under extreme climatic conditions, landslides and collapse usually occur at the same time. In order to analyze the occurrence of debris flow under extreme conditions, both collapse and landslide were considered as potential sources. Based on the landslide source area, the collapse disaster area at the head of the upstream gully was further identified. It was analyzed that the generation of large numbers of material sources provides great kinetic energy for the movement of debris flow. Compared with the debris flow movement process only considered to be affected by landslides, the debris flow movement process under such extreme conditions still has large energy at the mouth of the trench, and the simulation results are shown in Figure 10 and Figure 11.

By simulating the identification of source areas under four different scenarios of flood, collapse-induced debris flow, landslide-induced debris flow, and debris flow under the simultaneous influence of collapse and landslide, it was found that the DEM data with a resolution of 5 m were the most refined, which not only identified the potential source areas on both sides of the near-channel but also mistakenly identified many small-scale topographic undulations on the slope surface as source areas, resulting in simulated movement results that did not correspond to the actual situation. The DEM data with a precision of 20 m were too coarse, and the potential source areas in some branch channels were not well identified, and many small-scale source areas were ignored. The disaster area identified by the 10 m resolution DEM data, although still somewhat different from the actual situation, was the most consistent result compared to the 5 m and 20 m resolution results.

3.3. Risk Assessment of Debris Flow at Different Critical Slope Thresholds

The effect of critical terrain slope for source area identification was analyzed with the 10 m resolution DEM data. Three critical terrain slope thresholds of 15°, 25°, and 35° were analyzed separately to identify flood source areas without considering the effect of landslides from avalanches. It was found that areas smaller than this slope were excluded from the potential source area due to the increase in the source area initiation slope threshold, as shown in Figure 12 and Figure 13. Thus, as the critical terrain slope for source area identification increased, the identified source areas gradually decreased. In addition to some of the source areas in the upstream branches of the main ditch, some of the source areas in the middle and lower reaches of the ditch were also controlled by topographic features and could not be well identified when the source area initiation slope threshold was large. When the source area initiation slope threshold increased, the material that can undergo flow decreased, and the range and distance of movement spreading were relatively reduced. The evaluated disaster-prone area may be too small, and the obtained hazard evaluation was not reasonable.

Three critical terrain slope thresholds of 15°, 25°, and 35° were analyzed to identify debris flow source areas considering the combined effects of collapse and landslide. As the topographic slope threshold increased, the range of material sources at the identification site on the slope decreased, and some source areas larger than the topographic slope threshold could not be identified, as shown in Figure 14 and Figure 15. Therefore, the closest potential source accumulations need to be identified by selecting the appropriate terrain slope threshold based on the potential sources investigated in the field. The reduction in physical sources causes the debris flow material in the trench to have less kinetic energy of movement, the energy to reach the trench entrance and exit quickly disappears, and the distance of movement becomes shorter.

3.4. Risk Assessment of Debris Flow under the Spreading of Different Substances

The motion paths and distances of the same material source are often constrained by potential energy, topography, and other conditions, even if the same material source area often has different motion diffusion paths and diffusion ranges under different conditions. The motion diffusion path in the model is mainly controlled by the index x in the Holmgren model, and the four values of x were selected as 2, 4, 6, and 8, respectively, to analyze the influence of the index x in the Holmgren model on the motion process for the raster data with 10 m resolution. Figure 16 shows the spreading process of debris flow under the influence of both collapse and landslide, and it was found that the spreading range tended to decrease and the flow rate converge rapidly with the increase in the index x. Figure 17 shows the spreading range of motion of the flood, and it was similarly found that as the exponent x became smaller, its spreading range to both sides of the channel became wider. The two results reflected that the index x in the Holmgren model mainly affected the degree of diffusion to both sides during the movement process, and when the index x was larger, the diffusion range was larger, all other conditions being equal. The index x in the model is an indirect reflection of the representation of the way of motion diffusion. The model is based on a simple energy balance and, to a certain extent, can represent the energy of the moving material, and so the greater the energy of the moving material, the wider the path of motion, with a greater range of diffusion.

For the flood disaster area identified at 10 m resolution, the difference in its motion process simulated by changing the angle of arrival parameter in the simplified friction-limited model (SFLM model) was determined to analyze the effect of the motion distance of this parameter. The simplified friction limit model is based on the maximum possible distance of motion and is characterized by the smallest travel angle, also known as the angle of arrival, which is the angle between the connection line and the horizontal line between the source area and the farthest point reached by the debris flow along its path.

Because a large number of international studies have concluded that the value of this parameter is mostly in the range of 7~11°, three angles of 7°, 9°, and 11° were selected for comparative analysis, and the simulated flood hazard evaluation under the consideration of rainfall surface runoff only is shown in Figure 18. It was found that the maximum movement distance decreased as the angle increased. It was found that when the angle of arrival was set to 11°, the maximum movement distance had not reached the gully entrance and the movement process had stopped. By contrast, when the angle of arrival was set to 7°, the kinetic energy at the mouth of the ditch was not yet fully dissipated and would move further northward out of the ditch until the kinetic energy was reduced to 0, stopping the flow buildup to form a stacked fan. The results of the debris flow hazard evaluation under the consideration of the impact of landslide of avalanche are shown in Figure 19. Large numbers of material sources provided sufficient energy, and the debris flow still flowed out of the gully even when the arrival angle was set to 11°, but the comparison revealed that the energy at the mouth of the gully decreased compared to that at 7° and 9°, and it could be deduced that after flowing out of the gully, it had the shortest distance to move. Since the SFLM model is based on a simplified energy balance, the larger angle represents that the energy loss due to friction is also serious. Therefore, this parameter has a certain physical basis and has better practicality. When the particles of the debris material are rough and the friction between the particles is enhanced, the angle tends to be larger and the distance of movement is shorter.

4. Conclusions

Based on the acquired data of landslides in the Laolang gully, a 1:10,000 topographic map of the study area was digitized and 19 representative landslides were selected for this study. The thicknesses of the landslides were tested using a standard penetrometer. The accuracy of the measurement results were verified against the high-density electrical method. At last, a numerical model was established to simulate the occurrence and development of mudslides under different working conditions, and the propagation range was evaluated based on the propagation algorithm and the friction law. The simulation results show that:

(1)

A too large or too small DEM data resolution will cause the misjudgment of debris flow source area, and the diffusion range of debris flow is not very reasonable. The 10 m resolution DEM data were considered to be good simulation data in the region, and the simulated results matched well with the potential disaster events observed in the field, which enables a reasonable evaluation of the risk of mudslides.

(2)

By comparing the simulation results with the actual loose accumulations in the study area, the reasonable topographic conditions for debris flow occurrence were found. It was observed that a lower critical slope threshold allows more loose material to flow, leading to a wider spread and greater distance of the landslide.

(3)

By changing the angle of arrival parameter in the SFLM model, we simulated the difference in the landslide movement process under different working conditions. The greater the angle of arrival, the more energy is lost due to friction, resulting in a shorter landslide movement distance.