Cultural Heritage Risk Assessment Based on Explainable Machine Learning Models: A Case Study of the Ancient Tea Horse Road in China

Abstract

As the core carrier of historical and cultural identity, cultural heritage is facing multiple threats such as natural disasters, human activities and its own vulnerability. There is an increasing number of studies on cultural heritage risk assessment around the world, but the risk assessment of cultural heritage in China has not been fully explored. In this paper, the LightGBM model was used to quantitatively analyze the main influencing factors of cultural heritage risk along the Ancient Tea Horse Road in Sichuan, and spatial analysis was carried out by combining geographic information system (GIS) technology. In order to improve the interpretability of the assessment results, the SHAP method was introduced to systematically evaluate the contribution of each influencing factor to the risk of cultural heritage. This study identified seven major risk factors, including landslides, collapses, debris flows, earthquakes, soil erosion, urban road networks, and cultural heritage vulnerability, and constructed a risk assessment framework that comprehensively considers the vulnerability to natural and synthetic factors and the heritage itself. The results of the assessment divided the risk of cultural heritage sites into five levels: very low, low, medium, high and very high, and the results showed that 52.36% of the cultural heritage was classified as at medium and high risk and above, revealing the severe security situation faced by cultural heritage in the region and indicating the urgent need to take effective protective and management measures to deal with multiple risks and challenges.

1. Introduction

Cultural heritage is not only an important part of human civilization, but also carries profound historical significance and cultural value, which is related to the historical memory and identity of the nation. However, with the acceleration of modernization, cultural heritage is facing multiple threats from nature, society and the environment. Natural disasters such as earthquakes, floods, fires, etc. [1,2,3,4,5,6,7,8], anthropogenic factors such as urbanization, overtourism, etc. [9], and environmental changes, especially climate change and pollution [10,11], pose serious challenges. Globally, the international community has gradually recognized the urgency of this problem and has begun to take proactive measures, especially in reducing the risk of natural disasters and anthropogenic damage. In this context, the risk assessment of cultural heritage has become particularly important. By systematically analyzing the potential threats to cultural heritage from climate change, natural disasters and anthropogenic factors, risk assessment can provide a scientific basis for the development of conservation measures. This assessment helps to identify and quantify key risks and supports decision-makers in rationalizing the allocation of resources and focusing conservation priorities on the most vulnerable cultural heritage [12,13]. Therefore, effectively assessing the vulnerability of cultural heritage and formulating scientific and reasonable protection strategies have become the core topics of current research on cultural heritage protection.

In recent years, scholars have proposed a variety of assessment frameworks, covering vulnerability assessment, the impact of natural disasters on cultural heritage, and multi-hazard risk assessment, providing theoretical support and technical means for cultural heritage protection. For the impact of different types of disasters, researchers have proposed a variety of quantitative assessment tools and frameworks. For example, an evaluation model based on a comprehensive consideration of multiple influencing factors provided quantitative support for vulnerability assessment [14,15]. A multi-level seismic vulnerability assessment method was used for the risk assessment of historic buildings [16], while the Risk-UE project innovatively used macroscopic seismic models and mechanical models to assess the seismic vulnerability of historic buildings, thereby providing new ideas for disaster prevention and control [17]. In terms of flood hazard risk assessment, the MIVES method is widely used [18], and for fire hazards, a special vulnerability index framework was proposed [19]. In addition, the potential threat of landslide hazards has been systematically studied [20,21], and the impact of flood disasters on cultural heritage has been revealed through multiple studies [22,23,24]. These studies not only promote the theoretical development of cultural heritage vulnerability assessment, but also provide practical frameworks and methods for disaster prevention and control. In addition, remote sensing technology combined with GIS, analytic hierarchy processes, confidence factor models and other methods can successfully identify and evaluate threats such as landslides, floods, erosion, urban sprawl, modern road networks and fires [25,26,27], providing a scientific basis for disaster prevention and cultural heritage protection.

With the increasing threat of multiple natural disasters to cultural heritage, the multi-hazard risk assessment method has emerged and has become an important tool to assess the threat of multiple natural disasters to cultural heritage. Expert scoring [28] and analytic hierarchy process (AHP) [29] are the most common methods applied in this field, which take into account the interaction of different hazard factors through a combination of quantitative and qualitative methods. In recent years, climate change has been recognized as an emerging factor affecting cultural heritage, and researchers have proposed a comprehensive assessment approach by combining the impacts of climate change [30]. The interaction of these factors makes multi-hazard risk assessment more complex and sophisticated. In addition, for different types of cultural heritage and specific disasters, scholars have proposed a variety of risk assessment methods according to specific disaster scenarios. For example, a comprehensive assessment framework based on the three elements of “hazard–exposure–vulnerability” is widely used in the risk assessment of disasters such as earthquakes and floods [22]. In terms of post-disaster economic loss assessment, some scholars have proposed economic loss assessment indicators that specifically measure the loss of post-disaster cultural heritage assets [17,31].

While progress has been made in cultural heritage risk assessment, shortcomings remain. First, many methods rely too heavily on expert assessments, are susceptible to subjective influences, and struggle to deal with complex multivariate risks. Second, existing research ignores the fragility of cultural heritage itself, such as intrinsic factors such as the aging of building structures and materials. Most of the assessment methods focus on a single risk factor, and there is a lack of multi-factor scientific analysis.

The innovation of this study is reflected in three aspects: methodology, content and evaluation dimensions. Firstly, methodologically, a cultural heritage risk assessment framework based on LightGBM and SHAP models is proposed. Compared with the traditional analytic hierarchy process (AHP), this method avoids the bias caused by expert judgment by automatically identifying key factors and predicting risks, enhancing the objectivity and accuracy of assessment, and can cope with complex and volatile risk environments [32,33,34,35]. Secondly, in terms of content, this study breaks through the traditional framework of focusing only on external threats and includes the factors of cultural heritage itself, providing a more comprehensive risk analysis. Finally, in terms of the assessment dimension, a multi-factor and multi-level framework is proposed, which aims to integrate various risk factors and improve the objectivity and accuracy of the assessment.

This study has important academic and practical value. At the academic level, this study expands the methods and content of risk assessment of China’s cultural heritage, and proposes a new assessment framework and perspective. At the same time, the basic framework of cultural heritage risk management is analyzed in combination with the current laws and policies on cultural heritage protection, such as the relevant regulations of the State Administration of Cultural Heritage. These policies provide guidance on risk assessment, disaster preparedness and emergency management, but there is still room for improvement in spatialized risk assessment and multifactorial integrated analysis. This study used a comprehensive risk assessment approach, combined with machine learning technology and GIS, to provide a more refined risk quantification method, provide a scientific basis for the government and relevant departments to formulate protection measures, and facilitate the precise protection and sustainable management of cultural heritage.

2. Materials and Methods

2.1. Study Area

Sichuan Province is located in the southwest of China, with geographical coordinates of 97°21′~108°22′ E and 26°03′~34°19′ N, with a total area of 485,000 square kilometers. Since ancient times, Sichuan has become an important area for the study of Chinese history and culture due to its unique geography, climate and rich cultural heritage resources. There are a number of historical and cultural towns and traditional villages in the province, including 8 national and 27 provincial historical and cultural cities, more than 1165 traditional villages at or above the provincial level, 263 national key cultural relics protection units, and 3351 county-level cultural relics protection units. The geological structure of Sichuan is complex and diverse, and it has undergone many orogeny movements from the Paleozoic to the Cenozoic, forming a topographic pattern dominated by plateaus, basins and mountains.

As an important ancient transportation and trade route, the Ancient Tea Horse Road in Sichuan is not only a carrier of history and culture, but also carries rich architectural culture and historical value (Figure 1). The cultural heritage along the Sichuan Ancient Tea Horse Road is distributed in the mountains, basins and adjacent areas of Sichuan, covering multiple levels of history and culture. These cultural heritage sites are rich in types, including architectural groups of traditional settlements, religious sites, commercial centers, cultural facilities, and ancient road sites. Most of these heritage sites use traditional materials such as earth, stone, and timber, forming an architectural style with local characteristics and reflecting the uniqueness of local history and culture. Its history dates back to the Qin and Han dynasties, and even earlier, and has witnessed the rich history and cultural changes of Sichuan and its surrounding areas.

The Ancient Tea Horse Road originated from the tea and horse market in the southwest frontier during the Northern and Southern Dynasties, and after continuous development, it gradually formed a fixed trade route in the Tang Dynasty and became an important commercial channel. It is not only a channel for material exchanges, but also a center for the integration of multi-ethnic cultures. It connects Sichuan with the Central Plains, and also spans Yunnan, Guizhou, Tibet and other places, constituting an important trade route among Han, Tibet and other ethnic minorities in the southwest region. The cultural heritage along the route is distributed in complex mountainous, hilly and basin areas with complex topography and diverse geological structures, including alpine valleys and rapids. In addition, the Ancient Tea Horse Road crosses numerous large rivers and is surrounded by abundant water systems, including the Nu, Lancang, and Jinsha rivers, which are vulnerable to natural disasters such as landslides, collapses, and earthquakes. At the same time, due to the complex topography and the small area of available land, most of the settlements and buildings are located near mountains and rivers, and the spatial layout is compact, making full use of the limited natural conditions.

Therefore, systematic risk assessment, environmental protection and disaster prevention for these cultural heritage sites are important research efforts to ensure their sustainable protection and development. In this paper, the cultural heritage along the Ancient Tea Horse Road was selected as the research object, and the selection criteria were based on the three categories of traditional Chinese villages, national cultural relics protection units, and national historical and cultural towns, as well as relevant data from official data platforms, so as to ensure the representativeness of the research and the authority of the data. The spatial distribution, geographical features and risk profile of these cultural heritages were analyzed in detail using high-resolution remote sensing imagery and geographic information system (GIS) technology. The purpose of this study was to deeply explore the comprehensive value of these cultural heritage sites in multiple dimensions such as history, culture, economy, and society, and to put forward reasonable protection and utilization strategies, so as to provide a theoretical basis and practical guidance for the future protection and sustainable development of cultural heritage.

2.2. Data Sources

The basic data sources are shown in

Table 1. Basic data sources for the study area.

The Name of the Data	Data Type	Resolution or Scale	Data Sources
SRTM (Shuttle Radar Topography Mission) 30 M	Grid (.tif)	30 m	Geospatial data cloud (http://www.gscloud.cn/search, accessed on 23 December 2024)
NDVI	Grid (.tif)	250 m
Spatial distribution data of 1:1 million landform types in China	Grid (.tif)	1:1 million	Resource and Environment Science and Data Center, Chinese Academy of Sciences (https://www.resdc.cn/DataSearch.aspx, accessed on 15 December 2024)
Spatial distribution data of soil erosion in China	Grid (.tif)	1 km
Map of water system of Sichuan Province	Vector (.shp)	1:250 thousand	The National Earth Systems Science Data Center of China (http://www.geodata.cn, accessed on 15 December 2024)
Annual precipitation data set with 1 km resolution in China from 2000 to 2022	Grid (.tif)	1 km	Resource and Environment Science and Data Center, Chinese Academy of Sciences (https://www.resdc.cn/DataSearch.aspx, accessed on 18 December 2024)
Administrative division data, road data	Vector (.shp)	1: 1 million	Geospatial data cloud (http://www.gscloud.cn/search, accessed on 15 December 2024)
Data of landslide points in Sichuan Province over the years	Text (.txt)	–	Spatial distribution data of geological disaster sites in China (http://www.gisrs.cn, accessed on 27 December 2024)
Data of crumble points in Sichuan Province over the years	Text (.txt)	–	Spatial distribution data of geological disaster sites in China (http://www.gisrs.cn, accessed on 3 December 2024)
Comprehensive isoseismic map of China earthquake	Grid (.tif)	–
Chinese traditional villages, famous historical and cultural towns and villages in China	Text (.txt)	–	http://www.dmctv.cn/, accessed on 8 December 2024
National key cultural relics protection units	Text (.txt)	–	http://www.ncha.gov.cn/, accessed on 9 December 2024

. In addition, in order to determine the specific spatial location of the cultural heritage sites along the Ancient Tea Horse Road in Sichuan, the 30 m resolution historical remote sensing image data provided by Google Earth in January 2021 was obtained. The annual precipitation dataset, soil type map of Sichuan Province and water system map of Sichuan Province are from the National Earth System Science Data Center.

2.3. Methodology

The purpose of this paper was to conduct a comprehensive risk assessment of cultural heritage from the perspective of external and internal factors. The basic steps of risk assessment were divided into risk identification, risk analysis, and risk assessment. In the risk identification phase, landslides, mudslides, collapses, earthquakes and erosion were selected as natural disasters, and urban road networks were selected as human factors. In the risk analysis stage, various risk maps of the study area were drawn based on LightGBM, SHAP and GIS. Finally, a comprehensive assessment of all risks to the selected cultural heritage was carried out (Figure 2). In order to improve the stability and prediction ability of the model, parameters such as learning rate of 0.05, maximum depth of 7, number of leaf nodes of 31, and feature sampling rate of 0.7 were finally selected to prevent overfitting and merging and enhance the generalization ability of the model. At the same time, LightGBM was used to calculate the importance of all features, and the features with low contributions to the model were eliminated to reduce the interference of redundant information and improve the generalization ability of the model.

In this study, a weighted overlay analysis was used to evaluate the comprehensive risk of cultural heritage along the Ancient Tea Horse Road in Sichuan. The weights of each risk factor were determined by the feature importance calculated by LightGBM, and after the weights were normalized, they were multiplied and accumulated with the normalized risk factors to calculate the comprehensive risk index for each grid. Then, ArcGIS Spatial Analyst was used to perform a weighted overlay calculation, and the natural breakpoint classification method was used to classify the risk index into five levels: very low, low, medium, high, and very high, with higher risk values indicating greater threats to cultural heritage in the region.

In order to verify the reliability of the risk assessment results, the results were verified by field investigation. Field surveys and field surveys were carried out within the study area, and information on the destruction of cultural heritage was collected. Finally, the comprehensive risk analysis map clearly shows the spatial risk distribution of cultural heritage in the study area, which provides a basis for the formulation of risk management and protection measures.

2.3.1. LightGBM

Decision trees are machine learning models commonly used for classification and regression tasks [36]. Traditional decision trees are typically split by layer, which means that the entire training set is traversed at each iteration and all nodes in the current layer are split. This approach can lead to excessive splitting, which increases computational overhead [37,38]. In contrast, LightGBM adopts the strategy of splitting by leaves and preferentially selects the node with the largest gain for splitting so as to improve the training efficiency. LightGBM also incorporates a gradient boosting (GBDT) approach that allows it to maintain efficient performance when processing large amounts of data. Compared with traditional GBDT, LightGBM introduces techniques such as gradient-based one-side sampling and exclusive feature bundling to optimize the training process. Therefore, compared with the traditional decision tree split by layer, LightGBM has significant advantages in terms of training speed and prediction accuracy [39].

2.3.2. SHAP

Shapley Additive Explanations (SHAP) is a model interpretation method based on Shapley values in game theory proposed by Lundber [40] to understand the prediction results of machine learning models. This method is able to provide both global and local (single prediction) levels of interpretation. SHAP calculates the contribution value of each feature in all possible feature combinations and then derives the importance and influence of the feature on the model prediction. This contribution is called the SHAP value and is calculated in Equation (1). This approach helps quantify the role of each feature in the model’s decision-making, thereby improving the interpretability of the model.

$$Shapley(X_j)={\displaystyle {\sum }_{S\subseteq N\backslash \left.\left\{j\right.\right\}}\frac{k!\left(p-k-1\right)!}{p!}}\left(f\left(S\cup \left\{\left.j\right\}\right.\right)-f\left(S\right)\right)$$

(1)

where P is the total number of features, N\{j} is the set of all possible feature combinations except X_j, S is the set of features in N\{j}, f(S) is the model prediction of the features in S, and $f\left(S\cup \left\{\left.j\right\}\right.\right)$ is the model prediction j of the feature in S plus the feature X [41].

2.3.3. Disaster Risk Identification

According to the research of scholars at home and abroad, the risk assessment of cultural heritage mainly considers natural disaster factors such as landslides, erosion, earthquakes, floods, and avalanches, as well as anthropogenic factors such as urban sprawl, urban road networks, and tourism pressure [14,27]. However, the risk assessment of cultural heritage is not only a superposition of external factors, but is also affected by the characteristics of the heritage itself, including its spatial form, building density, distribution of infrastructure and other internal factors. Therefore, in order to better analyze the risks of cultural heritage along the Ancient Tea Horse Road, it is necessary to comprehensively identify and assess natural disasters, human changes and internal factors, especially highlighting the vulnerability of the heritage itself, which is essential for comprehensively assessing its conservation needs and formulating coping strategies.

The heritage risk along the cultural heritage route in Sichuan Province stems from its complex natural geography and climatic conditions. These sites range from historical trails along the Ancient Tea Horse Road to traditional settlements and architectural relics scattered throughout. Sichuan is located in a global seismic zone, with complex and active geological structures and mountainous and undulating terrain, especially the cultural heritage along the Ancient Tea Horse Road, which is generally distributed in steep mountains, canyons and areas around rivers. The geological structure of the region is complex, with broken rock formations and extensive fault zones, and heavy rainfall is frequent and intense. Heavy rainfall leads to an increase of soil moisture content and the softening or mudding of rock and soil, which reduces the stability of the slope and aggravates the instability of the mountain. The soft rock in the area is severely weathered, easy to loosen when exposed to water, and the frequent seismic activities bring about the development of rock mass cracks, which further weakens the shear strength of the rock and soil mass, and easily leads to the destruction of geological structures. The terrain is steep and the valleys are developed, and the strong currents brought by heavy rainfall exacerbate the erosion of the surface, carrying loose material downstream, creating the risk of landslides and mudslides. The concentration and frequency of heavy rainfall in the monsoon climate, as well as the destruction of vegetation by human activities, have weakened the soil consolidation capacity of the slope and aggravated the risk of geological disasters in the area. The combined effect of these factors has put the cultural heritage along the Ancient Tea Horse Road at risk of frequent and serious natural disasters.

In terms of human influence, the cultural heritage along the Sichuan Tea Horse Road is mainly threatened by modern transportation construction. The construction of modern transportation facilities such as highways often directly cuts into the route of the ancient road, causing damage to the ancient road itself and the heritage resources along the route, resulting in the destruction of the continuity and integrity of the ancient road. The vulnerability of cultural heritage is as much linked to its external influences as it is to its intrinsic characteristics and to the external effects of natural disasters and anthropogenic changes. For example, geographical location (e.g., slope area ratio, distance from rivers) is directly related to the threat of natural disasters. Factors such as building density, building quality, and the age of traditional settlements increase the risk of damage in disasters, especially older buildings that are more vulnerable to earthquakes, fires, or weathering. The coverage of disaster shelter facilities in buildings affects the ability to evacuate and recover from disasters, and the weak infrastructure makes disaster response more difficult.

The risks to cultural heritage along the Ancient Tea Horse Road in Sichuan Province can be summarized into three categories: natural disasters, human factors, and the inherent vulnerability of the heritage itself. Natural disasters include geological threats such as landslides, collapses, mudslides, earthquakes and soil erosion. The human factor is mainly due to the destruction of heritage caused by the expansion of urban road networks. The inherent vulnerability is closely related to the spatial layout, construction quality, infrastructure level and disaster response capacity of the site. These factors interact to threaten not only the safety of individual sites, but also the integrity and sustainability of the entire heritage route.

2.3.4. Disaster Risk Analysis

Earthquakes, landslides, mudslides, avalanches, erosion, urban road networks, and cultural heritage vulnerability risks were quantified and a grid of disaster risk levels was obtained. In this study, the LightGBM model was used to calculate the weights of the factors affecting the occurrence of landslides, and the landslide risk was quantified and classified by combining SHAP visual analysis and GIS spatial analysis technology. Firstly, the study area was gridded with fishing nets in GIS, and the administrative divisions along the Ancient Tea Horse Road were divided into several grid units, each of which represented a spatial analysis unit. The 3897 landslide sites used in the study (data source: http://www.gisrs.cn, accessed on 10 December 2024) were mapped to these grids. The number of landslide points within each grid was used to quantify the risk of landslides for that grid. In the influencing factor analysis stage, three categories of factors were selected: geographical environment, geological structure and human factors, and further subdivided into 24 subcategories as input characteristics for landslide risk assessment.

Through the LightGBM model, the relationship between landslide points and various environmental factors was studied, and the contribution of each factor to landslide risk was calculated. Based on the nonlinear relationship between the number of landslide points (target variables) and the influencing factors (characteristic variables), the weights of each factor were obtained. In order to determine the accuracy of the model, the root mean square error (RMSE) and the coefficient of determination (R²) were selected to compare the performance of the model. The closer R² is to 1, the better the fit of the model.

Table 2. Criteria for judging LightGBM training and test sets.

Type of Disaster	Training Set		Test Set
	RMSE	R2	RMSE	R2
Landslide	0.012	0.9212	0.043	0.9457
Collapse	0.046	0.9342	0.035	0.9091
Mudslide	0.039	0.9156	0.054	0.9364
Fragility of cultural heritage	0.087	0.8963	0.097	0.8997

shows the RMSE and R² values of the training and test sets in the model. In addition, the difference between RMSE and R² between the training set and the test set is small, which indicates that the evaluation accuracy and generalization ability of the model are high. Subsequently, the results of the LightGBM model were visualized in combination with the SHAP method, aiming to explain the specific contribution of each influencing factor in the landslide risk prediction (Figure 3). In the prediction of complex geological hazards such as landslide risk assessment, there is usually a nonlinear relationship between the characteristics and the prediction results. For example, characteristics such as slope and precipitation may not affect landslide risk in a simple linear way, but after a certain threshold, the risk changes more drastically, as shown in Figure 4, which shows the nonlinear graph of the most important indicators.

Finally, based on the weights and classification results of each influencing factor, the Spatial Analyst module of ArcGIS was used for weighted overlay analysis. The risk score for each grid was obtained by weighting and superimposing the scores of each influencing factor, and finally each grid was assigned a risk scale (1 to 5) based on the risk score. These risk levels correspond to different landslide risk intensities, with a risk level of 1 representing low risk and a risk rating of 5 representing high risk (Figure 5).

In the study, the analysis of collapse and debris flow hazards was similar to that of landslide hazards, and the LightGBM model was used. In terms of data, 1846 collapse points and 2153 debris flow points (data source: http://www.gisrs.cn, accessed on 21 December 2024) were used, which were mapped onto the corresponding grids. In terms of the selection of influencing factors, three categories were covered: geographical environment, geological structure and human factors, and they were subdivided into 22 and 20 subcategories, which were used as input features to assess the risk of collapse and debris flow.

The accuracy results of the models in the training and test sets are shown in Table 2. Among them, the collapse factors were analyzed. In order to quantify the contribution of each influencing factor, the study analyzed the importance of each factor in the model using the SHAP value (see Figure 6). Through the SHAP analysis, the nonlinear relationship between the features and the model output can be visualized (see Figure 7). In the process of calculating the risk score, the Spatial Analyst module of ArcGIS was used to weighted and superimposed each influencing factor, and finally assigned a risk level (1 to 5, level 1 is low risk, level 5 is high risk), and the results are presented in Figure 8. Then, the debris flow factors were analyzed, and the relevant graphs are shown as follows: Figure 9 shows the importance analysis of the influencing factors, Figure 10 shows the nonlinear relationship between the features and the model output, and Figure 11 shows the risk scoring results.

According to the Chinese Classification Standard for Soil Erosion Intensity [27], this study evaluated the soil erosion status in the study area. The spatial distribution data of soil erosion were obtained from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences, with a resolution of 1 km. The data were spatially tailored to the study area boundary to ensure consistency between the data and the study scope. Soil erosion in the study area was divided into five grades: micro-erosion, mild erosion, moderate erosion, intensity erosion and very strong erosion (Figure 12).

In this study, the impact of modern road engineering on the surrounding area was clarified through the analysis of the 500 m buffer zone. As urbanization progresses, the expansion of the road network has caused more areas to be disrupted by traffic. Noise, dust and traffic safety issues are often associated with major roads, which pose a threat to the cultural heritage of the neighborhood. Therefore, the area within 500 m of major transportation routes such as national highways, provincial roads, county roads, urban expressways and expressways was defined as the affected area (assigned a value of 1), and the rest of the area was regarded as an unaffected area (assigned a value of 0). The results of this analysis provide a scientific basis for assessing the potential impact of transport infrastructure on cultural heritage (Figure 13).

According to the GB18306-2015 standard, this study evaluated the seismic situation in the study area. The seismic intensity distribution data used are from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences and were spatially tailored to the boundaries of the study area to ensure data consistency. The seismic intensity of the study area was divided into seven grades according to the intensity: V (obvious tremor), VI (strong tremor), VII (severe), VIII (severe), IX (extremely strong), X (catastrophic), and XI (extremely catastrophic) (Figure 14).

In this study, 15 representative cultural heritage sites along the Ancient Tea Horse Road were selected and the vulnerability of the heritage was comprehensively assessed through field research and combined with the two dimensions of settlement space and residential space . Vulnerability was classified on a scale of 1 to 5, with data sources including field surveys and an analysis of remote sensing imagery (

Table 3. Data sources for indicators of cultural heritage vulnerability.

The Name of the Metric	Data Source	How to Obtain It
Settlement boundary shape index	Landsat 8 satellite imagery	The settlement boundary was extracted, and the shape index was calculated by GIS
Settlement boundary fragmentation	Landsat 8 satellite imagery	Raster or vectorization extracts boundary data and calculates fragmentation
Settlement building density	Landsat 8 satellite imagery	Extract the building distribution and calculate the building density
Slope area ratio	SRTM (Shuttle Radar Topography Mission) 30 M	Extract slope information and calculate the proportion of slope area
Distance from major river systems	National Center for Basic Geographic Information (https://www.ngcc.cn/dlxxzy/gjjcdlxxsjk/, accessed on 24 December 2024)	GIS analysis was used to calculate the shortest distance from the settlement center to the main river
Distance from main traffic roads	National Center for Basic Geographic Information (https://www.ngcc.cn/dlxxzy/gjjcdlxxsjk/, accessed on 21 December 2024)	GIS analysis was used to calculate the distance from the settlement center to the nearest major traffic road
The quality of the house	Field research	On-the-spot documentation of building materials and structures
The age of the house	Field research	Find the year the house was built and record it
Geological hazard points in front of and behind the house	Spatial distribution data of geological disaster sites in China (http://www.gisrs.cn, accessed on 14 December 2024), field research	Investigate the distribution of hidden danger points such as landslides and debris flows

), and quantified according to the formula in

Table 4. Calculation and analysis of cultural heritage vulnerability indicators.

Key Influencing Factors	Secondary Influencing Factors	Interpretation and Assignment of Indicators
Settlement space	Settlement boundary shape index	Interpretation: A measure of the complexity and irregularity of settlement boundaries.
		Assignment: Values can be assigned by calculating the perimeter area ratio (SI1), the ratio of the ellipse perimeter of the figure perimeter to the equal aspect ratio (SI2), and the fractal dimension of the landscape (SI3). The specific value needs to be determined based on the boundary data of the actual settlement. SI Composite Score: The composite score of SI can be calculated from the weighted average SI1, SI2, SI3. The formula is as follows: $S{I}_{final}=w_1•{\displaystyle \frac{P}{\sqrt{A}}}+w_2•{\displaystyle \frac{P_{actual}}{p_{ellipse}}}+w_3•\underset{\epsilon \to 0}{\lim }{\displaystyle \frac{\log N\left(\epsilon \right)}{\log \left(1/\epsilon \right)}}$ where P is the boundary perimeter of the settlement, A is the total area of the settlement, Pactual is the actual perimeter of the settlement, and Pellipse is the circumference of an elliptical shape equal to the area of the settlement. N(ε) is the number of boxes of the boundary measured by scale ε; ε is the box scale; and $w_1$, $w_2$, and $w_3$ are the index weights.
	Settlement boundary fragmentation	Interpretation: Refers to the degree to which the boundary of the settlement is divided, reflecting the degree of fragmentation of the settlement form.
		Assignment: The boundary fragmentation index is quantified by the dispersion or irregularity of the settlement boundary. A common calculation is based on the number of boundary fragments in relation to the surrounding environment.
	Settlement building density	Interpretation: Refers to the number of buildings per unit area, reflecting the building density of the settlement.
		Assignment: Values can be assigned by calculating the ratio of the floor area of the building within the settlement to the total area of the settlement.
	Slope area ratio	Interpretation: Refers to the proportion of sloping land area to total land area, reflecting the impact of topography on settlement development.
		Assignment: A value can be assigned by calculating the ratio of the slope area to the total land area.
	Distance from major river systems	Interpretation: Refers to the distance from the center of the settlement to the nearest river, reflecting the dependence and accessibility of the settlement on water resources.
		Assignment: Values can be assigned by measuring the straight-line distance from the center point of the settlement to the nearest river.
	Distance from main traffic roads	Interpretation: Refers to the distance from the center of the settlement to the nearest main traffic road, reflecting the convenience of the settlement to the outside world.
		Assignment: A value can be assigned by measuring the straight-line distance from the center point of the settlement to the nearest major traffic road.
Living space	The quality of the house	Wood = 5, Rammed Earth = 4, Clay Brick = 3, Green Brick = 2, Stone = 1. Each building material score is multiplied by the percentage of that building material and added.
	The age of the house	The construction age of traditional building clusters is 1 from before the founding of New China to 1980; 2 during the Republican period; 3 during the Qing Dynasty; 4 during the Ming Dynasty; Before the Ming Dynasty, it was 5.
	Geological hazard points in front of and behind the house	Interpretation: refers to whether there are hidden dangers around the structure that may cause geological disasters, such as high and steep slopes, unstable rock formations, etc. Assignment: Grading and assignment according to the severity of geological hazards, for example, no hazards are 1 point, minor hazards are 2 points, medium hazards are 3 points, and severe hazards are 4 points.

. In order to analyze the weights of factors affecting vulnerability, the LightGBM model was applied. The RMSE and R² values of the model in the training set and the test set are shown in Table 2, indicating that the model performs well in evaluating accuracy and generalization ability.

2.3.5. Risk Evaluation

LightGBM was used to construct seven main influencing factors, and the weights of each element were obtained (

Table 5. Indicator system of cultural heritage risk assessment.

Target Layer				Classification Criteria	Weight
	Classification of grades	Secondary factors	Weight
Landslide	Geographical environment	Elevation (m)	0.0881		0.1332
		Slope (°)	0.1202
		The terrain is undulating (m)	0.0166
		NDVI	0.0193
		Average annual rainfall (mm)	0.0494
	Geological formations	Semi-leached clay	0.0362
		Primary soil	0.0198
		Arid soils	0.0172
		Alpine soil	0.0037
		Leaching soil	0.0060
		Synthetic soil	0.0154
		Iron bauxite	0.0144
	The human factor	Built-up area	0.0188
		Woodland	0.0595
		Cultivated land	0.3535
		Grassland	0.0278
		Desert	0.0035
		Shrub	0.0248
		Waters	0.0360
		Population density	0.0191
		High-speed length	0.0016
		Length of national highways	0.0136
		The length of the county road	0.0138
		Dart length	0.0218
Collapse	Geographical environment	Elevation (m)	0.0833		0.0867
		Slope (°)	0.1085
		The terrain is undulating (m)	0.0973
		NDVI	0.0384
		Average annual rainfall (mm)	0.1146
	Geological formations	Semi-leached clay	0.0234
		Primary soil	0.0074
		Arid soils	0.0445
		Alpine soil	0.0076
		Leaching soil	0.0335
		Synthetic soil	0.0055
		Iron bauxite	0.0324
	The human factor	Built-up area	0.1574
		Woodland	0.0407
		Cultivated land	0.0278
		Grassland	0.0189
		Shrub	0.0179
		Waters	0.0087
		Population density	0.0474
		Length of national highways	0.0237
		Dart length	0.0371
		The length of the county road	0.0242
Mudslide	Geographical environment	Elevation (m)	0.0167		0.1190
		Slope (°)	0.0694
		The terrain is undulating (m)	0.0084
		NDVI	0.0341
		Average annual rainfall (mm)	0.0855
	Geological formations	Semi-leached clay	0.0222
		Primary soil	0.0114
		Leaching soil	0.0103
		Iron bauxite	0.1493
	The human factor	Built-up area	0.1883
		Woodland	0.0079
		Cultivated land	0.1368
		Grassland	0.0472
		Desert	0.0062
		Shrub	0.0946
		Waters	0.0272
		Population density	0.0381
		Length of national highways	0.0079
		Dart length	0.0316
		The length of the county road	0.0070
Earthquake				The tremor was noticeable	0.1392
				Strong tremors were felt
				Severe seismic sensation
				Severe earthquake
				Strong seismic sensation
				Catastrophic
				Supercatastrophic
Erosion				Micro-erosion	0.0420
				Mild erosion
				Moderate erosion
				Strength erosion
				Extreme erosion
Urban road network				Buffer	0.2380
				Other
Heritage fragility	Settlement space	Settlement boundary shape index	0.1518		0.2419
		Settlement boundary fragmentation	0.1718
		Settlement building density	0.1249
		Slope area ratio	0.0828
		Distance from major river systems	0.0661
		Distance from main traffic roads	0.0321
	Living space	The quality of the house	0.1366
		The age of the house	0.0632
		Geological hazard points in front of and behind the house	0.1979

). First, a feature set was constructed that included several major influencing factors, including landslides, avalanches, debris flows, earthquakes, erosion, urban road networks, and heritage vulnerability factors. The contribution of these factors to the risk of cultural heritage was automatically assessed through the ladder LightGBM algorithm, and the weight of each factor was calculated.

Based on the indicators and weighted scores provided, the risk to cultural heritage along the Ancient Tea Horse Road is mainly influenced by the urban road network and heritage vulnerability, which are identified as the most critical risk factors. While natural disasters such as landslides, collapses, and mudslides are frequent in the region and pose a threat to buildings and infrastructure, the construction and maintenance of urban road networks have a more far-reaching impact on the safety of cultural heritage. Areas with high traffic flow can exacerbate environmental pollution, increase human pressure, and lead to traffic accidents, all of which directly threaten the safety of cultural heritage. At the same time, the vulnerability of cultural heritage itself, especially in terms of settlement space and living space, further increases the risks it faces. In contrast, although soil erosion affects the ecological environment, it has a relatively small direct impact on the safety of cultural heritage.

In ArcGIS, risk data on heritage along the Ancient Tea Horse Road were first resampled and classified. Using a natural discontinuity taxonomy, the risk grid was divided into five levels: very high risk, high risk, medium risk, low risk, and very low risk. Next, spatial analysis was used to assign a corresponding risk level to each cultural heritage. Finally, a risk classification chart was generated and assigned a value of 1 to 5 based on five levels (very high risk, high risk, medium risk, low risk, very low risk), as shown in Figure 15.

3. Results

3.1. Results of Disaster Risk Analysis on the Ancient Tea Horse Road

3.1.1. Landslides

There is a significant spatial indifference in the landslide risk faced by cultural heritage sites along the Ancient Tea Horse Road in Sichuan, especially in very high-risk and high-risk areas. The area of the extremely high-risk landslide area is about 7825 km², accounting for 2.80% of the total area of the study area. The number of landslide hazards in these areas is dense, especially in the extremely high-risk areas, accounting for 58.86% of the total number of landslide points. Despite the small size of the very high-risk area, due to the complex geographical and geological conditions, landslides occur frequently and disasters are severe, which poses great challenges to the protection of cultural heritage.

Taking Ya’an and Chengdu as examples, landslide disasters in Jintang County, Dayi County, Pujiang County and Pengzhou City are particularly prominent. These areas are at high risk due to heavy rainfall, loose geology and steep slopes, with a high frequency of landslides and a dense distribution of cultural heritage. Some areas of Ya’an City are mainly hilly and mountainous, the geology is fragile, and the frequency of landslide disasters is exacerbated by transportation, development and construction. In high-risk areas, although the area is relatively small, the number of landslide sites still accounts for a certain proportion, mainly distributed in Meishan City and Maoxian, Lixian and Wenchuan counties in Aba Prefecture. The special geological and climatic conditions in the Aba region make landslides more common, especially at the intersection of mountains and river valleys, where cultural heritage is more threatened. Especially in Wenchuan and Li counties, natural factors such as steep slopes and heavy rainfall make landslides more likely, and cultural heritage sites in low-lying areas or along mountains are particularly vulnerable.

Although the incidence of landslides is low in medium- and low-risk areas, the protection of cultural heritage still requires attention. Landslide hazards in medium-risk areas are mainly distributed in some counties and cities in Aba Prefecture and Ganzi Prefecture, and although the frequency of landslides in these places is low, landslides may still occur in local areas due to the high mountains and hilly terrain, which will have a certain impact. As for the very low-risk areas, the number of landslide disaster sites is relatively small, but due to the large number of cultural heritage in these areas, accounting for 71.27% of the total, the potential landslide threat still exists. Especially at high altitudes, climate change is exacerbating the frequency of landslides, further compounding the difficulty of protecting cultural heritage.

According to the distribution results in

Table 6. Statistics on the risk allocation of landslides, collapses and debris flows in cultural heritage sites.

Landslide	Area (km2)	Area Percentage	Number of Cultural Heritage Sites	Percentage	Number of Landslide Points	Percentage
Extremely high risk	7825	2.80%	29	10.55%	2186	58.86%
High risk	2250	0.83%	12	4.36%	304	8.18%
Moderate risk	6025	2.23%	17	6.18%	270	7.23%
Low risk	11,850	4.38%	21	7.64%	482	12.97%
Very low risk	242,513	89.76%	196	71.27%	474	12.76%
	279,463	100	275	100	3716	100
Collapse
Extremely high risk	2275	0.81%	24	8.73%	859	49.14%
High risk	1300	0.47%	15	5.46%	208	11.90%
Moderate risk	975	0.35%	10	3.64%	117	6.70%
Low risk	3450	1.23%	27	9.82%	276	15.79%
Very low risk	271,463	97.14%	199	72.35%	288	16.48%
	279,463	100	275	100	1748	100
Mudslide
Extremely high risk	3275	1.28%	23	8.36%	602	29.12%
High risk	2200	0.79%	29	10.55%	212	10.26%
Moderate risk	6350	2.27%	33	12.00%	264	12.77%
Low risk	12,025	4.30%	22	8.00%	508	24.58%
Very low risk	255,613	91.36%	168	61.09%	481	23.27%
	279,463	100	275	100	2067	100

, the number of cultural heritage sites in the extremely high-risk landslide area is 29, accounting for 10.55% of the total, while the number of cultural heritage sites at medium risk and above is 21.09%. This shows that more than one-fifth of the cultural heritage sites along the Ancient Tea Horse Road are located in landslide-prone areas and face greater risks. In order to effectively address this challenge, governments and civil society organizations should take comprehensive precautionary measures as soon as possible and develop post-disaster recovery plans to ensure the safety of these precious cultural heritage sites.

3.1.2. Collapse

The area with a very high risk of collapse is 2275 km², accounting for only 0.81% of the total area of the study area, mainly distributed in Hanyuan County and Lushan County of Ya’an City; Jintang County, Dayi County and Pengzhou City of Chengdu City; Mao County, Li County and Wenchuan County of Aba County; and Danba County of Ganzi. First, cultural heritage in very high-risk areas is the most at risk (Table 6). Although the number of cultural heritage sites in these areas is not large, only 24, accounting for 8.73% of all cultural heritage sites, the density of landslide sites is unusually high, reaching 859, accounting for 49.14% of all landslide sites. This phenomenon reflects the complex geological structure, especially the instability caused by geological activities such as faults and folds. The mountainous and hilly terrain characteristics of these areas also make the occurrence of collapse events more frequent. Especially in the season of strong precipitation, heavy rainfall will not only increase the moisture content of the slope surface, but also easily induce soil sliding, which further aggravates the risk to cultural heritage.

Secondly, the high-risk areas are also facing a high landslide risk despite their small size, with a small number of cultural heritage sites (only 15), but 208 landslide sites, accounting for 11.90% of all landslide sites. The risk profile of these areas is also similar to that of very high-risk areas, mainly in terms of the vulnerability of geological conditions and the impact of inappropriate human activities.

In Jintang County, Dayi County, Pengzhou City and other places in Chengdu, long-term infrastructure construction and land development have destroyed the original vegetation, resulting in slope instability and increasing the probability of landslides. In medium-risk and low-risk areas, although the number of landslide sites and cultural heritage are small, there are still certain potential threats.

The area of the very low risk area is about 272,463 km², accounting for 97.14% of the total area. The number of cultural heritage sites in these regions is 199, accounting for 72.35% of the total cultural heritage. The number of landslide points was 288, accounting for 16.48% of the total landslide points. These data suggest that although the risk in these areas is relatively low, the number of landslide sites is not negligible due to their size. The protection of cultural heritage requires not only the prevention of natural disasters, but also the prevention of inappropriate land use, infrastructure construction and environmental protection.

3.1.3. Mudslides

According to the data analysis of debris flow risk areas along the Ancient Tea Horse Road in Sichuan (Table 6), the debris flow area of the extremely high-risk area is 3275 km², accounting for 1.28% of the total area, and the number of debris flow points is 602, accounting for 29.12% of the total debris flow points. The frequency of mudslides is high in these areas, especially in Shimian County, where mudslides are frequent due to complex geological formations and steep terrain, and although the number of cultural heritage sites is small, the threat of natural disasters is greater and requires special attention. The number of debris flow points in high-risk areas was 212, accounting for 10.26% of the total number of debris flow points.

Due to frequent geological activities and heavy precipitation, mudslides occur frequently in some areas of Aba Prefecture, Chengdu City and Ganzi Prefecture. Although the cultural heritage in these regions is widely distributed, there is a greater threat of mudslides, and the protection of these areas should be strengthened as a priority. Although the overall risk is low, due to the large area of the area, it is still necessary to pay attention to the impact of potential debris flow disasters on cultural heritage. Due to the relatively stable geological conditions, the probability of debris flow is relatively low, and the number of cultural heritage sites is also small, so the risk of natural disasters is relatively small. The vast majority of the area is very low risk, the probability of mudslides is extremely low, but the number of cultural heritage is large and the pressure to protect it is low. However, regular monitoring and preventive measures are still needed to ensure the long-term safety of cultural heritage. In general, the protection of cultural heritage along the Ancient Tea Horse Road in Sichuan needs to take corresponding measures according to different debris flow risk levels, especially in high-risk and extremely high-risk areas, and it is necessary to strengthen disaster prevention and cultural heritage protection.

3.1.4. Earthquakes

The seismic intensity distribution along the Sichuan Tea Horse Road is mainly concentrated in Chengdu, Meishan, Ganzi Prefecture and Aba Prefecture, and strong earthquakes are mainly felt in Meishan City, Danba County of Ganzi Prefecture and Malkang City of Aba Prefecture, while severe earthquakes are mainly felt in Chengdu City. The severe earthquakes and extremely severe earthquakes are concentrated in Litang County and Luhuo County of Ganzi Prefecture, Dujiangyan City of Chengdu and Li County of Aba Prefecture. The seismic intensity of these regions is closely related to the geological structure, topography and seismic wave propagation characteristics. From the perspective of the distribution of cultural heritage (

Table 7. Statistics on seismic intensity and soil erosion risk allocation of cultural heritage sites.

Earthquake Intensity	Area (km2)	Area Percentage	Number of Cultural Heritage Sites	Percentage
The tremor was noticeable	82,882	29.66%	24	8.73%
Strong tremors were felt	108,708	38.90%	102	37.09%
Severe seismic sensation	56,286	20.14%	86	31.27%
Severe earthquake	20,262	7.25%	38	13.82%
Strong seismic sensation	7764	2.78%	16	5.82%
Catastrophic	2420	0.87%	9	3.28%
Supercatastrophic	1141	0.41%	0	0%
	279,463	100	275	100
Soil erosion
Micro-erosion	121,806	43.59%	147	53.45%
Mild erosion	51,901	18.57%	60	21.82%
Moderate erosion	40,131	14.36%	34	12.36%
Strength erosion	6749	2.42%	16	5.82%
Extreme erosion	58,876	21.07%	15	5.46%
	279,463	100	275	100

), the area with strong earthquake sensations covers 38.09% of the area, and the number of cultural heritage sites accounts for the highest proportion, which is 37.09%. The area of severe earthquake detection covers 20.14% of the area, and the number of cultural heritage sites accounts for 31.27%. Cultural heritage in these high-intensity areas is at greater risk of earthquakes. The area with relatively low earthquake intensity was 29.66%, and the number of cultural heritage sites was 24.

In terms of geological structure, the complex geological conditions of Ganzi and Aba prefectures, especially the activity of the Xianshuihe fault zone, leads to high seismic intensity in these areas. The Chengdu area is close to the Longmenshan fault zone, with frequent seismic activity and large topographic fluctuations, which makes the seismic waves easy to amplify. In terms of topography, in mountainous areas such as the valleys and basins of Luhuo County, the energy of seismic waves does not decay when propagating in these places, resulting in increased intensity. The soil in the Chengdu Plain is soft, and the attenuation of seismic waves is slower, and even if it is far away from the epicenter, the propagation of seismic waves in the basin will still bring about strong vibrations. The characteristics of seismic wave propagation make the seismic intensity in these areas more complex, especially in places where the terrain is undulating, and the seismic waves are enhanced by the influence of the topography.

Although the seismic fortification standard for these areas is seven or eight degrees, and the building design has considered a certain seismic capacity, in the event of a strong earthquake, the intensity may exceed the fortification standard, and additional seismic measures are still required. Catastrophic and mega-catastrophic areas are smaller in size, at 1.28%, and, despite their small cultural heritage content, can still pose a threat to neighboring cultural heritage sites due to their high destructive power. Therefore, in these high-risk areas, seismic reinforcement and emergency response measures should be strengthened to effectively protect cultural heritage.

3.1.5. Erosion

The distribution of soil erosion along the Ancient Tea Horse Road in Sichuan shows obvious regional differences, and the distribution of soil erosion is closely related to natural factors, climatic conditions and human activities. According to the data analysis in Table 7, the micro-erosion zone accounts for 43.59% of the total area, and the number of cultural heritage sites is the largest, accounting for 53.45% of the total. Although the level of soil erosion in the region is relatively low, the high density of cultural heritage puts greater pressure on the conservation of these areas. This area is mainly distributed in Chengdu, Ya’an and other places.

The mild erosion area accounts for 18.57% of the total area, and the number of cultural heritage sites is 60, accounting for 21.82% of the total cultural heritage. Although the level of soil erosion is low, it has increased locally due to human activities.

The moderately eroded areas accounted for 14.36% of the total area, and the number of cultural heritage sites is 34, accounting for 12.36% of the total. The degree of soil erosion in this area is more significant, especially in the alpine valleys of Ganzi Prefecture and Aba Prefecture, such as Luhuo County and Litang County, due to the large undulating terrain and concentrated precipitation, the soil erosion phenomenon is more serious, and the vegetation coverage rate is low, resulting in the weak natural protection ability of the soil, which further aggravates the soil erosion.

The intense erosion zone and the extremely strong erosion zone accounted for 2.42% and 21.07% of the total area, respectively, and the number of cultural heritage sites is 16 and 15, respectively. Especially in the mountains and valleys, the complex terrain, steep slopes, frequent heavy rainfall, and the impact of human activities have made the problem of soil erosion more serious, and the protection of cultural heritage is facing high risks. These areas mainly involve Danba County and Daofu County.

In general, an increase in the degree of soil erosion poses a great challenge to the protection of cultural heritage, especially in areas of intensity and extreme erosion, and the protection of cultural heritage is significantly more difficult and effective soil protection and heritage protection measures are urgently needed.

3.1.6. Urban Road Networks

The urban road network along the Sichuan Tea Horse Road has had a significant impact on the preservation of cultural heritage. According to the statistics (

Table 8. Statistics on the risk allocation of the road network in cultural heritage cities.

	Area (km2)	Area Percentage	Number of Cultural Heritage Sites	Percentage
Buffer	21574.1	7.72%	179	65.10%
Other	257888.9	92.28%	96	34.90%
	279463	100	275	100

), the buffer zone covers an area of 21,574.1 km², accounting for 7.72% of the total area of the study area, including 179 cultural heritage sites, accounting for 65.10% of the total cultural heritage in the area. The cultural heritage of this area is highly dense and has been greatly influenced by the expansion of the city’s road network. With the advancement of urbanization, road construction may trigger land use changes, which in turn threatens the integrity of historical sites and the stability of the ecological environment. In addition, the increase in traffic flow may lead to a series of negative effects such as environmental pollution and noise pollution, which will further exacerbate the vulnerability of cultural heritage areas.

The cultural heritage along the Ancient Tea Horse Road is often an important tourist destination, and the improvement of transportation convenience may increase the flow of tourists [42], thereby increasing the pressure of human activities on heritage protection. Therefore, in the planning and construction of transportation networks, the protection needs of cultural heritage should be fully considered to ensure the sustainability of cultural heritage during the process of urban development and modernization.

3.1.7. Heritage Fragility

After using LightGBM to calculate the index weight of heritage vulnerability, the comprehensive index method was used to calculate the objective level vulnerability index of the cultural heritage historical environment; the larger the index, the higher the vulnerability.

$$OV={\displaystyle \sum _{q=1}^5\left({\displaystyle \sum _{i=1}^{n}O{w}_i•X_i}\right)}•O{W}_q$$

(2)

where: OV represents the objective historical environmental vulnerability index; $O{w}_i$ represents the weight of the ith indicator at the objective level; X_i denotes the normalized value of indicator i; and OW_q represents the weight of the q-criterion layer at the objective level.

The objective vulnerability index of the historical environment of the heritage was calculated by using the above method, and the natural breakpoint method was used to divide it into five vulnerability levels: very low vulnerability, low vulnerability, medium vulnerability, high vulnerability and very high vulnerability.

The vulnerability of cultural heritage along the Ancient Tea Horse Road in Sichuan presents obvious spatial distribution characteristics, which is affected by the natural environment, geographical conditions and human activities. According to the vulnerability assessment (

Table 9. Statistics on the distribution of cultural heritage vulnerability risks.

	Number of Cultural Heritage Sites	Percentage
Extremely vulnerable	88	32.00%
High vulnerability	68	24.73%
Medium vulnerability	46	16.73%
Low vulnerability	34	12.36%
Very low vulnerability	39	14.19%
	275	100

), 32.00% of the cultural heritage sites are classified as extremely high vulnerability, most of which are located in mountainous and sloping areas where natural disasters are frequent, with complex and fragmented settlement boundaries, high building density and mostly traditional materials, poor building quality, lack of modern reinforcement measures, and vulnerability to natural disasters such as landslides, mudslides and earthquakes. At the same time, 24.73% of the cultural heritage is distributed in high-vulnerability areas, where buildings are in disrepair, and the hidden danger of geological disasters is more prominent, and some settlements are close to major rivers and face a high risk of flooding. An estimated 16.73% of the cultural heritage is in the medium vulnerability zone, and although the quality of construction and disaster prevention facilities have improved, it still faces the threat of natural disasters, especially in the context of extreme weather and intensified tourism development. Only 12.36% of the cultural heritage sites are in a low vulnerability zone, which have good construction quality, relatively complete disaster prevention facilities, and low vulnerability at the time of disaster, but they are still affected by overexploitation and human activities. Finally, 14.19% of the cultural heritage sites are located in areas of very low vulnerability, which are mainly distributed in relatively stable geographical environments, with high construction quality, wide coverage of disaster prevention facilities, and low overall disaster risk. In general, the vulnerability of cultural heritage along the Sichuan Tea Horse Road is affected by multiple factors, such as geographical environment, construction quality, and frequency of natural disasters.

In summary, the cultural heritage along the Sichuan Tea Horse Road is mainly concentrated in extremely vulnerable and high-fragile areas, which face multiple natural and artificial threats, and it is urgent to strengthen disaster prevention and control, rationally plan infrastructure, improve disaster prevention facilities and enhance disaster resilience. In addition, cultural heritage in areas of medium and low vulnerability should be given sustained attention to ensure that its long-term protection is not neglected.

3.2. Analysis of the Risk Assessment Results

The risk classification of the study area showed that the very low risk area and the low risk area occupied most of the area, which were 145,963 km² (52.23%) and 106,775 km² (38.21%), respectively (

Table 10. Risk assessment and classification results of the cultural heritage of the Ancient Tea Horse Road.

	Zoning Area/km2	Area Ratio	Number of Cultural Heritage Sites	Ratio of the Number of Cultural Heritage Sites
Low risk area	145,963	52.23%	68	24.73%
Low-risk areas	106,775	38.21%	63	22.91%
Medium risk area	16,900	7.00%	66	24.00%
High-risk areas	5375	6.05%	42	15.27%
Very high-risk areas	4450	1.59%	36	13.09%
	279,463	100%	275	100%

), indicating that most of the study area in the Ancient Tea Horse Road Cultural Heritage Area was at low risk. However, most of the cultural heritage sites are located in areas of medium risk and above, accounting for 52.36% of the total, indicating that more than half of the cultural heritage is at high risk, facing greater threats of natural disasters and anthropogenic changes, indicating a need to strengthen categorical protection measures (Figure 15).

Based on the summary and analysis of the spatial distribution characteristics of cultural heritage in risk areas at all levels, the risk level of cultural heritage along the Sichuan Tea Horse Road is closely related to the geographical environment, geological activities, climatic conditions and human activities. The extremely high-risk areas are mainly distributed in Ya’an City and parts of Pengzhou, Dujiangyan and Chongzhou in Chengdu, covering an area of 4450 km², accounting for 1.59% of the total area. The Nu River, the Lancang River, the Jinsha River, the Dadu River and other rivers cut the plateau into many peaks and valleys, the terrain is rugged, and the traffic conditions are difficult. Despite its small size, the number of cultural heritage sites is 36, accounting for 13.09% of the total cultural heritage. The high-risk areas are mainly distributed in Mao County, Li County and Wenchuan County in Aba Prefecture, accounting for 6.50% of the total area. Located in the Minshan Mountains, Aba is an important part of the western route of the Ancient Tea Horse Road, with frequent seismic activity and complex terrain, which is prone to geological disasters such as landslides and mudslides, which increases the risk to cultural heritage sites.

The medium-risk areas are mainly distributed in Heishui County and Jiuzhaigou County in Aba and Luding; and Luhuo County, Daofu County, Litang County and Batang County in Ganzi, with an area of 16,900 km², accounting for 7.00% of the total area. The terrain in these areas is undulating, there are many canyons, the geological structure is complex, the water flow is turbulent, and the climate is changeable, and it is very easy to induce disasters such as collapse and debris flow, especially in the mountain and valley terrain areas such as Daofu County. The risk of water erosion is enhanced, and the disasters caused by extreme weather phenomena are frequent, which increases the risk to cultural heritage.

The low-risk area accounts for 38.21% of the total area, and the number of cultural heritage sites is 63, accounting for 22.91% of the total cultural heritage. Very low-risk areas account for 52.23% of the total area. The amount of cultural heritage in these regions is relatively small, accounting for 24.73% of the total cultural heritage. The geological stability is good, and the probability of disasters is low, so the protection of cultural heritage is relatively simple, and the main focus is on the impact of human activities.

In summary, the distribution of cultural heritage risks along the Sichuan Tea Horse Road shows significant spatial heterogeneity. Although the low-risk and very low-risk areas cover most of the area, the number of cultural heritage sites in the high-risk and very high-risk areas accounts for more than half of the sites, and their protection challenges are particularly severe due to the frequent occurrence of natural disasters. Therefore, during the process of cultural heritage protection, it is particularly necessary to strengthen the protection of high-risk and extremely high-risk areas and optimize disaster prevention and emergency response mechanisms to ensure the safety of important cultural heritage sites in these areas.

3.3. Spatial Autocorrelation of Risk

Spatial autocorrelation measures the dependencies and correlations between the attribute values of point elements and their neighboring point elements, including the global Moran I. Spatial autocorrelation was first proposed by P.A.P. Moran in 1950. It can be used to explore the correlation between element properties and adjacent area elements. Global Moran’s I was used to verify the presence of spatial correlations between elements within the study area [27]. Using the spatial statistics tools of ArcGIS 10.5, a global spatial autocorrelation analysis of the comprehensive risk value of cultural heritage was carried out. The results show that the global Moran I value is about 0.455 and the p value is 0.000000, which is much less than 0.05. The 99% confidence test shows that the Z-value is about 64.333, which is much greater than 1.65, indicating that the comprehensive risk value of cultural heritage has a significant positive spatial distribution and is clustered.

4. Discussion

This study used the LightGBM model to comprehensively assess the cultural heritage risks along the Ancient Tea Horse Road. By integrating multiple factors such as natural disasters, human activities and the vulnerability of the heritage itself, a multi-dimensional risk assessment framework was constructed. The results of the study reveal the complexity of the risks faced by cultural heritage along the Ancient Tea Horse Road.

(1) This study reveals the vulnerability of cultural heritage along the Ancient Tea Horse Road to the impact of natural disasters, especially earthquakes, landslides, avalanches and mudslides. The complex geological structure, high altitude and frequent seismic activity in the study area have made the cultural heritage face a high disaster risk for a long time. This conclusion is consistent with existing research that cultural heritage is more vulnerable to natural disasters in areas with intense geological activity, steep terrain slopes and complex geological condition [1,16,21]. In addition to natural factors, human activities are also altering the original ecological environment and increasing the risk exposure of cultural heritage. Large-scale infrastructure construction has destroyed natural barriers, resulting in a significant increase in the frequency of landslides and mudslides [43], This conclusion is highly consistent with the results of previous studies [44,45]. Therefore, rational planning of the relationship between human development activities and cultural heritage protection and formulating scientific land use policies are important contents of cultural heritage protection.

This study further analyzed the impact of settlement spatial structure (e.g., settlement boundary shape index, building density) and environmental factors (e.g., slope, distance from major rivers) on cultural heritage risk. The results show that the spatial pattern and environmental characteristics of settlements have a significant impact on the risk level, especially in earthquake- and landslide-prone areas. For example, due to soil instability and serious soil erosion in steep slope areas, the probability of landslide disasters has increased significantly, which in turn increases the vulnerability of cultural heritage. This finding is consistent with the conclusion that slope is regarded as a key factor of ecological risk in previous studies [46]. In addition, high-density buildings and complex boundary settlements may make evacuation more difficult after a disaster, thus increasing the risk of loss of cultural heritage. This is especially true in high-density settlement areas such as the Hengduan Mountains. Settlements close to major rivers are also at high risk to their cultural heritage due to soil erosion and flooding.

In terms of risk assessment, the LightGBM model was used to improve the accuracy and stability of risk prediction. Compared with the traditional analytic hierarchy process (AHP), the LightGBM model can automatically process high-dimensional variables, reduce the influence of human subjective factors on the evaluation results, and show strong prediction ability in complex environments. Its efficient training speed and ability to adapt to large-scale datasets make it outstanding in capturing nonlinear relationships. Combined with SHAP value analysis, the contribution of different influencing factors to cultural heritage risk can be more clearly quantified so as to improve the interpretability of the model [47]. Its core strength lies in its ability to provide both global and local explanations, which not only reveal the contribution of each feature to the overall model prediction, but also analyze how specific environmental factors affect the prediction results for a single sample. In addition, SHAP is suitable for Gradient Boosting Decision Tree (GBDT) models, which can accurately measure the contribution of input features to the prediction results, making model decisions more transparent. Therefore, this study used SHAP to conduct feature importance analysis, identify key influencing factors, and visually demonstrate the specific contribution of different environmental factors to individual disaster susceptibility prediction.

Although some progress has been made in this study, there are still some limitations. First, the effectiveness of the LightGBM model depends on data quality and completeness, and especially the lack of historical disaster data may affect the accuracy of risk assessment. Second, although LightGBM is capable of addressing most risk assessment scenarios, it may still be necessary to combine other modeling approaches to optimize the assessment accuracy in more complex environments. Future research may consider introducing multi-source data, such as remote sensing monitoring data and field survey data, to improve the generalization ability of the model and further explore the long-term impact mechanism of different risk factors on cultural heritage security.

Cultural heritage is not only about physical structures, but also about the natural environment in which it is located and the culture of the community. This study focused on the impact of natural disasters, infrastructure and settlement characteristics on cultural heritage, with less emphasis on social, spiritual and community values. However, the vulnerability of cultural heritage is not only influenced by geological and environmental factors, but also by socio-economic development, community awareness and cultural heritage. For example, community attitudes toward the preservation of cultural heritage, visitor activities, religious beliefs and cultural identities may all affect the long-term preservation of heritage. Future research can combine information such as social surveys, cultural identity assessments, and community participation to further improve the heritage vulnerability assessment system, make the assessment results more comprehensive, and provide more scientific support for the protection and management of cultural heritage.

(2) This study highlights the need for systematic protection measures to protect cultural heritage. Effective risk management relies on strict regulations, a sound early warning system and professional management. All localities should strengthen the implementation of laws and regulations, optimize monitoring and early warning, and regularly maintain protection facilities, especially in high-risk areas such as those prone to earthquakes and landslides. In addition, unnatural factors such as a deterioration in environmental quality, agricultural pollution and overtourism also exacerbate cultural heritage risk [27]. There is a need to raise public awareness of disaster preparedness to enhance community participation and governance effectiveness. In the future, we should rely on remote sensing monitoring, big data and other technologies to optimize risk assessment, promote collaborative governance among the government, scientific research and society, improve the adaptability of cultural heritage, and provide a reference for high-risk areas.

5. Conclusions

This study used the LightGBM-based risk assessment method to conduct a comprehensive analysis of the cultural heritage along the Ancient Tea Horse Road in Sichuan Province. Taking into account the vulnerability to natural disasters, human factors and cultural heritage itself, this study showed that the vulnerability of urban road networks and cultural heritage itself are two key factors affecting the risk of cultural heritage along the Ancient Tea Horse Road. First, the expansion of urban road networks and anthropogenic activities have increased the direct threat to heritage sites, particularly environmental changes caused by transport construction and industrialization, increasing their exposure to natural disasters. Second, the fragility of cultural heritage itself (e.g., aging structures, lack of necessary protection measures, etc.) makes it vulnerable to serious damage in the event of natural disasters such as landslides, collapses, and earthquakes. Natural disasters themselves, particularly landslides, collapses, earthquakes and soil erosion in mountainous and valley areas, further exacerbate the risks of the site. These factors are intertwined, putting multiple pressures on the protection of cultural heritage. Therefore, on the basis of the traditional natural disaster risk assessment, this study incorporated the inherent vulnerability factors of cultural heritage into an assessment system, providing a more comprehensive and accurate assessment perspective. Some of the cultural heritage along the Ancient Tea Horse Road is at a very high level of vulnerability due to complex geological structures, steep terrain and human activities. At the same time, while some sites are less threatened by natural disasters, they are still potentially threatened by urbanization and overexploitation of tourism. The risk assessment framework of this study provides scientific decision-making support for future cultural heritage protection and emphasizes the importance of comprehensive consideration of internal and external factors in cultural heritage protection.

On this basis, in order to more effectively reduce the risks faced by cultural heritage, it is recommended to start from multiple aspects. First of all, during the process of urban development and transportation construction, rational planning should be made to reduce direct interference with cultural heritage, and the implementation of protection policies should be strengthened, so as to achieve parallel protection and development. Second, a geological hazard monitoring system can be established to improve risk early warning capabilities, and appropriate restoration technologies, such as 3D scanning and digital archiving, can be used to ensure the long-term preservation of cultural heritage information. In view of the potential impact of tourism, measures can be taken to limit the flow of tourists and guide tourists to participate in low-impact ecotourism to reduce the pressure of overexploitation. In addition, governments, communities and relevant institutions need to work together to develop more operational strategies for the protection of cultural heritage to ensure its sustainable development.