Preamble

ARID LAND GEOGRAPHY Vol. 48 No. 8 Aug. 2025

Comprehensive Assessment and Regionalization of Social Vulnerability for Natural Disasters in Gansu Province

YU Han¹,², MENG Zhihua³, WANG Jing'ai²

¹ School of Agriculture and Forestry Economics and Management, Lanzhou University of Finance and Economics, Lanzhou 730101, Gansu, China
² Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China
³ School of Accountancy, Lanzhou University of Finance and Economics, Lanzhou 730101, Gansu, China

Abstract

Reducing social vulnerability is crucial for natural disaster risk prevention and mitigation. At the county-level scale, this study constructs a social vulnerability assessment index system tailored to Gansu Province, focusing on three disaster-bearing bodies: population, economy, and crop farming. We evaluate regional social vulnerability to natural disasters in Gansu, including exposure, sensitivity, adaptability, and comprehensive social vulnerability. Based on the paradigm of natural geographic regionalization, we systematically develop a comprehensive regionalization scheme for natural disaster social vulnerability at the county level. The results demonstrate: (1) The comprehensive social vulnerability index in Gansu Province generally exhibits a macro-pattern of high in the east and low in the west, high in the south and low in the north. High-vulnerability areas are mainly concentrated in eastern, central, and southern Gansu, showing clustering characteristics in densely populated disaster-bearing areas such as cities and adjacent counties. (2) The comprehensive regionalization includes two levels: The first level comprises four dominant natural disaster type zones, including the sandstorm-dominant disaster zone in the western Hexi Corridor, the drought-dominant disaster zone in Lanzhou, the rainstorm-flood-landslide/debris flow-dominant disaster zone in southern Gansu, and the multi-hazard disaster zone covering the central-eastern Hexi Corridor and the Gannan Tibetan Autonomous Prefecture. The second level includes 14 comprehensive vulnerability level zones with different structures. This regionalization scheme systematically expresses the macro-spatial differentiation patterns of social vulnerability structures under regionally dominant natural disaster types, which can serve diverse regional differences and needs in reducing social vulnerability to natural disasters.

Keywords: social vulnerability; natural disaster; comprehensive regionalization; arid region; Gansu Province

1.1 Research Area Overview

Gansu Province is located at the geographic center of China's land territory, extending in a northwest-southeast orientation (1655 km long, 32°11′~42°57′N, 92°13′~108°46′E) with an average north-south width of 530 km. The province lies at the intersection of the Tibetan Plateau, Loess Plateau, and Inner Mongolia Plateau, belonging to three major water systems (Yangtze River, Yellow River, and inland rivers) and spanning three climatic zones (eastern monsoon region, northwest arid/semi-arid region, and Qinghai-Tibet cold region). The diversity and complexity of the natural environment are prominent, resulting in significant spatiotemporal variation in disaster-forming environments, hazard factors, and disaster-bearing body elements across the province's natural disaster system.

By the end of 2021, Gansu's total population was approximately 24.9 million, with a provincial area of 425,800 km², making it a typical region with vast territory and sparse population. Population distribution is characterized by dense southeast and sparse northwest, with population, economy, and other elements in the northwestern part of the province concentrated in oases and other favorable locations, showing severe imbalance. These factors collectively contribute to significant regional differentiation in natural disaster vulnerability across Gansu, making comprehensive social vulnerability regionalization essential for systematically understanding these differentiation patterns and providing scientific basis for regionally tailored vulnerability reduction strategies.

Considering the complex and diverse natural disaster systems in Gansu and the significant disaster chain phenomena, this study categorizes multiple natural disasters according to disaster chain types. Based on the China Natural Disaster Newspaper Database, which records actual disaster events, we unified drought and aridity into a single drought category, and combined landslides and debris flows triggered by rainstorms into the rainstorm-flood category. The final regionalization includes five disaster types: earthquake, drought, rainstorm-flood-landslide/debris flow, and sandstorm, which together account for approximately 89.6% of total disaster occurrences. The disaster-bearing bodies selected for this study are population, economy, and crop farming, which are most affected by these disaster types.

1.2 Data Sources

The basic research units are 87 counties and districts in Gansu Province. Population data were obtained from the Seventh National Population Census Bulletin. Economic and crop farming-related data primarily came from the Gansu Statistical Yearbook and China County Statistical Yearbook, with missing data supplemented by prefecture-level city statistical yearbooks and county government bulletins. Due to significant data fluctuations in 2020 caused by the COVID-19 pandemic, 2019 data were used as substitutes. Regional disaster data were derived from the China Provincial Newspaper Disaster Database provided by the Regional Geography Laboratory of Beijing Normal University. This database was established by reviewing disaster-related news reports from 18 national provincial newspapers, classifying and summarizing information including disaster start/end times, types, and severity, covering 28 natural disaster types. The database records actual disaster events and reflects regional natural disaster exposure levels. Land use data were obtained from the GlobeLand30 global land cover data released by the Ministry of Natural Resources (Table 1).

1.3.1 Natural Disaster Social Vulnerability Assessment

Index System Construction. Referencing existing research and considering data availability, we established a county-level social vulnerability assessment index system for natural disasters in Gansu based on three core dimensions: exposure, sensitivity, and adaptability (Table 2). Exposure includes population exposure, economic exposure, crop farming exposure, and historical disaster exposure. Sensitivity and adaptability are divided into population, economic, and crop farming sub-dimensions.

Weight Calculation. This study employs the entropy method for objective weighting, using information entropy as the criterion for weight evaluation to eliminate subjectivity. The general procedure involves: (1) Data standardization to eliminate differences in measurement units and scales (forward indicators use formula (1), negative indicators use formula (2)); (2) Calculation of the proportion of each county's indicator value to the total for each indicator (formula (3)); (3) Calculation of information entropy values for each indicator (formula (4)); and (4) Calculation of weights for each indicator (formula (5)).

$$
Y_{ij} = \frac{x_{ij} - \min(x_j)}{\max(x_j) - \min(x_j)} \quad (1)
$$

$$
Y_{ij} = \frac{\max(x_j) - x_{ij}}{\max(x_j) - \min(x_j)} \quad (2)
$$

$$
P_{ij} = \frac{Y_{ij}}{\sum_{i=1}^{n} Y_{ij}} \quad (3)
$$

$$
E_j = -\frac{1}{\ln(n)} \sum_{i=1}^{n} P_{ij} \ln(P_{ij}) \quad (4)
$$

$$
W_j = \frac{D_j}{\sum_{j=1}^{m} D_j}, \quad D_j = 1 - E_j \quad (5)
$$

Where $Y_{ij}$ is the standardized value; $x_{ij}$ is the actual indicator value; $\min(x_j)$ and $\max(x_j)$ are the minimum and maximum values for each indicator column; $P_{ij}$ is the percentage of indicator value for county $i$ under indicator $j$; $E_j$ is the information entropy value; $W_j$ is the weight; $D_j$ is the information entropy redundancy; $n$ is the number of records (total counties in Gansu); and $m$ is the number of indicators.

Comprehensive Vulnerability Calculation Model. Based on the assessment framework, we calculate social vulnerability and its dimensions for Gansu's counties using formulas (6)-(9):

$$
SoVI = EI + SI - AI \quad (6)
$$

$$
EI = \sum_{j=1}^{k} W_j Y_{ij} \quad (7)
$$

$$
SI = \sum_{j=k+1}^{k+l} W_j Y_{ij} \quad (8)
$$

$$
AI = -\sum_{j=k+l+1}^{m} W_j Y_{ij} \quad (9)
$$

Where $SoVI$ is the comprehensive social vulnerability index; $EI$, $SI$, and $AI$ are exposure, sensitivity, and adaptability indices, respectively; $m$ is the total number of indicators; $k$, $l$, and $s$ are the numbers of exposure, sensitivity, and adaptability indicators, respectively. $EI$ and $SI$ together express potential vulnerability, while $AI$ has an offsetting effect. The three components constitute the county-level regional natural disaster social vulnerability index. Adaptability indicators were processed as negative indicators during normalization, so the adaptability index should take negative values when expressed separately.

1.3.2 Comprehensive Regionalization of Natural Disaster Social Vulnerability

Regionalization Purpose and Principles. The primary purpose is to systematically reveal macro-regional differentiation patterns of natural disaster social vulnerability at the county level in Gansu, providing scientific support for disaster prevention and risk management, insurance market layout, pricing, and risk control. Based on natural disaster regionalization practices and social vulnerability differentiation mechanisms, we establish the following principles: (1) Highlight multi-dimensional characteristics of regional natural disaster social vulnerability, expressing not only comprehensive vulnerability levels but also regional differentiation patterns of constituent elements; (2) Maintain county-level administrative unit integrity following the regional conjugacy principle to ensure unified application within administrative regions; (3) Employ bottom-up merging combined with top-down division, using high-resolution spatial data and multiple methods to construct a comprehensive quantitative regionalization model.

Regionalization Levels and Index System. Considering Gansu's natural disaster system characteristics, spatiotemporal scales, regional features, and data availability, we adopt a two-level system: Level I zones are dominant natural disaster type zones, using historical disaster occurrence frequency as the indicator to reflect macro-spatial differentiation of specific disaster types. Level II zones are comprehensive social vulnerability level zones, using exposure, sensitivity, and adaptability indices as indicators to reflect structural differences in social vulnerability under specific dominant disaster backgrounds.

Regionalization Process and Methods. The process involves: (1) Specifying regionalization principles according to regional and data characteristics; (2) Determining preliminary zoning schemes using hierarchical classification and multi-indicator k-means clustering for bottom-up merging to obtain reasonable zone numbers and boundaries; (3) Finalizing comprehensive zone boundaries by overlaying Level I and Level II zones, supplemented by top-down systematic spatial division, with fine adjustments based on regional geographic features, including enclave and fragmented patch merging, to produce the final regionalization scheme with naming and characteristic parameter statistics.

2.1 Spatial Differentiation Pattern of Social Vulnerability Index

We calculated exposure, sensitivity, adaptability, and comprehensive vulnerability indices for all 87 county units using natural breaks classification. The results are shown in Figure 1.

2.1.1 Differentiation Patterns. The exposure index shows a macro-pattern of high in the east and low in the west, high in the south and low in the north. Highest exposure concentrates in prefecture-level city urban areas, followed by surrounding counties. This reflects the concentration of population, economy, and crop farming in cities and their peripheries with better natural conditions in northwest arid/semi-arid regions. Medium exposure areas are mainly in counties surrounding cities and along the Hexi Corridor. Although western Gansu has vast territory with sparse population, disaster-bearing bodies concentrate in oases, resulting in medium exposure levels. Low exposure areas are concentrated along the northern edge of the Qaidam Basin and western Gannan Plateau, where population, economy, and crop farming are all small-scale.

The sensitivity index also shows high east/low west, high south/low north patterns. High sensitivity areas concentrate in northern Tianshui City, the region with Gansu's largest population and crop farming proportion. Sub-high sensitivity areas show contiguous distribution in Longnan City and scattered distribution in central Gansu and the Hexi Corridor. This is because Longnan has favorable natural conditions with relatively large population and crop farming proportions, while medium sensitivity areas along the Hexi Corridor and in central/eastern Gansu correspond to their medium-level population and crop farming proportions. Low sensitivity areas are mainly along the Qilian Mountains in the Hexi Corridor and the Gannan Plateau, where population, economy, and crop farming scales are relatively small.

The adaptability index shows an overall pattern of high in the west and low in the east, high in the north and low in the south. High values concentrate in the central-western Hexi Corridor and Baiyin City urban area. The central-western Hexi Corridor has much smaller population, agriculture, and economic scales compared to other regions, resulting in the highest per capita resource allocation for adaptability improvement. Lanzhou and Baiyin urban areas have higher urbanization levels and greater resource investment for adaptability enhancement under the "strong provincial capital" development model. Medium adaptability areas concentrate in Dingxi, southern Tianshui, and Longnan, where large population-economic scales and crop farming-dominant agriculture result in fewer per capita adaptability resources than high-value areas. Low adaptability areas are mainly in Gannan Tibetan Autonomous Prefecture, eastern Dingxi, and northern Qingyang, where relatively backward population and economic development result in smaller per capita adaptability resource investment.

The comprehensive vulnerability index shows high east/low west, high south/low north patterns. High-vulnerability areas mainly concentrate in eastern, central, and southern Gansu, clustering around densely populated disaster-bearing areas like cities. This results from high exposure and sensitivity but insufficient per capita adaptability resources due to large population-economic scales. Medium-vulnerability areas are mainly along the Hexi Corridor and in Linxia Hui Autonomous Prefecture and northern Lanzhou, corresponding to medium exposure and medium-high adaptability. Low-vulnerability areas are concentrated along the Qilian Mountains and Gannan Plateau, where small-scale disaster-bearing bodies result in low exposure and sensitivity but relatively high adaptability.

2.1.2 Regional Differentiation Characteristics. Gansu's natural disaster social vulnerability shows distinct regional differentiation: First, counties in western Gansu with vast territory and sparse population have small-scale population and economy with relatively concentrated distribution patterns, clearly differing from central and eastern regions. Statistics on area proportions of different vulnerability levels (Table 3) show that western counties contribute large proportions to low exposure, low sensitivity, high adaptability, and low vulnerability categories. Their distribution patterns indicate that direct use of county area for area-average indicators is inappropriate for these sparsely populated western counties.

Second, structural differences in comprehensive vulnerability formation show clear regional variations in the combination characteristics of exposure, sensitivity, and adaptability levels. This suggests that Gansu's comprehensive vulnerability element composition differences can be roughly divided into four regions: the Hexi Corridor shows medium exposure, medium sensitivity, and high adaptability; the central Loess Plateau shows high exposure and high sensitivity; the Gannan Plateau shows low exposure, low sensitivity, and low adaptability; and the southeastern mountainous region shows high exposure, high sensitivity, and low adaptability.

2.2.1 Level I Zoning Scheme

Level I zones are dominant natural disaster type zones. We counted disaster occurrences by county and performed spatial k-means clustering on the occurrence data. Based on preliminary research, we set the cluster number to 4. After obtaining preliminary results and referring to regionalization principles and disaster diversity indices, we adjusted fragmented patches to obtain Level I zones (Figure 2): (I) Western Hexi Corridor sandstorm-dominant disaster zone, accounting for 40.98% of Gansu's total area. This region is deep inland with extensive deserts and gobi, where sandstorm disaster impacts far exceed those of earthquakes, rainstorm-floods, and droughts. (II) Lanzhou drought-dominant disaster zone, accounting for 3.04% of the area. This zone has highly concentrated population, economy, and crop farming disaster-bearing bodies affected by multiple disasters, with drought impacts being relatively prominent. (III) Southern Gansu rainstorm-flood-landslide/debris flow-dominant disaster zone, accounting for 7.54% of the area. This region belongs to the Yangtze River basin and, influenced by the southeast monsoon and mountainous terrain, shows significantly higher occurrences of rainstorm-floods and secondary landslides/debris flows than other disaster types. (IV) Central-eastern Hexi Corridor and Gannan Tibetan Autonomous Prefecture multi-hazard disaster zone, accounting for 48.43% of the area. This large zone experiences diverse natural disasters including earthquakes, sandstorms, droughts, and rainstorm-floods with secondary landslides/debris flows, with no single dominant disaster type, though earthquake occurrences are slightly higher than in other zones.

2.2.2 Level II Zoning Scheme

Level II zones are social vulnerability level zones. For application convenience and reasonable zone numbers, we first classified comprehensive vulnerability indices into high, medium, and low categories using natural breaks. We then performed spatial clustering using exposure, sensitivity, and adaptability indices as indicators to obtain preliminary zones, which were fine-adjusted under vulnerability level constraints to form high, medium, and low vulnerability zones with different structural characteristics. The adjusted Level II zones can be roughly divided into 7 regions (Figure 3): 4 high-vulnerability sub-zones including Lanzhou (urban area), Wuwei, Qingyang, and the Dingxi-Longnan region; 3 medium-vulnerability sub-zones mainly covering the Hexi Corridor-central Gansu, Pingliang, and eastern Longnan regions; and 2 low-vulnerability sub-zones including the western Qilian Mountains and Gannan Tibetan Autonomous Prefecture region.

2.2.3 Comprehensive Regionalization Scheme

By overlaying Level I and Level II zones and performing fine adjustments according to regionalization principles, spatial proximity, and regional geographic features (including enclave and fragmented patch merging), we obtained the final comprehensive vulnerability regionalization scheme (Figure 4). The naming principle combines place names (counties, geographic units) with dominant disaster types for Level I, and place names with vulnerability levels for Level II. Zone codes adopt a structure of Level I type (Roman numerals I-IV) + Level II vulnerability level (A-C for low-medium-high) (Table 4). Original key data from county or raster units were averaged by new zone units to create zone characteristic parameter tables.

To facilitate practical application in reducing social vulnerability, we expressed the structural characteristics of the three vulnerability components as averages using bar charts (Figure 5). This visualization intuitively displays the systematic element differentiation patterns of Gansu's natural disaster social vulnerability. The regionalization scheme expresses not only comprehensive vulnerability levels but also structural differences in constituent elements, revealing that vulnerability reduction priorities should differ by zone, thereby providing scientific basis for zone-specific countermeasures.

Western Hexi Corridor Region shows overall low vulnerability with high adaptability and low exposure, making vulnerability level primarily dependent on sensitivity. Therefore, the key to reducing vulnerability in this zone lies in effectively decreasing sensitivity.

Central-eastern Hexi Corridor and Lanzhou surrounding areas are mainly medium-vulnerability zones characterized by relatively high sensitivity and medium exposure and adaptability. As these regions are affected by multiple natural disasters, vulnerability reduction should focus on decreasing sensitivity while also reducing exposure to multiple hazards and improving adaptability.

Eastern Loess Plateau and Longnan regions show high overall vulnerability with high sensitivity and low adaptability. Therefore, when facing multiple natural disasters in the eastern Loess Plateau and rainstorm-flood-landslide/debris flow hazards in Longnan, these regions should further invest in disaster prevention and mitigation resources, prioritizing adaptability improvement while simultaneously reducing sensitivity.

3 Discussion

Potential Inappropriateness of Area-Average Indicators. This study used cultivated land and artificial land as benchmarks for some area-average indicators to avoid errors from large uninhabited areas in western Gansu. However, for densely populated eastern Gansu, disaster-bearing bodies like population and economy are not limited to cultivated and artificial lands, so our area-average calculation method may introduce some errors. Future research should consider dividing Gansu into eastern, central, and western sub-regions and calculating area-average indicators separately according to disaster-bearing body distribution characteristics in each region.

Distinction Between Potential and Actual Exposure. Exposure can be divided into potential exposure (disaster-bearing body quantity reflecting possible hazard impact scale) and actual exposure (the portion of potential exposure truly impacted after accounting for protection measures). This study considered multiple natural disasters but could not obtain data on protection measures and their quantities. Additionally, the provincial newspaper disaster database mainly derives from news reports with insufficient detailed disaster data to express the actual scale of disaster-bearing systems after removing protective measures (e.g., irrigated areas in crop farming are not significantly impacted by drought). Therefore, our exposure indicators represent potential exposure only. Future research should further distinguish actual exposure within potential exposure as data and methods improve.

4 Conclusion

1) Gansu's natural disaster social vulnerability shows a macro-pattern of high in the east and low in the west, high in the south and low in the north. High-vulnerability areas concentrate in eastern, central, and southern Gansu, clustering around densely populated disaster-bearing areas like cities. This results from high exposure and sensitivity levels but relatively insufficient per capita adaptability resource investment due to large population-economic scales. Low-vulnerability areas are mainly distributed along the Qilian Mountains and in Gannan Tibetan Autonomous Prefecture counties, where small-scale disaster-bearing bodies create low exposure and sensitivity but relatively high per capita adaptability resource investment.

2) Dominant natural disasters show significant spatial differentiation across Gansu regions. The Level I regionalization indicates four dominant disaster type zones: western Hexi Corridor sandstorm-dominant, Lanzhou drought-dominant, southern Gansu rainstorm-flood-landslide/debris flow-dominant, and central-eastern Hexi Corridor and Gannan multi-hazard zones.

3) Different zones should prioritize different approaches to reducing regional social vulnerability and improving comprehensive risk prevention. The 14 Level II zones indicate differentiation in vulnerability levels and element structures: the western Hexi Corridor shows low exposure and high adaptability with vulnerability depending mainly on sensitivity; central-eastern Hexi Corridor and Lanzhou surrounding areas are medium-vulnerability zones with high sensitivity; eastern Gansu and Longnan regions are high-vulnerability zones formed by high sensitivity and low adaptability.

References

[1] National Climate Center, China Meteorological Administration. Blue Book on Climate Change in China 2024[M]. Beijing: Science Press, 2024: 1-10.

[2] Shi Peijun. Vigorously enhance the ability to prevent and control systemic risks of global change[J]. China Journal of Disaster Reduction, 2024(1): 14-15.

[3] Shi Peijun. Theory and practice on disaster system research in a fourth time[J]. Journal of Natural Disasters, 2005, 14(6): 1-7.

[4] Zhou Yao, Wang Jing'ai. A review on development of vulnerability curve of natural disaster[J]. Advances in Earth Science, 2012, 27(4): 435-442.

[5] Huang Xiaojun, Huang Xin, Cui Cailan, et al. The concept, analytical framework and assessment method of social vulnerability[J]. Progress in Geography, 2014, 33(11): 1512-1525.

[6] Li Di, He Kai, Sun Jing, et al. Research on social vulnerability assessment from the perspective of meteorological disaster process[J]. Journal of Inner Mongolia Normal University (Natural Science Edition), 2023, 52(4): 374-381.

[7] Yuan Dinghuan, Luo Junyue, Guo Jun, et al. Vulnerability and natural disaster loss analysis: Taking Guangdong Province from 2011 to 2020 as an example[J]. Journal of Catastrophology, 2024, 39(3): 94-101.

[8] Wang Jing'ai, Shang Yanrui, Wu Yaoyao, et al. Integrated risk governance of regional agricultural drought: From vulnerability, resilience, adaptability to consilience[J]. Journal of Hebei Normal University (Natural Science Edition), 2021, 45(4): 400-409.

[9] Shi Peijun, Sun Shao, Wang Ming, et al. Climate change regionalization in China (1961—2010)[J]. Science China: Earth Sciences, 2014, 44(10): 2294-2306.

[10] Cutter S L, Mitchell J T, Scott M S. Revealing the vulnerability of people and places: A case study of Georgetown County, South Carolina[J]. Annals of the Association of American Geographers, 2000, 90(4): 713-737.

[11] Zhou Y, Li N, Wu W X, et al. Assessment of provincial social vulnerability to natural disasters in China[J]. Natural Hazards, 2014, 71(3): 2165-2186.

[12] Wang Jun, Chen Jingye, Gong Qinghua, et al. Urban social vulnerability assessment and optimization strategies from the perspective of resilience: A case study of the Pearl River Delta urban agglomeration[J]. Tropical Geography, 2023, 43(3): 474-483.

[13] Zhao Ruidong, Fang Chuanglin, Liu Haimeng. Progress and prospect of urban resilience research[J]. Progress in Geography, 2020, 39(10): 1717-1731.

[14] Zeng Bing. Evaluating the evolution patterns of provincial economic resilience in China[J]. Journal of Dalian University of Technology (Social Science Edition), 2023, 44(2): 28-36.

[15] Ning Jing, Zhu Ran, Zhang Xinyuan, et al. Evaluation and analysis of urban resilience of districts and counties in Inner Mongolia[J]. Arid Land Geography, 2023, 46(7): 1217-1226.

[16] Zhang Zhen, Zhang Yichen, Zhang Jiquan, et al. Urban resilience assessment based on entropy weight method and TOPSIS model: Take Changchun City as an example[J]. Journal of Catastrophology, 2023, 38(1): 213-219.

[17] Bai Limin, Xiu Chunliang, Feng Xinghua, et al. A comprehensive assessment of urban resilience and its spatial differentiation in China[J]. World Regional Studies, 2019, 28(6): 77-87.

[18] Wu Shaohong, Pan Tao, Liu Yanhua, et al. Comprehensive climate change risk regionalization of China[J]. Acta Geographical Sinica, 2017, 72(1): 3-17.

[19] Wang Jiwei, Wang Jun, Ye Tao, et al. Integrated regionalization of crop natural disaster insurance: A case study on late rice in Hunan Province, China[J]. Journal of Natural Disasters, 2016, 25(3): 1-10.

[20] Ling Mingfang, Xie Baopeng, Chen Ying, et al. Spatial and temporal patterns of vulnerability of rural systems in Gansu Province[J]. Journal of Arid Land Resources and Environment, 2024, 38(2): 112-122.

[21] Li Feifei, Zhou Xia, Zhou Yuxi. Vulnerability assessment and spatiotemporal distribution of agricultural drought in northwest China[J]. Arid Zone Research, 2023, 40(4): 663-669.

[22] Zhou Su, Zhang Mingjun, Wang Shengjie, et al. Assessment of vulnerability in natural social system in Hexi, Gansu[J]. Resources Science, 2018, 40(2): 452-462.

[23] Li Jie, Zhao Ruifeng, Xie Zuolun. The comprehensive assessment of vulnerability in social-ecological system of Gansu Province[J]. Economic Geography, 2015, 35(12): 168-175.

[24] Ye Wenli, Yang Xinjun, Wu Kongsen, et al. Spatiotemporal characteristics and influencing factors of social-ecological system resilience in the Loess Plateau[J]. Arid Land Geography, 2022, 45(3): 912-924.

[25] Wang Jianing, Xue Dongqian, Ma Beibei, et al. Vulnerability and population response of mineral resource-based cities in the Loess Plateau[J]. Arid Land Geography, 2020, 43(6): 1679-1690.

[26] Zheng Du, Ou Yang, Zhou Chenghu. Understanding of and thinking over geographical regionalization[J]. Acta Geographical Sinica, 2008, 63(6): 563-573.

[27] Shi Peijun, Lü Lili, Wang Ming, et al. Disaster system: Disaster cluster, disaster chain and disaster compound[J]. Journal of Natural Disaster, 2014, 23(6): 1-12.

[28] Jia Huicong, Wang Jing'ai, Yang Yang, et al. Analysis of natural disaster chain in northwest China[J]. Journal of Catastrophology, 2016, 31(1): 72-77.

[29] Office of the Leading Group of the State Council for the Seventh National Population Census. Tabulation on 2020 China Population Census by County[M]. Beijing: China Statistics Press, 2022: 3-983.

[30] Wang Nai'ang, Huang Yinzhou, Luo Dan. Geography of Provinces, Cities, and Districts in China: Gansu Geography[M]. Beijing: Beijing Normal University Press, 2020: 3-123.

[31] Shi Peijun, Wang Jiwei, Zhang Gangfeng, et al. Research review and prospects of natural disasters regionalization in China[J]. Geographical Research, 2017, 36(8): 1401-1414.

[32] Fang Jiayi, Chen Wenfang, Kong Feng, et al. Measuring social vulnerability to natural hazards of the coastal areas in China[J]. Journal of Beijing Normal University (Natural Science Edition), 2015, 51(3): 280-286.

[33] Chen Shaohua, Zhang Li, Zuo Hongzhen, et al. Measurement of social vulnerability of rural natural disasters in Hunan Province in 2018[J]. Meteorology and Disaster Reduction Research, 2020, 43(4): 299-306.

[34] Ruan Xinxin, Fu Xiaolin, Hou Jundong, et al. Comprehensive measurement of social vulnerability to natural disasters and characteristics of space-time evolution in Hubei Province[J]. Safety and Environmental Engineering, 2019, 26(2): 52-61.

[35] Zhang M, Xiang W B, Chen M L, et al. Measuring social vulnerability to flood disasters in China[J]. Sustainability, 2018, 10(8): 1-16.

[36] Yu Han, Guo Hao, Wang Jing'ai. Comprehensive regionalization of natural disaster risk: A case study of population and economic risk caused by typhoon in Guangdong Province[J]. Journal of Catastrophology, 2022, 37(2): 131-137.

[37] Ye Tao, Shi Peijun, Wang Jun, et al. Integrated Risk Governance: Agricultural Natural Disaster Risk Assessment and Insurance Regionalization[M]. Beijing: Science Press, 2017: 44-80.