Avalanche Hazard Dynamics and Causal Analysis Along China’s G219 Corridor: A Case Study of the Wenquan–Khorgas Section

Abstract

Investigating avalanche hazards is a fundamental preliminary task in avalanche research. This work is critically important for establishing avalanche warning systems and designing mitigation measures. Primary research data originated from field investigations and UAV aerial surveys, with avalanche counts and timing identified through image interpretation. Snowpack properties were primarily acquired via in situ field testing within the study area. Methodologically, statistical modeling and RAMMS::AVALANCHE simulations revealed spatiotemporal and dynamic characteristics of avalanches. Subsequent application of the Certainty Factor (CF) model and sensitivity analysis determined dominant controlling factors and quantified zonal influence intensity for each parameter. This study, utilizing field reconnaissance and drone aerial photography, identified 86 avalanche points in the study area. We used field tests and weather data to run the RAMMS::AVALANCHE model. Then, we categorized and summarized regional avalanche characteristics using both field surveys and simulation results. Furthermore, the Certainty Factor Model (CFM) and the parameter Sensitivity Index (Sa) were applied to assess the influence of elevation, slope gradient, aspect, and maximum snow depth on the severity of avalanche disasters. The results indicate the following: (1) Avalanches exhibit pronounced spatiotemporal concentration: temporally, they cluster between February and March and during 13:00–18:00 daily; spatially, they concentrate within the 2100–3000 m elevation zone. Chute-confined avalanches dominate the region, comprising 73.26% of total events; most chute-confined avalanches feature multiple release areas; therefore the number of release areas exceeds avalanche points; in terms of scale, medium-to-large-scale avalanches dominate, accounting for 86.5% of total avalanches. (2) RAMMS::AVALANCHE simulations yielded the following maximum values for the region: flow height = 15.43 m, flow velocity = 47.6 m/s, flow pressure = 679.79 kPa, and deposition height = 10.3 m. Compared to chute-confined avalanches, unconfined slope avalanches exhibit higher flow velocities and pressures, posing greater hazard potential. (3) The Certainty Factor Model and Sensitivity Index identify elevation, slope gradient, and maximum snow depth as the key drivers of avalanches in the study area. Their relative impact ranks as follows: maximum snow depth > elevation > slope gradient > aspect. The sensitivity index values were 1.536, 1.476, 1.362, and 0.996, respectively. The findings of this study provide a scientific basis for further research on avalanche hazards, the development of avalanche warning systems, and the design of avalanche mitigation projects in the study area.

1. Introduction

An avalanche is a significant natural disaster that occurs in areas with seasonal snowfall. It is characterized by snow masses rapidly descending from mountainsides under the influence of gravity, temperature, and other factors [1]. Avalanche hazards are characterized by their sudden onset, latent nature, unpredictability, rapid movement, and significant impact force. These characteristics can severely affect the sustainable development of transportation networks in mountainous regions [2]. For instance, in 1998, a catastrophic avalanche event occurring along a roadway near Evolene village, Switzerland, resulted in 12 fatalities as the snow mass destroyed dwellings along its trajectory and overwhelmed the transportation corridor [3]. In 2018, a snow avalanche incident in the Indian-administered Jammu and Kashmir region resulted in 10 fatalities after a light passenger vehicle was engulfed by the rapidly moving snow mass, with the catastrophic event completely overwhelming local transportation infrastructure [4]. The catastrophic avalanche event at the Duoxiongla Tunnel portal in Nyingchi, Tibet Autonomous Region, China, on 17 January 2023 resulted in 28 fatalities following the complete inundation of this critical transportation artery, subsequently being designated as one of China’s Ten Most Significant Natural Disasters for that calendar year [5]. The progressive expansion of cryospheric economic activities coupled with accelerated alpine transportation infrastructure development has driven human settlements deeper into avalanche-prone terrains, thereby exponentially heightening disaster risk exposure levels across these vulnerable mountain corridors [6,7]. In response to the increasingly severe avalanche hazards, avalanche early warning systems and engineering mitigation measures are the primary means for managing risks in areas prone to frequent avalanche events. Furthermore, conducting surveys and analyzing the causes of these disasters are essential tasks for advancing research on avalanche early warning and mitigation strategies.

For the needs of disaster research, disaster warning, and prevention and control engineering design, countries generally adopt methods such as on-site survey, snow layer structure analysis, unmanned aerial vehicle investigation, numerical simulation, and remote sensing interpretation, monitoring, and early warning for investigation and research [8,9]. As one of the earliest countries to conduct avalanche research, Switzerland continues to rely on field observations as a crucial method for acquiring avalanche data to this day [10]. Snowpack stratigraphic profiling forms the basis for snowpack stability assessment [11]. Herla et al. [12] have demonstrated that manual snowpack stratigraphic excavation, coupled with systematic measurement of density, mechanical strength, and thermal gradient parameters, enables the identification of critical weak-layer configurations (e.g., depth hoar layers, discontinuity interfaces), thereby facilitating quantitative stability evaluations of snow stratigraphy and probabilistic forecasting of avalanche initiation timing. Advancements in remote sensing have enabled novel methodologies for avalanche investigation. Denissova et al. [13] demonstrated effective avalanche monitoring through satellite remote sensing integrated with causal analysis, enhancing predictive capabilities and risk assessment. Acoustic signal monitoring has been operationally deployed in Norway, where avalanche detection systems successfully documented multiple events through acoustic signature analysis [14].

Numerical simulation has become an important method in avalanche research. The RAMMS::AVALANCHE model is now widely used in avalanche event reconstruction, risk assessment, and sensitivity analysis. Christen et al. [15,16] conducted rigorous validation of the entrainment module within the RAMMS::AVALANCHE framework through back-calculation procedures of historical avalanche events, demonstrating the model’s fidelity in reconstructing mass flow dynamics during avalanche back-calculation processes. Teich et al. [17] conducted successful implementation of the RAMMSS::AVALANCHE modeling framework to simulate low-magnitude avalanche events within forested terrain, establishing novel implementation protocols that extend the model’s operational capacity in arboreal environments; Kazemzadeh et al. [18] leveraged the RAMMS::AVALANCHE numerical platform through the integration of meteorological drivers and topographic determinants, conducting systematic identification of dominant controlling factors in avalanche initiation mechanisms and generated key parameters related to avalanche dynamics; Liu et al. [19] leveraged the RAMMS::AVALANCHE modeling system to reconstruct historical avalanche events in the Tianshan Mountains region, which not only demonstrated the model’s applicability in this area but also optimized the input parameters based on local snow characteristics. Through these studies, the practical applications of RAMMS::AVALANCHE in avalanche reconstruction, causality analysis, and risk assessment have been progressively refined.

Avalanches, as a complex natural disaster with multiple contributing factors, share similarities with other natural disasters in that the disaster-prone environment has a significant impact on the occurrence of the event. Atwater et al. [20] conducted seminal work in avalanche hazard genesis identification and identified ten major influencing factors. In recent years, with the gradual advancement of avalanche research, the number of influencing factors considered in avalanche analysis has expanded and fluctuated, ranging from 8 to 14 factors [21,22]. Current studies generally classify avalanche causes into three categories: terrain and geomorphology, climate and weather, and snow accumulation characteristics [23]. Advancements in analytical methodologies and avalanche research have shifted hazard assessment toward explicating qualitative relationships among contributing factors. Uroš Durlević et al. [24] employed the Analytic Hierarchy Process (AHP) for avalanche susceptibility zoning. Satish et al. [25] validated the applicability of the fuzzy–frequency ratio model through avalanche analysis in the Himalayan region. Building upon field investigations, Liu et al. [26] integrated the Geodetector model with the Certainty Factor (CF) approach to produce susceptibility maps for this study area, where the CF model achieved an AUC of 0.836—demonstrating robust predictive capability. Comparative analysis confirms the CF model’s computational efficiency and high reliability. Consequently, this study incorporates in situ field testing and statistical analysis with updated avalanche inventories, applying the CF model for causal analysis along this transport corridor.

The G219 highway section between Wenquan County (Bortala Mongolian Autonomous Prefecture) and Khorgas City (Ili Kazakh Autonomous Prefecture) in Xinjiang, China, is situated in a region characterized by heavy winter snowfall concentration, drastic diurnal temperature fluctuations, and frequent avalanche hazards in mountainous terrain [27]. Falling snow from mountain slopes often blocks roads and damages construction machinery, seriously affecting winter construction safety. To ensure safety during construction and after completion, avalanche monitoring, early warning, and prevention design are urgently needed in this area; however, there is limited data on the causes, distribution, and hazard potential of avalanche disasters in this area. Therefore, a detailed investigation of avalanche distribution, along with an analysis of disaster characteristics and causality, holds significant engineering and theoretical value.

This study conducts a detailed field investigation of the project area, analyzing the distribution and types of avalanches along the route, and provides statistical data on avalanche disasters within the region. Building upon the field investigation, this study utilizes meteorological data, video surveillance footage, drone-based aerial imagery, and in situ field experiments to analyze the characteristics and causative factors of avalanche disasters in the study area. Utilizing the RAMMS::AVALANCHE numerical platform, avalanche disaster reconstruction was conducted along the entire route, yielding key parameters such as maximum flow velocity, flow pressure, flow height, and deposition height. The results serve to support avalanche disaster research and the development of early warning models and provide essential parameters for the design of avalanche mitigation engineering.

2. Overview of the Study Area

2.1. Geographic Overview

The study area (80°39′–80°41′ E, 44°27′–44°38′ N) is situated within the Ili Kazakh Autonomous Prefecture of Xinjiang Uygur Autonomous Region, China, bounded by Khorgas City to the north, extending southward to the Qiet’akesu Gully mouth, and adjacent to Sayram Lake to the east (Figure 1). The G219 National Highway’s Wenquan–Khorgas section traverses the study area along the topographic axis of the Qiet’akesu Gully, exhibiting a distinct meridional orientation (N–S) with elevation gradients descending from northern highlands to southern lowlands, encompassing a longitudinal span of 15 km within the research area. The study area features a temperate continental climate characterized by cold winters, hot summers, short spring and autumn seasons, significant annual temperature variation, concentrated seasonal precipitation, and low annual rainfall. However, the Ili region (Ili Kazakh Autonomous Prefecture, Xinjiang Uygur Autonomous Region) exhibits unique microclimatic conditions in the Sayram Lake basin, where abundant precipitation and elevated humidity levels manifest as a combined effect of Atlantic westerlies transport and orographic forcing from the Tianshan Mountains’ geomorphic configuration [28].

2.2. Overview of the Hazard-Prone Environment

2.2.1. Topography and Geomorphology

The study area is predominantly characterized by mountainous terrain and river valleys. The hillslopes within the region exhibit significant surface erosion, with steep gullies and ravines distributed throughout. Perennial snowmelt from surrounding mountains feeds the Qiet’akesu River, resulting in swift water flow and deeply incised “V”-shaped valley formations. The adjacent mountain slopes rise sharply on both sides of the valley, with relatively steep inclinations, further enhancing the area’s susceptibility to geomorphological hazards such as avalanches. Avalanche gullies and debris flow channels are distributed along both sides of the Qiet’akesu Valley within the study area. These gullies exhibit sparse vegetation coverage, with the surface primarily composed of loose gravel and weathered soil, which serve as the dominant underlying materials. Steep, exposed mountain slopes provide ideal terrain conditions for avalanche development. Empirical investigations by Zhong et al. [29] reveal that snowpack accumulation in non-vegetated or sparsely vegetated zones exhibits up to 53% greater depth compared to forested areas. In the study area, the combination of exposed slopes and abundant snowfall provides ample material sources for the formation of avalanche release zones.

2.2.2. Meteorological and Climatic Conditions

Five six-element meteorological stations, numbered 1 through 5, are arranged sequentially from south to north within the study area (Figure 1b). Based on data obtained from the meteorological stations, the maximum snow depth during winter in the study area exceeds 150 cm, with an average snow cover duration of approximately 160 days (Figure 2). In comparison to other regions within Ili Prefecture, the project area demonstrates greater natural snow accumulation depth and prolonged average snow cover duration [30]. The study area, located in mountainous terrain, experiences significant diurnal temperature fluctuations. During early March, snowpack undergoes daytime melting due to temperature rise and solar radiation, followed by nocturnal refreezing. These repeated freeze-thaw cycles coupled with internal snowpack temperature gradients lead to reduced snowpack stability and increased avalanche risk. Concurrently, the region’s unique climatic conditions result in an annual average relative humidity of 66.89%. During the snow season from 1 November to 25 May of the following year, the average relative humidity measures 61.49%, with daily maximum relative humidity reaching 98.49% within this period (Figure 3). Atmospheric water vapor infiltrates slope snowpack, increasing internal liquid water content, which leads to elevated snow density, crystal regrowth, reduced snow layer strength, and consequently heightened avalanche risk.

2.2.3. Snowpack Characteristics

Xinjiang, categorized as a continental snow climate zone, exhibits ‘dry’ and ‘cold’ snowfall characteristics compared to maritime snow climate zones. The continental dry-cold snowpack leads to significant internal temperature gradients and thicker depth hoar layer development [31]. During the early snow season, the low liquid water content within the snowpack results in a loose granular structure with reduced interlayer cohesion, leading to predominant occurrences of loose-snow avalanches. During the late snow season, snowpack ablation coincides with extensive development of depth hoar layers and diminished cohesion within the snowpack [32]. Such snowpack characteristics render avalanche risk more prone to occur in Xinjiang’s snow cover when environmental changes transpire. The Ili region exhibits heavier winter snowfall compared to other areas in Xinjiang, leading to frequent avalanche occurrences and larger-scale events within the study area.

3. Data and Methods

3.1. Data

3.1.1. Field Reconnaissance

Field reconnaissance constitutes one of the critical methodologies in mountain hazard investigation [33]. To obtain detailed and reliable avalanche data, this study conducted avalanche disaster investigations at multiple stationed sites starting from 2023. This study conducted an integrated field reconnaissance and UAV aerial surveys along the entire project corridor-enabled documentation of historical avalanche traces and residual avalanche deposits, leading to preliminary identification of 86 avalanches (

Table 1. Statistics of avalanche hazards in the study area.

Category	Count	Proportion
Chute-confined avalanches	23	26.74%
Unconfined slope avalanches	63	73.26%

). These were classified into 63 chute-confined avalanches and 23 unconfined slope avalanches based on flow morphology and substrate terrain characteristics. The study area’s surface vegetation is predominantly characterized by meadows and scrublands. Extensive snow accumulation platforms above the chutes provide favorable snow storage capacity, while the terrain exhibits wide and steep gullies with exposed bedrock features.

3.1.2. Meteorological Data and Video Surveillance Footage

Meteorological data is extensively utilized in avalanche hazard analysis [34], with five six-parameter meteorological stations deployed in the study area recording data at 10 min intervals. Analysis of meteorological parameter fluctuations and snow depth variations pre- and post-avalanche events enables reconstruction of the avalanche initiation process and identification of causative factors in avalanche hazard analysis. Video surveillance cameras serve as a critical methodology for avalanche data acquisition [35], with 15 units deployed across the study area (Figure 1b). Each video monitoring station conducts pre-programmed cruise monitoring of multiple risk-prone locations at 30 or 60 min intervals. Application of video surveillance footage enables precise determination of the temporal occurrence and magnitude of avalanche events while providing supplementary validation and refinement of field reconnaissance findings. Key technical parameters and specifications of the monitoring stations are detailed in

Table 2. Key technical parameters and specifications of meteorological and video monitoring stations.

Monitoring Parameters	Instrument/Sensor Model	Main Technical Indicators	Data Intervals	Equipment Manufacturer	Production Address
Video surveillance cameras	DS-2SK8C144MH-D/SP/GLT/DG	Focal length range: Panoramic 4 mm; detail 6.0~150 mm	30~60 min	Hangzhou Hikvision Digital Technology Co., Ltd.	Hangzhou, China
Temperature and humidity	HMP155 Air temperature and humidity sensor	Range: 0~100% (RH); −80~+60 °C	10 min	Beijing Truwel Instruments Inc.	Xi’an, China
Air pressure	PTB110 Air pressure sensor	Range: 500~1100 hPa; Accuracy: ±1.0 hPa	10 min	Beijing Truwel Instruments Inc.	Xi’an, China
Wind speed	W10 Wind speed sensor	Range: 0~60 m/s; Starting wind speed: 0.17 m/s	10 min	Beijing Truwel Instruments Inc.	Xi’an, China
Wind direction	W20 Wind direction sensor	Range: 0~360°; Accuracy: 0.088°	10 min	Beijing Truwel Instruments Inc.	Xi’an, China
Snow depth	SnowVUE10 Digital snow depth sensor	Accuracy: ±1 cm; Resolution: 0.25 mm	10 min	Beijing Truwel Instruments Inc.	Xi’an, China

.

3.2. Methods

3.2.1. In Situ Field Testing

In situ snowpack field testing provides critical data on snow layer physical properties, offering fundamental insights into avalanche initiation processes, causative mechanisms, and hazard recurrence in avalanche hazard analysis [36]. This study conducted in situ field testing from 1 March 2024 to 15 April and from 1 November 2024 to 3 March 2025 at 11:00 a.m. daily. The snow cover is measured layer by layer every 20 cm, and the weak layer is measured separately. The data obtained from the weak layer is used as the snow cover characteristics of the avalanche instability interface. The snow cover is measured layer by layer every 20 cm, and the weak layers are measured separately. The data obtained from the weak layers are used as the snow cover characteristics of the avalanche instability interface. Natural snowpacks in wind-sheltered shaded areas of the study area were selected for excavation of stratigraphic profiles (Figure 4a). Measurement of physical property variations across snow layers enables rapid identification of internal weak layers (Weak Layers), thereby facilitating determination of avalanche initiation depth. Through implementation of direct shear tests (Figure 4b), snow density measurements using cutting ring samplers (Figure 4c), liquid water content determination (Figure 4d), temperature gradient profiling (Figure 4e), extended column tests (Figure 4f), and crystal morphology imaging (Figure 4g), the snowpack parameters listed in

Table 3. On site in situ test table.

Pilot Projects	Direct Shear Tests	Cutting Ring Sampler Density Measurements	Liquid Water Content Determination	Temperature Gradient Profiling	Crystal Morphology Photography	Extended Column Tests
Parameter	Shearing strength; Cohesion; Internal friction angle	Density	Moisture content	Snow temperature	Crystal morphology	Weak layer position

were obtained (Table 3).

3.2.2. RAMMS::AVALANCHE Simulation

RAMMS::AVALANCHE is a professional numerical simulation software for avalanche dynamics, which simplifies the complex three-dimensional flow process of avalanches into two-dimensional problems through depth-averaged avalanche dynamics equations. The RAMMS::AVALANCHE model employs the Voellmy rheological framework to simulate avalanche dynamics. This formulation accounts for turbulent flow regimes during high-velocity motion—characterized by the turbulent friction coefficient ξ—and Coulomb frictional resistance, which dominates in low-velocity regimes and is proportional to the normal stress. Calibration of avalanche flow behavior is achieved by adjusting the turbulent friction coefficient ξ and the Coulomb friction coefficient μ, enabling accurate simulation of avalanche dynamics across a wide range of flow velocities.

In the RAMMS::AVALANCHE model, the frictional resistance during avalanche motion is expressed by the following equation:

$$S=\mu N+\left(1-\mu \right)Cexp\left(-{\displaystyle \frac{N}{C}}\right)+{\displaystyle \frac{\rho g{U}^2}{\xi }},$$

(1)

In the equation, S represents the resistive force acting during avalanche motion, which governs the flow behavior of the avalanche; μ denotes the Coulomb friction coefficient; N denotes the normal stress of the avalanche flow, which, in conjunction with the friction coefficient μ, characterizes the dominant dynamic frictional force during high-velocity avalanche motion; C represents the cohesive strength of the snowpack, which helps reduce numerical diffusion during simulation, thereby producing clearer and more accurate flow outlines; ρ denotes the average snow density within the avalanche release zone; g is the gravitational acceleration; U represents the avalanche flow velocity; and ξ is the viscous/turbulent friction coefficient.

The RAMMS::AVALANCHE model requires input data including topographic datasets, release zone parameters, frictional coefficients, forest cover indices, computational configurations, and orthophoto imagery. Critical parameters are listed as follows:

• Release Zone Parameters

The avalanche release zone refers to the area where avalanches initiate and detach, typically situated on steep mountain slopes. When the snowpack within the release zone reaches a metastable equilibrium state, even minor disturbances can trigger gravitational failure, initiating downward snowmass movement that subsequently entrains lower snowpack layers through momentum transfer mechanisms. Since the avalanche release zone provides the initial kinetic energy for the entire avalanche and its spatial extent governs the event magnitude, accurate delineation of the release zone boundaries is critical in avalanche simulation. Different avalanche types exhibit distinct release zone characteristics. In the study area, slab avalanche release zones are typically characterized by areal features (Figure 5a) with prominent pre-failure fracture lines. In contrast, loose-snow avalanche release zones often initiate from point triggers (Figure 5b), frequently observed beneath protruding bedrock outcrops or isolated trees within the release zones. As illustrated in Figure 5, this study predominantly utilizes UAV aerial imagery to delineate the upper boundary of release zones. Integrated with terrain analysis, the lower boundary is defined at slope gradient transition zones, with the total release area constrained to ≤1/3 of the watershed area to ensure geomorphological plausibility.

2.

Fracture depth

Fracture depth refers to the mean depth of displaced snowpack within the avalanche release zone during snow layer failure. The determination of fracture depth is inherently linked to topographic conditions and snowpack stratigraphic properties.

To simplify the analysis, this study reduces the problem of full-depth avalanche fracture depth to a snowpack slope stability problem. Under this assumption, the gravitational force acting on the snowpack can be decomposed into two components: the downslope component T and the normal component N, which are parallel and perpendicular to the slope surface, respectively. The expressions for T and N are given as follows:

$$\left\{\begin{array}{c}T=\rho hlsin\alpha \\ N=\rho hlcos\alpha \end{array}\right.,$$

(2)

In the equations, h denotes the snowpack thickness on the slope; l represents the slope length; and α signifies the slope angle.

Although snow mechanics constitutes a complex and unique rheological system, avalanche research primarily focuses on internal stresses governed by boundary conditions and material self-weight. Consequently, snowpack instability can be effectively characterized as a Mohr–Coulomb failure mechanism [37]. When the normal component T of gravitational force exceeds the combined resistance of snow cohesion, shear strength, and static friction, the snowpack begins to slide downslope. The resistance of snowpack to downslope gravitational forces is predominantly determined by its shear strength. Through fitting shear strength data obtained under varying normal stresses, the Mohr–Coulomb failure criterion can be derived, expressed as follows:

$$\tau =C+\sigma tan\varphi ,$$

(3)

In the equation, τ is the snowpack shear strength, representing the snow’s resistance to shear deformation under gravitational loading; σ is the snowpack normal stress, here expressed as N/l; and ϕ is the snow internal friction angle.

Therefore, the fracture depth h_k during full-depth avalanche snowpack failure can be derived as follows:

$$h_k=C/\left(\rho \right(sin\alpha -cos\alpha \ast tan\varphi \left)\right),$$

(4)

Within the study area, the average snow density, cohesion, and shear strength parameters were determined through in situ experimental investigations, while frictional coefficients and turbulent friction coefficients were referenced from snowpack characterization studies conducted by the Chinese Academy of Sciences (CAS) in the Tianshan Mountains region. All parameters are detailed in

Table 4. Physical properties of accumulated snow in the region.

Average Snow Density (g/cm3)	Cohesion (g/cm2)	Frictional Coefficients	Shear Strength (gf/cm2)	Turbulent Friction Coefficients (m2/s)
3.0~3.4	2.3~3.7	0.3~0.35	3.0~5.12	1000–3500

.

Weak layers within the snowpack typically form the failure plane for slab avalanches, thereby allowing the fracture depth of shallow avalanches to be conceptualized as the stratigraphic position of weak layers within the snow column [38]. Based on observational data collected by Hao Jiansheng et al. [39] in the Tianshan region, the weak layer within the snowpack exhibits the following functional relationship with snow depth:

$$h_W=237 ln h-1190,$$

(5)

In the equation, h_W denotes the weak layer height.

Therefore, the release depth h_k′ for shallow avalanches can be expressed as follows:

$${h_k}^{\prime }=h_W-h,$$

(6)

3.

Friction Coefficient

By integrating the topographic characteristics, release zone elevation, avalanche magnitude, and temporal recurrence intervals of the study area with the friction coefficient reference table (

Table 5. RAMMS::AVALANCHE Friction coefficient reference table.

Avalanche Magnitude	Topography	Elevation/m	Temporal Recurrence Intervals/Years	μ	ξ
Medium (25~60,000 m3)	chute-confined avalanches	>1500	10	0.35	1350
	Unconfined slope avalanches	>1500	10	0.2	3250

) provided in RAMMS::AVALANCHE, the following parameters are determined:

Since RAMMS::AVALANCHE mainly simulates dry avalanches, studies by Platzer et al. [40,41] show wet snow has greater cohesion and friction than dry snow. Wet-snow avalanches exhibit lower Coulomb friction coefficients (μ) than dry-snow counterparts during rapid flow phases. However, under ξ-controlled low-velocity motion (where ξ represents the turbulent friction coefficient), their turbulent effects intensify significantly. Formula 1 indicates that parameter calibration requires either reducing μ or increasing ξ. Ultimately, the adopted μ/ξ parameter combinations for chute-confined avalanches and unconfined slope avalanches in this study are 0.34/1350 and 0.25/3250, respectively.

3.2.3. Sensitivity Analysis of Influencing Factors

The study employed a Certainty Factor Model (CFM) coupled with Sensitivity Indices (Si) to conduct categorical and hierarchical classification analysis of four critical avalanche-influencing factors within the research area: elevation, slope gradient, slope aspect, and maximum snow depth.

(1)

Hierarchical Classification of Influencing Factors

Through field reconnaissance and manual interpretation of imagery data, four critical influencing factors—elevation, slope gradient, slope aspect, and maximum snow depth—were selected from three primary avalanche genesis categories as the dominant controlling parameters for snow avalanche hazards in the study area. The original-resolution Digital Elevation Model (DEM) was resampled using a Geographic Information System (GIS) platform to optimize spatial analysis. In GIS platforms, pixel size significantly influences analytical outcomes, making appropriate pixel size selection a critical preliminary task in hazard risk assessment [42]. This study employs the following formula to determine optimal pixel resolution:

$$G_S=7.49+0.0006S-2.0\ast {10}^{-9} S^2+2.9\ast {10}^{-15} S^3,$$

(7)

In the equation, G_S denotes the pixel size of the resampled dataset (meters) and S represents the scale denominator of the original topographic data (dimensionless).

In this study, the original dataset processed within the GIS platform maintained a scale denominator of 1:35,000. Through application of Equation (7), the resampled pixel size was calculated as 21.05 m, which was subsequently rounded to the nearest integer value of 20 m to facilitate computational processing. The four selected influencing factors were subjected to hierarchical grading computation, generating corresponding raster layers. As shown in Figure 6, elevation was evenly divided into eight classes at 300-m intervals (Figure 6a); slope gradient was classified into nine categories at 10° intervals (Figure 6b); slope aspect was categorized into eight directional sectors—N, NE, E, SE, S, SW, W, and NW—at 45° intervals (Figure 6c), with ‘N’ corresponding to the range (337.5°, 22.5°]. Snow depth zoning was based on maximum seasonal snow depth data recorded by snow depth sensors at five meteorological stations. Using the Inverse Distance Weighting (IDW) interpolation method in a GIS platform, a continuous snow depth distribution was generated for the entire study area (Figure 6d), and the region was further divided into ten zones at 10 cm intervals.

(2)

Certainty Factor Model

The Certainty Factor Model in natural hazard assessment quantifies the contributing influence of environmental factors on hazard occurrence, integrating expert knowledge and observational data to achieve dynamic risk evaluation. The core of this model lies in transforming uncertainty into computable metrics—Certainty Factor (CF) values. The relationship between hazard-inducing factors and avalanche occurrence is quantified through these numerical credibility measures, reflecting each factor’s support degree for avalanche susceptibility. The mathematical expression is defined as follows:

$$C{F}_a=\left\{\begin{array}{c}{\displaystyle \frac{P{P}_a-P{P}_s}{P{P}_a\left(1-P{P}_s\right)}},P{P}_a\ge P{P}_s\\ {\displaystyle \frac{P{P}_a-P{P}_s}{P{P}_s\left(1-P{P}_a\right)}},P{P}_a<P{P}_s\end{array}\right.,$$

(8)

In this equation, CF_a represents the certainty factor of avalanche occurrence under the influence of factor a. PP_a denotes the conditional probability of the event occurring within the classification of the input factor, expressed as the spatial density of avalanches in that classification—that is, the number of avalanches per unit area. PP_s is the prior probability of the event, represented by the overall avalanche density across the entire study area. According to Equation (8), the value of CF ranges from −1 to 1 and can be interpreted as the relative likelihood of avalanche occurrence under the given factor. A CF value approaching 1 indicates a strong positive correlation between the factor and avalanche occurrence, suggesting a high level of certainty. A CF value near 0 implies that the conditional probability closely matches the prior probability, indicating that the factor provides little to no predictive power regarding avalanche occurrence. Conversely, a CF value approaching −1 signifies a strong negative correlation, reflecting a low level of certainty that an avalanche will occur under the influence of that factor.

(3)

Sensitivity Index

The Sensitivity Index (S_a) is widely applied in the fields of geological hazard assessment, environmental science, and engineering risk analysis. This index can be used to quantify the degree to which influencing factors affect avalanche occurrence. The formula is expressed as follows:

$$S_a={CF}_{a,max}-{CF}_{a,min}$$

(9)

In the equation, S_a represents the Sensitivity Index of influencing factor a with respect to avalanche occurrence; CF_a,max denotes the maximum Certainty Factor value among the classified intervals of factor a; CF_a,min represents the minimum Certainty Factor value within the same classification.

3.2.4. Drawing Software and Workflow

The series of thematic maps used in this paper are mainly produced by ArcMap 10.8 software in the ArcGIS platform. Among them, by importing tilt photography and digital elevation maps, the maps are obtained by exporting the simulation results of RAMMS::Avalanche to an asc file and superimposing them on DEM.

All procedures and approaches used for the purpose of this research are presented in the flowchart given in Figure 7.

4. Results

4.1. Hazard Characteristics

4.1.1. Type and Quantitative Characteristics

Through the integration of field reconnaissance data, video surveillance footage, UAV aerial photographs, and analysis of avalanche deposit residues, 86 avalanche initiation zones were identified and rectified (Figure 8a), including 23 unconfined slope avalanches and 63 chute-confined avalanches. Through manual interpretation of video surveillance data from the project area, where each slope snowpack release event was recorded as a single avalanche occurrence, a total of 202 avalanche events were documented. Using a 3% snow liquid water content threshold (LWC ≥ 3% for wet snow), the avalanches were classified into dry snow avalanches (28 events, 13.86% of total) and wet snow avalanches (174 events, 86.14% of total). This indicates that chute-confined avalanches and wet snow avalanches are the predominant avalanche types within the study area. The study area contains 86 documented avalanches, yet manual interpretation revealed 145 distinct avalanche release zones (Figure 8a). This discrepancy primarily stems from chute-confined avalanches frequently exhibiting two or more release zones (Figure 8b). Geomorphological analysis indicates that the valley’s flanking slopes predominantly feature Y-shaped gullies. The upper sections of these gullies contain expansive snow accumulation basins that form multiple release zones, while downstream topographic constrictions cause gully convergence, ultimately coalescing into unified avalanche tracks and deposition zones.

4.1.2. Magnitude Characteristics

In the study area, numerous avalanche events are distributed across varying terrain, with differences in contributing catchment area, fracture depth, and flow path, making it challenging to directly assess avalanche magnitude. Given that avalanche deposits persist over time and can visually represent avalanche scale, this study adopts the European Avalanche Warning Services (EAWS) classification standard [43]. Based on deposit volume as the primary criterion, avalanche magnitude is categorized into five levels, as detailed in

Table 6. Classification of avalanche magnitude.

Avalanche Magnitude	SLUFF	Medium	Large	Very	Extremely
Volume (m3)	1~102	102~103	103~104	104~105	>105
Potential damage	Unlikely to bury a person, except in unfavorable runout zones. In extreme terrain there is a danger of falling.	Can bury, injure, or kill people. Many avalanches that kill people are classified as ‘medium’.	Can bury and destroy cars, damage trucks, destroy small buildings and break a few trees. Many avalanches that kill people are classified as ‘large’.	Can bury and destroy trucks and trains Can destroy fairly large buildings and small areas of forest. Very large avalanches can occur at danger level 3 and are typical of danger levels 4 and 5.	Can devastate the landscape and has catastrophic destructive potential. Typical for danger level 5

.

According to the classification criteria outlined in the table above, the avalanche classification results within the study area are illustrated in the following figure (Figure 9):

As shown in the figure, avalanche disasters within the study area are predominantly of medium and large scales, accounting for 45 and 29 occurrences, respectively. The maximum avalanche deposit volume reaches 24,546.19 m³, while the minimum is 50.275 m³. These events pose significant potential threats to road traffic safety and human life within the region.

4.2. Spatiotemporal Distribution Characteristics

4.2.1. Time Distribution Characteristics

Avalanche events within the study area were categorized and statistically analyzed on a monthly basis, with the results summarized in

Table 7. Avalanche events monthly distribution table.

Month		November	December	January	February	March	April
Count	Dry snow avalanche	0	4	9	10	3	2
	Wet snow avalanche	1	11	9	55	82	16
	Total	1	15	18	65	85	18

. The findings indicate that the peak period for avalanche occurrences is in February and March. During this period, a total of 150 avalanche events were recorded, accounting for approximately 74% of all avalanche incidents throughout the snow season. When classified by hour, as shown in

Table 8. Avalanche events hourly distribution table.

Time		10	11	12	13	14	15	16	17	18	19	20	21	22
Count	dry snow avalanche	1	2	18	15	14	21	26	33	21	12	7	3	1
	wet snow avalanche	2	2	2	3	2	3	3	2	2	2	1	2	2
	total	3	4	20	18	16	24	29	35	23	14	8	5	3

, avalanche events predominantly occurred between 15:00 and 18:00, with 111 incidents reported, representing 54.95% of the total. McClung and Schaerer [44] proposed that the strength of the snowpack is influenced by its water content; as the water content within the snowpack increases, its overall strength correspondingly decreases. As shown in Figure 10, air temperatures in the study area begin to rise in February. Although the region is located in the UTC+6 time zone, it follows UTC+8 for timekeeping; therefore, the actual avalanche occurrences correspond to local afternoon hours between 13:00 and 16:00. This indicates that February to March and the hours of 13:00 to 16:00 represent the peak periods for avalanche activity. The increase in temperature leads to a rise in snowpack water content, resulting in decreased snow strength and reduced stability. Consequently, this thermal effect induces a pronounced temporal clustering of avalanche events.

4.2.2. Spatial Distribution Characteristics

Using the GIS platform, the elevation distribution of avalanche release areas was statistically analyzed, resulting in a histogram of avalanche release zone frequency across elevation bands (Figure 11). Avalanches within the study area are distributed across mountainous terrain ranging from 1500 to 3600 m in elevation, with the majority concentrated between 2100 and 3000 m, accounting for 74.48% of all recorded events. The maximum elevation of avalanche release areas is 3505 m, while the minimum is 1714 m. By integrating the elevation zoning map and the snow depth distribution map (Figure 6a,d), it can be observed that snow accumulation is significantly greater within the 1800–3000 m elevation range, providing the material basis necessary for avalanche initiation. This indicates a strong spatial correlation between avalanche occurrences and the distribution of maximum snow depth.

4.3. Numerical Simulation Characteristics

Similarity analysis between simulation results and actual avalanche events quantitatively assesses simulation accuracy [8]. This evaluation integrates shape ratio, size, and directional similarities to determine overall congruence. Results demonstrate >80% similarity (S) between numerical simulations and field investigations, meeting similarity criteria. Therefore, RAMMS::AVALANCHE exhibits high reliability for avalanche numerical modeling. The RAMMS::AVALANCHE model was employed to simulate avalanche activity across the entire study area. Then, to facilitate a detailed examination of flow dynamics, one representative chute-confined avalanche and one unconfined slope avalanche were randomly selected for individual analysis. These two cases were used to characterize key simulation features, including maximum flow velocity, runout distance, flow depth, and impact pressure, thereby providing insights into the distinctive behaviors of different avalanche types under local topographic and snowpack conditions.

4.3.1. Simulated Avalanche Characteristics Across the Entire Region

According to simulation results, avalanche release zones are predominantly located on relatively gentle snow accumulation platforms near the upper slopes. The maximum flow height typically occurs at the front edge of the deposition zone (Figure 12a), where the terrain gradient decreases, resulting in reduced flow velocity and flow pressure, causing the snow mass to accumulate maximally. The highest recorded flow height across the study area reaches 15.43 m. The maximum flow velocity generally occurs in the lower to middle sections of the flow path (Figure 12b), with a maximum velocity of 47.6 m/s observed within the study region. Maximum flow pressure typically coincides with the point of peak flow velocity (Figure 12c), with the highest value reaching 679.79 kPa. Simulation results indicate that most avalanches deposit along the valley floor, with some residual snow remaining at the lower end of the flow paths (Figure 12d). Depositional bodies tend to spread along the river valley, often connecting with adjacent avalanche deposits, leading to extensive blockage of the valley. The maximum deposition height recorded is 10.3 m.

4.3.2. Simulated Characteristics of Unconfined Slope Avalanche

The specific input parameters for the selected unconfined slope avalanches are presented in

Table 9. Input parameters of unconfined slope avalanche.

Avalanche Type	Fracture Depth (cm)	Average Density (g/cm3)	μ	ξ	Volume (m3)
Full-depth wet snow avalanche	45	3.2	3250	0.20	205.646

.

Based on the simulation results, unconfined slope avalanches are less influenced by terrain constraints during their motion. The affected area shows no significant signs of channeling or narrowing (Figure 13). The broad affected area enables the avalanche to entrain additional unstable snow from the slope surface, progressively increasing in width after release from the release zone. As the avalanche enters the deposition zone, changes in terrain induce lateral spreading of the flow, accompanied by a reduction in velocity and an expansion in flow width. Flow pressure and velocity exhibit similar spatial distributions and characteristics (Figure 13c). After motion ceases, aside from the snow deposited within the deposition zone, relatively little snow remains in the affected area. Most of the entrained snow is transported from the slope surface to the deposition zone due to entrainment effects (Figure 13d).

4.3.3. Simulated Characteristics of Chute-Confined Avalanche

The specific input parameters for the selected chute-confined avalanches are presented in

Table 10. Input parameters of chute-confined avalanches.

Avalanche Type	Fracture Depth (cm)	Average Density (g/cm3)	μ	ξ	Volume (m3)
Full-depth wet snow avalanche	76	3.2	1350	0.35	205.646

.

The avalanche is released from the snow accumulation zone located above the chute, and, influenced by the topography, the avalanche flow becomes confined to the base of the gully. Unlike unconfined slope avalanches, there is no apparent lateral expansion of the flow. Once the avalanche enters the chute, the terrain forces the moving snow mass to converge toward the center of the path. Compared to slope-type avalanches, the snow in the chute-confined-type event exhibits a more pronounced concentration toward the flowline center (Figure 14). Although the selected chute-confined avalanche is slightly larger in overall scale than the unconfined slope case, the increased friction and collision between snow particles, as well as between the snow and ground surface, result in lower maximum flow velocity and flow pressure (Figure 15b,c). Additionally, while most of the snow is deposited in the terminal deposition zone, a small amount of residual snow remains along the base of the chute path (Figure 15d).

4.4. Analysis of Avalanche Causes and Influencing Factors

4.4.1. Elevation

According to the study by Marcia et al. [45], approximately 68% of avalanche events occur at elevations between 2000 and 3000 m. This pattern is closely related to snowpack stability and the distribution of permafrost. Huijun et al. [46] reported that the lower limit of permafrost on the sunny slopes of the central Tianshan Mountains is around 3250 m above sea level. Above this elevation, ground temperatures remain below 0 °C year-round, which promotes the long-term stability of snowpacks on mountain slopes. Consequently, most avalanche occurrences in the study area are concentrated below the permafrost boundary. In regions above the permafrost limit, the snow cover tends to persist year-round with minimal internal temperature gradients, resulting in a more stable snow structure and reduced avalanche susceptibility. In contrast, at elevations below 1800 m, the vegetation cover increases significantly, leading to a reduction in both favorable terrain and snow accumulation necessary for avalanche initiation, thereby substantially lowering the likelihood of avalanche occurrence.

As shown in Figure 16, the Certainty Factor (CF) values for elevation are greater than 0 within the altitude range of 1800–3000 m, indicating a high susceptibility to avalanche occurrence in this elevation band, which accounts for 55.86% of the total area and is thus classified as an avalanche-prone zone. The highest CF values are observed between 2100 m and 2400 m (CF = 0.476), identifying this range as the most avalanche-susceptible zone. In contrast, elevations below 1800 m and above 3000 m exhibit CF values less than 0, indicating low avalanche susceptibility. Notably, elevations above 3600 m exhibit the lowest CF value (CF = −1), representing areas with minimal avalanche risk. According to Equation (9), the sensitivity index of elevation is S_a = 1.476, confirming that elevation is one of the dominant controlling factors influencing avalanche occurrence.

4.4.2. Slope Gradient

Topography is the only factor that remains constant throughout the avalanche process, making it critically important for understanding avalanche formation. Bussion et al. [47] emphasized the strong relationship between slope angle and snowpack stability, noting that the likelihood of avalanches increases with slope steepness within a certain range. The lower threshold of this range is generally considered to be a slope angle of 30° [48]; below this threshold, snow is less prone to instability, the critical failure depth increases, and avalanches are unlikely to occur. According to data published by the USDA Forest Service, the upper threshold is around 60°; beyond this angle, snow has difficulty adhering to the slope and typically slides off in the form of sluffs rather than full-scale avalanches.

As shown in Figure 17, the Certainty Factor (CF) values for slope angle are greater than 0 within the range of 30° to 50°, indicating that avalanches are more likely to occur in this slope range, which can be classified as an avalanche-prone zone. This zone accounts for 60.52% of the total study area. The highest CF values are observed between 40° and 50° (CF = 0.362), identifying this interval as the area of highest avalanche risk. In contrast, slopes below 30° and above 50° exhibit CF values less than 0, designating them as avalanche-unlikely zones. Particularly, slopes below 10° and above 60° show the lowest CF values (CF = −1), indicating regions where avalanche occurrence is least likely. According to Equation (8), the sensitivity index for slope angle is S_a = 1.362, confirming that slope angle is one of the primary dominant factors influencing avalanche occurrence in the study area.

4.4.3. Slope Aspect

Slope aspect governs snowpack metamorphism processes, while snowpack metamorphism constitutes a critical hypothesis within avalanche genesis mechanisms [49]. Due to the presence of temperature gradients within natural snowpacks, moisture migrates internally, leading to continued development of snow crystals at the snow-ground interface. This process causes changes in the shape, size, and mass of the snow grains as they adjust to evolving environmental conditions. Colbeck [50] suggested that solar radiation absorption results in the melting of snow crystals, forming a cohesionless surface layer that reduces snowpack stability. In the study area, significant diurnal temperature variation exists—temperatures rise during the day due to solar radiation and drop at night as heat is released. However, since snow is a poor conductor of heat, temperature increases are generally confined to the snow surface layer during daytime, resulting in steep internal temperature gradients, which can reach up to 0.5 °C/cm. According to Wang Yanlong et al. [51], the critical temperature gradient for snow metamorphism in Xinjiang is 0.2 °C/cm; thus, the occurrence of high gradients in this region is highly conducive to the development of weak layers. Since aspect determines the amount of solar radiation a slope receives, it significantly influences avalanche susceptibility through its effect on snow metamorphism.

As shown in Figure 18, the CF values for aspect zones are greater than zero for NE-, E-, and W-facing slopes, indicating that avalanches are more likely to occur in these orientations. These zones account for 35.89% of the total study area and are classified as avalanche-prone. Among them, the W-facing slopes exhibit the highest CF value (CF = 0.46), suggesting the greatest avalanche risk and thus constituting the most susceptible aspect zone. In contrast, slopes facing N, SE, S, SW, and NW show negative CF values, indicating lower avalanche susceptibility. The S-facing slopes display the lowest CF value (CF = −1), suggesting minimal avalanche hazard in these areas. According to Equation (9), the sensitivity index for aspect, Sa = 0.966, indicates that aspect is not a dominant factor influencing avalanche occurrence in this region. Notably, while avalanches predominantly occur on sun-facing slopes in general scenarios, the specific terrain configuration of north–south trending valleys in this study area constrains avalanche release zones to predominantly east–west orientations.

4.4.4. Maximum Snow Depth

As snow depth increases, the load exerted by the snowpack also rises. When the snow depth reaches a critical threshold, the gravitational component acting downslope induces shear failure and instability within the snow layers, ultimately leading to avalanche initiation. Therefore, snow depth serves as a key parameter in assessing snowpack stability. When the overlying snow depth above a weak layer becomes sufficiently large, the shear strength of the weak layer may no longer be able to support the weight of the snowpack, resulting in an avalanche that propagates along the weak layer. As snow depth increases, the probability of avalanche occurrence in the region also rises, indicating a positive correlation between snow depth and avalanche frequency.

As shown in Figure 19, when snow depth exceeds 65 cm within the study area, the Certainty Factor (CF) becomes greater than zero and increases with increasing depth. This indicates that snow avalanches are more likely to occur in zones with snow depths greater than 65 cm, which are therefore classified as avalanche-prone areas, covering 99.72% of the total area. Among these, the zone with snow depths between 145 cm and 155 cm exhibits the highest CF value (CF = 0.763), representing the most avalanche-prone region. Conversely, the 55–65 cm depth zone has the lowest CF value (CF = −0.773), indicating a low likelihood of avalanche occurrence. According to Equation (9), the sensitivity index of snow depth is Sa = 1.536, confirming that maximum snow depth is one of the primary controlling factors influencing avalanche occurrence in this region.

5. Discussion

In this study, we characterized avalanche hazards in the research area through integrated methodologies—including field reconnaissance, UAV aerial photography, video imagery analysis, in situ field testing, RAMMS::AVALANCHE numerical simulations, the Certainty Factor (CF) model, and Sensitivity Index (Sa) analysis. This comprehensive approach revealed avalanche typology and event frequency, spatiotemporal distribution patterns, dynamic flow characteristics, dominant controlling factors, and individual factor influence magnitudes.

Compared with Denissova et al.’s research [13], our study demonstrates significant agreement regarding avalanche causality: elevation, terrain configuration, and snowfall collectively exert substantial influence on avalanche initiation. McClung and Schaerer [44] postulated critical moisture content thresholds as determinants of avalanche probability. Increased snowpack moisture during diurnal warming cycles reduces structural stability, elevating failure likelihood. This mechanism explains the pronounced temporal clustering of avalanches during February–March warming periods observed in our study. Prior spatial distribution analyses established distinct upper and lower avalanche boundaries, which we precisely quantified through rigorous analysis of snow distribution patterns and physical properties within the study area.

While the integrated approach combining field surveys, simulations, and causal analysis successfully reveals regional avalanche characteristics and triggers, methodological limitations persist. Primarily, data acquisition is constrained by terrain accessibility, resulting in sampling bias toward road-proximal slopes and underrepresentation of remote/inaccessible zones. Future research should supplement methodologies with multi-platform terrain-independent data—such as satellite remote sensing and UAV aerial photography—to establish comprehensive avalanche inventories and achieve complete spatial documentation of release locations.

Through investigation and causal analysis of avalanche hazards in the study area, avalanche-prone zones and high-frequency periods were identified. Subsequent mitigation efforts can leverage numerical simulation results to implement targeted prevention for discrete avalanches. For small-to-medium avalanches amenable to engineering solutions, combining biological and structural countermeasures reduces hazards to roads and vehicles. For large avalanches resistant to engineering controls, continuous monitoring of elevation, slope angle, aspect, and maximum snow depth—based on identified prone zones and dominant factors—determines critical instability thresholds to enable early avalanche warnings.

6. Conclusions

This study investigates avalanche hazard characteristics and dominant controlling factors along the Wenquan–Khorgas section of China’s National Highway G219. Methodologically, we employed established techniques—field reconnaissance, UAV aerial photography, video imagery analysis, and in situ field testing—to collect avalanche data. Subsequent RAMMS::AVALANCHE numerical simulations analyzed avalanche dynamics. Finally, integrated application of the Certainty Factor (CF) model, Sensitivity Index (Sa), and avalanche causality hypotheses enabled quantitative assessment of multiple contributing factors, yielding the following conclusions:

• Avalanche events exhibit distinct spatiotemporal distribution patterns. Temporally, these events predominantly cluster in February and March, with peak hourly frequencies occurring between 13:00 and 16:00 local time. Spatially, the avalanches are distributed across elevation zones ranging from 1800 to 3300 m above sea level, showing concentrated occurrence between 2100 and 3000 m. The research results provide data support for the key prevention and monitoring of avalanche disasters in the study area.
• Avalanches were classified into chute-confined avalanche and unconfined slope avalanche categories based on movement morphology and underlying terrain characteristics, with 63 chute-confined avalanches (73.26% of total) identified. Using a 3% snow water content threshold, events were categorized as dry snow (14.86%) or wet snow avalanches (174 occurrences, 86.14%). According to the EAWS classification system, medium- and large-scale avalanches predominated in the study area, posing significant hazards to road construction and vehicular safety. The final classification result can provide an important reference for the prevention and control of engineering structures.
• Parameters obtained from field experiments were input into the RAMMS::AVALANCHE model to simulate avalanche dynamics along the entire route, yielding maximum flow heights of 15.43 m, maximum flow velocities of 47.6 m/s, maximum flow pressures of 679.79 kPa, and maximum deposition heights of 10.3 m. Comparative analysis of unconfined slope avalanches versus chute-confined avalanche simulations revealed that although unconfined slope avalanches exhibit smaller volumes, they demonstrate elevated flow velocities and intensified pressure dynamics. This hydrodynamic behavior suggests slope-type avalanches pose heightened hazards in open-slope terrain configurations due to their capacity for rapid momentum transfer and concentrated energy release. Based on the results of dynamic simulation, the minimum reference value for the structural strength of engineering structures has been proposed for prevention and control.
• Within the analyzed influence factor zones, elevation ranges of 1800–3000 m, slope angles of 30–50°, aspects oriented NE, E, and W, and maximum snow depths ≥65 cm demonstrated CF > 0, indicating these thresholds define avalanche-prone zones in the study area. Sensitivity index (Sa) analysis identified elevation, slope angle, and maximum snow depth (Sa > 1) as dominant controlling factors for avalanche initiation. The four factors were ranked in descending order of predictive significance: maximum snow depth > elevation > slope gradient > aspect.

In summary, this study provides a scientific research foundation for exploring the characteristics, causes, and post-disaster risk mitigation of avalanche disasters in the Wenquan–Khorgas section of National Highway G219 as the research area.