Exploring city-level climate risks for immovable cultural heritage in Beijing

Beijing is a world-renowned historical and cultural city, boasting over 3000 years of urban development and more than 800 years of history as the national capital. It ranks among the Chinese cities with the highest number of World Heritage Sites and the greatest concentration of cultural relics. According to the Law of the People’s Republic of China on the Protection of Cultural Relics35, immovable cultural relics are classified into six categories: sites of ancient culture (SOAC), ancient tombs (AT), ancient architectural structures (AAS), cave temples and stone carvings (CTASC), important modern and contemporary historic sites and typical buildings (IMACHSTB), and murals. The first five categories are predominantly exposed to outdoor environmental conditions, whereas murals are generally located indoors and thus less susceptible to direct climate-change-induced environmental stressors. Consequently, this study focuses on these five outdoor heritage categories.

A total of 3619 immovable cultural heritage sites are included in the analysis, distributed across 16 administrative districts characterized by alternating plains, hills, and mountains under a temperate monsoon climate with pronounced seasonality and inter-annual variability. These sites exhibit a distinct spatial pattern, with high concentrations observed in Dongcheng and Xicheng districts as well as in foothill transition zones, forming a clear “central concentration—peripheral dispersion” structure. China’s immovable cultural relics are classified according to protection level into national-level, municipal-level, district-level, and district-wide surveyed heritage sites. Among the 3619 cultural heritage sites included in this study, 147 are designated at the national level, 263 at the municipal level, 789 at the district level, and 2420 fall under the category of district-wide surveyed cultural relics. To facilitate understanding of the spatial scope and classification framework applied in the subsequent risk assessment, Fig. 1 illustrates the overall spatial distribution of immovable cultural heritage in Beijing.

Fig. 1: Spatial distribution map of cultural heritage in Beijing.

This figure illustrates the geographical distribution of cultural heritage sites across different protection levels in Beijing, with the background colors depicting the city’s topographical features.

Research framework

To systematically assess the risks to Beijing’s cultural heritage under the context of climate change, this study established a three-dimensional indicator system centered on “Hazard-Vulnerability-Exposure”. The detailed methodological framework is illustrated in Fig. 2.

Fig. 2: The methodological framework for climate risk assessment of cultural heritage in Beijing.

This figure presents a comprehensive overview of the methodological framework for climate risk assessment of cultural heritage in Beijing.

Risk assessment framework

The study employed the IPCC risk framework to quantify the climate risk of immovable cultural heritage in Beijing.

In the equation, R represents the climate change risk, H denotes the hazard of climate change, E represents the exposure of the elements at risk, and V denotes the vulnerability of the elements at risk. \({{\rm{W}}}_{{\rm{h}}}\)、\({{\rm{W}}}_{{\rm{e}}}\)、\({{\rm{W}}}_{{\rm{v}}}\) are the respective weights of each component.

Risk assessment indicators

Based on a comprehensive review of existing research and data availability, this study selected 14 key indicators to represent the hazard, exposure, and vulnerability. The indicator system exhibits strong logical coherence with respect to data availability, spatial adaptability, and risk response mechanisms, thereby providing a robust foundation for the development of a city-scale climate risk assessment model for cultural heritage.

In the hazard dimension, emphasis was placed on capturing both the intensity and temporal characteristics of climatic factors relevant to cultural heritage deterioration. According to existing literature, climatic factors, including temperature, precipitation, wind, lightning, and freeze-thaw cycles, exert significant influences on the deterioration of cultural heritage36,37,38,39. Climate-related risks arise not only from sudden and acute physical damage caused by extreme climatic events such as heavy rainfall and typhoons, but also from long-term shifts in key climatic variables such as temperature, precipitation, and humidity. These gradual changes contribute to accelerated deterioration through “dose-response” mechanisms, manifesting as flaking, blistering, cracking, color alteration, and rot39,40,41. Climate risk assessments typically require long-term future climate projections. However, current climate models remain limited in their ability to reliably simulate phenomena such as wind patterns and lightning frequency over extended time scales. Therefore, this study focuses exclusively on temperature and precipitation as the two primary climatic variables. Annual mean temperature and annual mean precipitation were selected as representatives of long-term cumulative effects, reflecting thermal stress and moisture perturbation, respectively. Considering the pronounced abrupt impacts of extreme events on cultural heritage, maximum annual mean temperature and maximum annual mean precipitation were introduced as proxy variables for extreme heat and extreme precipitation.

In the exposure dimension, heritage value represents the social influence and historical significance of cultural heritage, with high-value sites generally exhibiting higher irreplaceability and greater potential losses, thus experiencing higher exposure. The Heritage scale reflects the physical spatial extent, where larger areas involve more pathways for climatic disturbances, resulting in more pronounced risk accumulation effects.

In the vulnerability dimension, intrinsic vulnerability is determined by the construction age and material composition of heritage sites. Construction age broadly reflects the duration of accumulated environmental exposure, with older structures generally exhibiting higher vulnerability. However, significant variations among individual cultural heritage sites may arise due to differences in materials and construction techniques. Traditional Chinese architectural craftsmanship, such as the dougong bracket system, often provides exceptional mechanical adaptability and resilience. Therefore, construction age should be regarded as a generalized proxy for long-term deterioration potential rather than an absolute determinant of intrinsic vulnerability. Material type influences thermal and moisture resistance as well as structural porosity; for instance, wood and rammed earth are highly sensitive to fluctuations in temperature and humidity.

Environmental vulnerability incorporates slope, distance to environmental risk points, vegetation cover, and land use type, collectively reflecting the capacity of the geographic environment to amplify or buffer climate hazards. Steep terrain, proximity to hazard points, and degraded ecosystems tend to exacerbate climatic impacts.

Regarding socioeconomic vulnerability, the fiscal revenue of the administrative region in which the heritage site is located serves as a proxy for protective resource capacity, with higher revenue reflecting greater institutional capacity for management and maintenance, thereby contributing to reduced vulnerability. The population exerts a multifaceted influence on the vulnerability of cultural heritage. In densely populated areas, heritage sites often benefit from stronger community engagement in conservation efforts, yet they are simultaneously more susceptible to human-induced threats such as urban encroachment, tourism pressure, and infrastructure development. Conversely, sites in sparsely populated regions experience fewer anthropogenic disturbances but frequently lack institutional maintenance and monitoring, particularly for heritages with lower conservation grades, thereby increasing their exposure to natural degradation processes. This study employs population density within a 3-km radius as a proxy indicator for socio-cultural vulnerability. Higher population density within this buffer is associated with greater potential for local stewardship and thus reflects reduced human-related vulnerability.

Risk scenarios

This study selects two climate scenarios, SSP2-4.5 and SSP5-8.5, to assess risk disparities under different climatic conditions. SSP2-4.5 represents a moderate radiative forcing scenario, stabilizing at approximately 4.5 W/m² by 2100, which is widely regarded as a plausible future pathway. SSP5-8.5 corresponds to a high radiative forcing scenario, with forcing reaching up to 8.5 W/m² by 2100, and is generally considered to represent the upper bound of potential future climate change.

Weight of the indicators

This study employs a hybrid approach integrating both subjective and objective methodologies to determine indicator weights. In the hazard dimension, the Entropy Weight Method is employed to automatically assign indicator weights based on the sample distribution, ensuring both the distinguishability among climate disturbance variables and the objectivity of the weights. For other weights, it is challenging to compute them using purely objective weighting methods. In this study, the Analytic Hierarchy Process was employed to determine these weights. A pairwise comparison consistency matrix was constructed by conducting mutual evaluations of seven indicators using seven AI platforms—ChatGPT, Doubao, KIMI, Wenxin Yiyan, DeepSeek, Hailuo Wenwen, and Tongyi Qianwen—in order to enhance assessment accuracy. To ensure the comparability of the evaluations from each AI, a unified set of questions was presented to all platforms. As different AI models yielded distinct comparison matrices, their consistency was tested, typically using the following formula:

In the equation, \({CI}\) denotes the consistency index, n represents the order of the matrix, and \({{\rm{\lambda }}}_{\max }({\rm{A}})\) is the maximum eigenvalue of matrix A. When \({CI}\) = 0, it indicates perfect consistency; the larger the \({CI}\), the greater the degree of inconsistency. To measure the magnitude of \({CI}\), the random consistency index (\({RI}\)) is introduced, which is calculated using the same method as \({CI}\). On this basis, the random consistency ratio is then computed as follows:

When \({CR}\) < 0.1, the pairwise comparison matrix is considered to have satisfactory consistency, or its degree of inconsistency is deemed acceptable; otherwise, the matrix must be adjusted until satisfactory consistency is achieved. In this study, all AHP judgment matrices passed the consistency test. The results are shown in Table 1.

Data sources

This study employs both historical and projected climate data to evaluate the climate risk of cultural heritage sites in Beijing. Historical climate data (1901–2022) were obtained from the National Qinghai–Tibet Plateau Science Data Center (https://data.tpdc.ac.cn/home). The data have a spatial resolution of 1 km (0.008°) and a temporal resolution of monthly, covering temperature and precipitation variations across China42,43. Future climate data (2021–2100) were also sourced from the National Qinghai–Tibet Plateau Science Data Center. The dataset is based on the CMIP6 multi-model ensemble under three representative Shared Socioeconomic Pathways (SSP1-1.9, SSP2-4.5, and SSP5-8.5), simulated by global climate models including EC-Earth3, GFDL-ESM4, and MRI-ESM2-0. A Delta downscaling method was applied to achieve a 1 km resolution, with the period 1981–2010 used as the baseline for systematic bias correction43,44.

The Exposure and Intrinsic vulnerability indicators were derived from the ‘Beijing Cultural Relics Map’ platform (https://maptable.com/s/p/cnzodzkujocg/map). The dataset includes geographical information, age information, and the types of heritage. Environmental vulnerability is quantified at the individual site scale using GIS, based on a 30 m resolution DEM from the European Space Agency’s Copernicus program (2022), Beijing geological hazard points (2019) and land-use data (2023) from the Chinese Academy of Sciences Resource and Environmental Science Data Center (https://www.resdc.cn/ data.aspx?DATAID = 290), and monthly NDVI indices (2023) from NASA (https://www.earthdata.nasa.gov). Socio-economic vulnerability is represented by population density at 1 km resolution from WorldPop and regional fiscal revenue in 2023 from the Beijing Statistical Yearbook (https://nj.tjj.beijing.gov.cn/nj/main/2024-tjnj/zk/indexch.htm).

Risk modelling processHazard

This study selects four indicators – annual mean temperature, maximum annual mean temperature, annual mean precipitation, and maximum annual mean precipitation – to analyze their spatial distributions and temporal trends under two emission scenarios: the intermediate scenario (SSP2-4.5) and the high-emission scenario (SSP5-8.5).

Figure 3 illustrates the comparison of temperature under different climate scenarios. During the historical period, the spatial pattern of annual mean temperature (Fig. 3a) displays a “higher in urban centers, lower in mountainous areas” distribution. The central districts (Dongcheng, Xicheng, and Chaoyang) generally exceed 14 °C, while the northwestern mountainous regions (Yanqing and Miyun) remain below 12 °C. Under the SSP2-4.5 scenario (Fig. 3b), these mountainous regions are projected to experience warming of approximately 3.0 °C; under the SSP5-8.5 scenario (Fig. 3c), temperatures increase generally exceed 3.5 °C, with localized areas surpassing 4.0 °C. The spatial heterogeneity of maximum annual mean temperature is even more pronounced. During the historical period (Fig. 3d), high temperature values are predominantly concentrated in the southeastern plain districts, including Tongzhou, Daxing, and Shunyi, approaching 33 °C. Under the SSP2-4.5 scenario (Fig. 3e), temperatures in these regions increase by approximately 2 °C. In the SSP5-8.5 scenario (Fig. 3f), most plain areas exceed 35 °C, with localized zones approaching 36 °C. Both the frequency and intensity of extreme heat events are projected to rise substantially, posing significant thermal stress risks to heat-sensitive heritage, such as wooden structures and rammed-earth sites.

Fig. 3: Comparison of temperature under different climate scenarios.

a Annual mean temperature during the historical period (1901–2022); b Annual mean temperature under the SSP2-4.5 scenario (2023–2100); c Annual mean temperature under the SSP5-8.5 scenario (2023–2100); d Maximum annual mean temperature during the historical period (1901–2022); e Maximum annual mean temperature under the SSP2-4.5 scenario (2023–2100); f Maximum annual mean temperature under the SSP5-8.5 scenario (2023–2100).

Figure 4 illustrates the comparison of precipitation under different climate scenarios. During the historical period, the spatial pattern of annual mean precipitation (Fig. 4a) displays a “higher in the southwest, lower in the northeast” distribution, with southwestern mountainous regions, such as Fangshan and Mentougou, receiving 600–650 mm annually. Under future scenarios, precipitation is projected to increase overall, with more pronounced increments under the high-emission pathway. Specifically, under the SSP2-4.5 scenario (Fig. 4b), increases in the southwestern mountains and southeastern plains range from 50 to 150 mm, whereas under the SSP5-8.5 scenario (Fig. 4c), localized increments exceed 200 mm. For maximum annual mean precipitation (Fig. 4d), high values during the historical period are concentrated in the southwestern mountain valleys, reaching approximately 800 mm. Under the SSP2-4.5 scenario (Fig. 4e), the high-precipitation zones expand; under the SSP5-8.5 scenario (Fig. 4f), they extend markedly into the plain districts of Fengtai, Daxing, and Tongzhou, with localized values exceeding 1000 mm, nearly double the historical levels. Such short-duration, high-intensity precipitation events are highly likely to trigger flooding and secondary geological hazards, posing significant risks of surface erosion and foundation scouring for permeable heritage structures, including stone carvings and masonry.

Fig. 4: Comparison of precipitation under different climate scenarios.

a Annual mean precipitation during the historical period (1901–2022); b Annual mean precipitation under the SSP2-4.5 scenario (2023–2100); c Annual mean precipitation under the SSP5-8.5 scenario (2023–2100); d Maximum annual mean precipitation during the historical period (1901–2022); e Maximum annual mean precipitation under the SSP2-4.5 scenario (2023–2100); f Maximum annual mean precipitation under the SSP5-8.5 scenario (2023–2100).

Exposure

In this study, exposure is defined as a weighted combination of heritage value and site scale, comprehensively capturing the social-historical significance and the physical-spatial susceptibility of the heritage system, thereby constituting a key component of climate risk assessment.

Cultural value reflects the historical significance, artistic merit, and societal influence of heritage sites, serving as a key basis for determining protection levels and allocating management resources. Typically, cultural heritage value decreases progressively from national-level to municipal-level, district-level, and district-wide surveyed sites. This study calculates the proportions of these heritage categories in Beijing as 1:1.8:5.4:16.4. Under the assumption of equal total cultural value across levels, corresponding weights are derived and presented in Table 2. This weighting system reflects the increased potential exposure of higher-level heritage sites, attributable to their irreplaceable cultural value and greater restoration challenges. In the GIS-based spatial analysis, each site is assigned a value based on its protection level and spatially represented, thereby providing discrete data inputs for the exposure model.

The heritage scale defines the spatial extent of a site in geographic space. Larger areas are associated with higher structural complexity, greater microclimatic diversity, and wider potential pathways for climate impacts, thus increasing exposure45. In this study, boundaries of national-level and municipal-level heritage units were extracted from the Beijing Cultural Heritage Map, and their areas were calculated using a GIS platform. Sites were then classified into graded intervals based on area, with the assigned values presented in Table 3.

Vulnerability

Vulnerability measures the sensitivity and adaptive capacity of cultural heritage to climatic disturbances, reflecting the compound property of whether a heritage asset can resist the shock and, if impacted, whether it can recover effectively46,47. In this study, vulnerability is defined as the system’s response capacity, shaped by the interplay of the heritage’s intrinsic attributes, its surrounding natural environment, and the prevailing socio-economic and management conditions.

Intrinsic vulnerability characterizes the responsiveness of cultural heritage to climatic stressors based on its antiquity and material composition. In this study, the construction era of heritage assets is categorized into five chronological stages, ranging from the Pre-Qin period to the modern era. Older heritage sites exhibit greater susceptibility to long-term natural weathering and structural fatigue18,48; therefore, higher vulnerability values are assigned to earlier periods (Table 4). This chronological classification reflects the general trend of deterioration. Given variations in architectural styles, material systems, and construction techniques, significant differences in vulnerability may exist among individual cultural heritage sites.

The materials used in cultural heritage can be broadly categorized into five types: timber, stone, brick, earth, and concrete. Due to differences in water resistance, pore structure, and other intrinsic properties, the climate-induced vulnerability of these materials varies significantly. Water resistance decreases in the following order: stone, concrete, brick, timber, and earth, indicating that stone and concrete are generally more resistant to precipitation-induced soaking damage than timber and earth. The porosity of timber and earth typically exceeds that of brick, concrete, and stone, making wood and earth more susceptible to moisture-related environmental stressors. Furthermore, the organic nature of timber renders it prone to fungal degradation, while the calcium-rich composition of stone increases its vulnerability to bio-corrosion by calciphilic organisms49. For earthen sites, biological damage is primarily caused by the mechanical disruption from plant root growth on ancient structures. A comprehensive assessment indicates that rainfall vulnerability decreases approximately in the sequence: earth, timber, brick, stone, and concrete50. The existing literature50 analyzed the material composition of 1500 national and provincial-level cultural heritage sites, calculated the proportion of each material within different heritage categories, and derived corresponding material vulnerability levels for heritage types. This study adopts their findings to assign material-based vulnerability scores to cultural heritage assets, as presented in Table 5. Overall, the sites of ancient culture are predominantly composed of earth materials, which confer the highest level of vulnerability. In contrast, modern and contemporary buildings are primarily constructed using more durable materials such as brick, stone, and concrete, resulting in relatively lower vulnerability. Ancient tombs contain a significant proportion of earthen structural components, making them more vulnerable than ancient architectural structures built from wood and brick.

Environmental vulnerability refers to the degree to which natural geographic conditions surrounding cultural heritage sites either exacerbate or mitigate the impacts of climate-related events51,52. To assess this vulnerability, this study integrates four key indicators into the modeling framework: slope, distance to geological hazard points, vegetation cover as measured by the Normalized Vegetation Index (NDVI), and land-use type.

Topography is a critical determinant of environmental stability for cultural heritage. Among its parameters, slope provides a direct measure of terrain steepness and is strongly associated with the risk of landslides and debris flows, making it an essential indicator of vulnerability in the modeling framework53,54. This study employs 30-meter resolution digital elevation model (DEM) data to compute the slope distribution across Beijing. Vegetation cover, quantified by the NDVI, serves as a measure of ecosystem resilience; higher NDVI values indicate greater environmental regulatory capacity and are associated with reduced vulnerability, thus classifying it as a negative vulnerability indicator. Land use types serve as proxies for biodiversity and the intensity of biological activities. Microbial activity can induce physical structural changes in cultural heritage, including expansion, contraction, and cracking, and can also promote chemical processes such as corrosion and mold formation55. Similarly, animal activity poses significant threats to cultural heritage. Overall, regions with higher biological activity exhibit greater environmental vulnerability. Based on this relationship, vulnerability weights are assigned to different land use categories, as summarized in Table 6. Forests are assigned the highest weight, followed by grasslands and water bodies, whereas urban, industrial, and residential land receive the lowest weight.

The distance to geological hazard sources is used to assess the spatial pressure imposed by debris flows, landslides, ground subsidence, and collapses on cultural heritage sites. To quantitatively evaluate the geological hazard risks facing heritage assets, this study employs GIS-based spatial analysis to calculate the distance from each heritage site to the nearest hazard location. The shorter the distance to geological hazard sources, the higher the vulnerability value assigned to cultural heritage sites.

Socioeconomic vulnerability reflects spatial disparities in heritage protection resources and management capacity, primarily indicated by the potential for fiscal and human resource investment56. Fiscal revenue serves as a critical enabler for the routine maintenance and disaster response of cultural heritage, ensuring the sustained implementation of conservation measures and timely intervention during emergencies. Significant disparities exist in fiscal revenue levels across Beijing’s districts. In 2023, the public budget revenues of Chaoyang, Haidian, and Xicheng Districts exceeded 40 billion yuan, whereas those of Yanqing, Mentougou, Huairou, Pinggu, and Miyun remained below 4.5 billion yuan. For the large number of district-level and district-wide surveyed cultural heritage sites, higher revenue districts are able to allocate substantially greater resources toward maintenance, restoration, and disaster prevention. In contrast, lower-revenue districts face persistent underfunding in heritage conservation, leading to limited protective capacity and heightened vulnerability.

The 3-kilometer buffer distance corresponds to the typical spatial extent of a rural settlement in the Beijing context, reflecting the scale at which local communities engage in daily activities and interact with their surrounding environment. When a heritage site is located far from human settlements such as villages, the absence of nearby residents restricts regular monitoring of its condition, thereby increasing its susceptibility to undetected deterioration or intentional damage. Conversely, higher population density within the same radius facilitates sustained on-site observation and informal protection through active community involvement. In Beijing, the population distribution is highly uneven, with dense concentrations in the urban core and southeastern plain regions, while the northwest mountainous areas are sparsely populated. As a result, cultural heritage sites in these peripheral, low-density areas are more likely to be socially marginalized and receive insufficient institutional or public attention, leading to elevated levels of vulnerability.