Multi-Scale Assessments and Future Projections of Drought Vulnerability of Social–Ecological Systems: A Case Study from the Three-River Headwaters Region of the Tibetan Plateau

Abstract

The vulnerability of Social–Ecological Systems (SES) is a frontier research topic in the field of geography. Research on drought vulnerability has emerged as a key area of focus in the study of SES vulnerability, and it has increasingly been recognized as a critical step in formulating policies for drought prevention and mitigation. In this study, the indicator system for drought vulnerability evaluation of SES in the Three-River Headwaters Region (TRHR) was established. This paper revealed the drought vulnerability evolution process and characteristics, and key driving indicators of SES at county-town-village spatial scales in six time periods of 1990, 2000, 2010, 2015, 2020, and 2023, and predicted the drought vulnerability of SES in 2050 under two scenarios. Results indicate that the average drought vulnerability in the TRHR decreased from 0.526 in 1990 to 0.444 in 2023. Compared to 1990, among the 82 selected towns, 85.37% experienced a decline in 2023, and among the 152 selected villages, 95.39% showed a reduction in 2023. Hot spots of drought vulnerability were concentrated in the southeast of the TRHR, while cold spots were in the northwest. From 1990 to 2000, the drought vulnerability of counties and towns in the TRHR increased, but it decreased between 2000 and 2023. In 1990, Henan County exhibited the highest drought vulnerability at the county level. Waeryi Town in Jiuzhi County had the highest vulnerability among towns, while Suojia Town in Zhidoi County had the lowest. Of the 152 selected villages, 41.45% exhibited relatively high or high levels of drought vulnerability, while 23.68% showed relatively low levels. In 2023, Jiuzhi County became the most vulnerable county, with Baiyu Town in Jiuzhi County ranking highest among towns and Suojia Town in Zhidoi County remaining the least vulnerable. At the village level, 22.37% exhibited relatively high or high vulnerability, whereas 42.11% showed relatively low or low levels. Drought disaster records, the proportion of agricultural and animal husbandry output value, the proportion of grassland, the proportion of large livestock, and the per capita disposable income surface are the key factors influencing drought vulnerability in the TRHR. By 2050, under the first scenario, the average drought vulnerability of the TRHR is projected to be 0.428, indicating a medium level, while the second scenario predicts a further reduction to 0.350, representing a relatively low level. The adaptive governance strategies to mitigate drought vulnerability in the TRHR include developing an integrated drought management system; establishing an ecological management, protection, and financial support model; and so on. Overall, this paper can provide scientific references and policy recommendations for policymakers and researchers on the aspects of drought vulnerability and sustainable development of SES.

1. Introduction

Achieving sustainable development is recognized as a major global challenge within the United Nations 2030 Agenda for Sustainable Development Goals (SDGs) [1,2]. The research on Social–Ecological Systems (SES) occupies the forefront of sustainability research and also constitutes an important approach and frontier in the comprehensive research of geography and ecology [3]. Vulnerability, as a core concept in SES research, is among the frontier scientific research propositions in the field of geographical science research and one of the key aspects of global change research, contributing to the realization of sustainable development [4,5,6]. Research on drought vulnerability represents the latest hotspot in the study of SES vulnerability and has drawn attention within the geographical community both domestically and abroad [7,8,9,10,11]. Drought vulnerability refers to the state in which multiple elements of nature, society, and economy within an SES are disrupted by drought, thereby influencing the system’s ability to cope with drought. It is the outcome of the interaction within social–ecological coupling systems. Research on drought vulnerability has gradually emerged as the primary element in policy formulation for drought prevention and mitigation [12,13,14,15].

The SES theory, originating from ecology [16], is one of the most influential theories in interdisciplinary research concerning the human-environment relationship. It emphasizes revealing the process and form of the nonlinear evolution of system elements and their structural configurations, thereby providing novel perspectives and ideas for the study of the evolution of the territorial human-environment relationship system [17,18,19,20]. Conducting research on SES is regarded as an essential means for science to achieve sustainable development [20,21,22,23]. With the introduction of global sustainable development goals, key concepts of SES, such as vulnerability, resilience, and adaptability, have become integral to in-depth analysis and exploration within sustainability science, offering references for the direction and approach of sustainable development [20,24]. The SES vulnerability analysis framework proposed by representative scholars like Turner has been widely adopted [7,25,26]. This framework encompasses components such as exposure, sensitivity, and resilience within the context of human activities, the natural environment, and their interactions, and explores the interaction of various elements at three scales: global, regional, and local [25,27]. Drought and its associated conditions are issues of global significance. Compared to other extreme events or natural disasters, drought is characterized by slow development, long duration, and extensive impact range [28,29]. Drought vulnerability research has recently emerged as a research hotspot, focusing on the attribute of SES being susceptible to drought losses and impacts due to constraints from various elements including nature, society, and economy [7]. Currently, research on drought vulnerability of SES has been conducted across multiple scales, including global, continental, national, regional, and local [30,31,32,33,34]. For instance, some scholars utilize per capita GDP, cultivated land reclamation rate, total per capita renewable energy, and per capita water resources to evaluate global drought vulnerability [35]. Other scholars calculate the drought vulnerability index of Turkey by using the mean value of four standardized indicators, namely irrigated land, total agricultural land, population density, and municipal water use [14]. Five indicators, including climate, topography, waterway density, land use, and groundwater resources, are selected to evaluate drought vulnerability in Iran [36]. Population density, agricultural land area, urban water demand, agricultural water demand, and industrial water demand are employed to analyze drought vulnerability in South Korea [37]. Parameters such as NDVI, rainfall, slope, seepage zone, soil depth, land use and land cover change, catchment structure, geomorphology, drainage density, and groundwater level fluctuation are considered for analyzing drought vulnerability in western India [11]. Additionally, scholars select indicators for characterization based on three aspects: exposure, sensitivity, and adaptability. For example, indicators such as drought frequency are commonly used to characterize exposure, indicators related to the area of agricultural and pastoral land are often used to represent sensitivity and indicators like GDP and drought resistance management plans are frequently used to signify adaptability [38,39]. The drought vulnerability of SES is influenced by the interaction of the natural environment and social–economic elements, with greater dependence on exposure, sensitivity, and adaptive capacity. Based on the analysis of influencing factors, scholars have proposed countermeasures and suggestions to reduce the drought vulnerability of SES, such as strengthening risk management measures, adjusting land use structure, and rationally allocating water resources [10,35].

In summary, the aforementioned scholars have provided theoretical support and methodological guidance for this study. However, drought vulnerability remains a complex and interdisciplinary matter that demands more in-depth research. Firstly, current research on drought vulnerability of SES at local scales, such as towns and villages, is relatively limited. Secondly, multi-scale assessments can comprehensively and systematically analyze the overall characteristics of drought vulnerability of SES in the TRHR. At the same time, multi-scale assessment can capture the evolution characteristics of drought vulnerability at different scales and discover local features, patterns, demands, and disturbances, which helps to understand the diversity and complexity of SES, improve the reliability and accuracy of drought vulnerability assessment, and enhance the scientific rigor and effectiveness of policy making. Conducting research on the drought vulnerability of SES across multiple temporal and spatial scales is an area that requires attention in subsequent drought vulnerability research. Thirdly, there is a scarcity of research on formulating criterion decisions from the perspective of adaptive governance and subsequently simulating future drought vulnerability scenarios.

The Three-River Headwaters Region (TRHR), located in the hinterland of the Tibetan Plateau, is the source place of the Yangtze River, Yellow River, and Lancang River. It encompasses various types of ecosystems, including mountains, waters, forests, fields, lakes, grasslands, and deserts. This region is one of the areas with the highest concentration of high-altitude biodiversity in the world. It also serves as a crucial area for ensuring national ecological security and water conservation. Moreover, it is an ecologically fragile area and a climate change-sensitive area. In October 2021, the TRHR National Park was officially established. Currently, the ecological situation in the TRHR has exhibited a trend of initial containment and partial improvement [40,41,42,43], yet it still confronts the challenge posed by drought [44]. In the face of drought and its development trend in the TRHR, it is of utmost urgency to conduct research on the vulnerability of SES in relation to their ability to cope with drought.

Therefore, the research objectives of this paper are as follows: (1) To construct a drought vulnerability evaluation indicator system from the three dimensions of exposure, sensitivity, and adaptive capacity. (2) To investigate the spatiotemporal evolution characteristics of drought vulnerability in TRHR in six time periods of 1990, 2000, 2010, 2015, 2020, and 2023, considering three spatial scales of drought vulnerability: county-level, town-level, and village-level. (3) To examine the key influencing factors of drought vulnerability in TRHR, thereby providing reference experience for exploring effective paths to reduce drought vulnerability. (4) To predict the drought vulnerability of TRHR in 2050 and provide reference experience for future drought vulnerability research aimed at sustainable development and adaptive governance.

The remainder of this paper is structured as follows. Section 2 presents the study area, the construction of the indicator system, data sources, and research methods. In Section 3, the spatiotemporal differentiation characteristics of drought vulnerability are measured and analyzed, the drivers of drought vulnerability are identified, and the future drought vulnerability is predicted. The discussion and conclusions of the paper are presented in Section 4 and Section 5, respectively.

2. Materials and Methods

2.1. Study Area

This study chooses the TRHR on the Tibetan Plateau of China as the study area (Figure 1). The TRHR serves as a catchment in the source region of the Yangtze River, Yellow River, and Lancang River, constituting a crucial water conservation area within China. It encompasses a total area of approximately 39.5 × 10⁴ km², incorporating 4 Tibetan autonomous prefectures, 21 counties, and 161 towns, with an average altitude exceeding 4000 m. Based on the data retrieved from the official website of the Qinghai Meteorological Bureau (http://qh.cma.gov.cn/, accessed on 8 November 2024), the annual precipitation in the TRHR is 404.0 mm, predominantly concentrated in the period from June to August. The annual mean temperature of the TRHR ranges between −5.4 °C and 7.5 °C, and the annual sunshine hours fall within the interval of 2300 to 2900 h.

Presently, the TRHR is encountering the challenges posed by drought. To commence with, drought events frequently transpire in the TRHR. For instance, the combination of high temperatures and scarce rainfall in July 2017 gave rise to a meteorological drought across the entire TRHR, with the core area attaining a state of severe meteorological drought [45]. Moreover, the drought situation in the eastern and southern TRHR has been exacerbated [44,46,47], while the warming and drying tendencies are prominent in the southern portion of the Yangtze River source region and the Yellow River source region [48,49]. Additionally, the rate of temperature increase in the TRHR is more than twice the global average, and rising temperature may potentially trigger a drying of TRHR in the future [50,51].

2.2. Data Sources

The data used in this paper regarding the three dimensions of exposure, sensitivity, and adaptive capacity were sourced from a statistical yearbook, government gazette on the drought disaster, data from a meteorological station, climate communique, ecological meteorological bulletin, ecological environment status bulletin, climate change monitoring bulletin of the Tibetan autonomous prefectures, counties, and towns within the TRHR.

2.3. Construction of Indicator System

This study examines the spatiotemporal dynamics of drought vulnerability in the TRHR in six time periods 1990, 2000, 2010, 2015, 2020, and 2023, focusing on the dimensions of exposure, sensitivity, and adaptive capacity.

Exposure refers to the extent to which a system is subject to changes in external natural environmental conditions or social and political pressures, encompassing the frequency, duration, and intensity of these external pressures on social–ecological systems [52]. Eight indicators were selected to measure exposure: drought disaster record, population density, livestock quantity, yields of grain and oil crop, area of cropland, area of grassland, area of wetland, and the proportion of agricultural and animal husbandry output value.

Sensitivity refers to the extent to which a system is influenced by disturbances or pressures, reflecting the interaction between climatic events and social systems as well as the inherent characteristics of social–ecological systems in their pre-disturbance state [53]. Eight indicators were selected to assess sensitivity: the proportion of cropland, the proportion of grassland, the proportion of wetland, grain yield per unit area, the proportion of large livestock, the proportion of employees in agriculture and animal husbandry, urbanization rate, and water consumption.

Adaptive capacity refers to a system’s ability to respond to actual or anticipated pressures, encompassing its capabilities to accommodate and recover from disturbances within social–ecological contexts [27]. Fourteen indicators were selected to assess adaptive capacity: per capita disposable income, per capita cropland area, per capita grassland area, per capita wetland area, grain yield per capita, per capita livestock quantity, area of water conservancy facilities, total power of agricultural machinery, working-age population, education, transportation, communication, finance, and the savings deposits in financial institutions.

Table 1. Comprehensive evaluation indicator system of drought vulnerability in the TRHR [19].

Domains	Indicators	Description	Variable
Exposure	Drought disaster record	the number of occurrences of drought	a1
	Population density	person/km2	a2
	Livestock quantity	the number of livestock	a3
	Yields of grain and oil crop	kg	a4
	Area of cropland	km2	a5
	Area of grassland	km2	a6
	Area of wetland	km2	a7
	The proportion of agricultural and animal husbandry output value	the ratio of agricultural and animal husbandry output value to GDP (%)	a8
Sensitivity	The proportion of cropland	the ratio of cropland area to total land area (%)	b1
	The proportion of grassland	the ratio of grassland area to total land area (%)	b2
	The proportion of wetland	the ratio of wetland area to total land area (%)	b3
	Grain yield per unit area	kg	b4
	The proportion of large livestock	the ratio of the number of cattle, horses, mules, and donkeys to the number of livestock (%)	b5
	The proportion of employees in agriculture and animal husbandry	the ratio of the number of employees in agriculture and animal husbandry to the total population (%)	b6
	Urbanization rate	the ratio of urban population to total population (%)	b7
	Water consumption	ton	b8
Adaptive	Per capita disposable income	CNY 10,000	c1
capacity	Per capita cropland area	km2/person	c2
	Per capita grassland area	km2/person	c3
	Per capita wetland area	km2/person	c4
	Grain yield per capita	kg/person	c5
	Per capita livestock quantity	the ratio of the number of livestock to the number of total populations	c6
	Area of water conservancy facilities	km2/person	c7
	Total power of agricultural machinery	kw	c8
	Working-age population	the number of people aged 16–59 who have the ability to work	c9
	Education	the number of teachers and students	c10
	Transportation	mileage of highways opened to traffic	c11
	Communication	the number of mobile phone users	c12
	Finance	financial expenditure (CNY 10,000)	c13
	The savings deposits in financial institutions	CNY 10,000	c14

presents 30 indicators used to assess drought vulnerability, providing a comprehensive basis for evaluating exposure, sensitivity, and adaptive capacity. All indicators need to be standardized using Formula (2) in Section 2.3. After the range standardization, the range of values for all indicators is from 0 to 1, and then subsequent analysis of the 30 indicators can be conducted using other formulas in Section 2.3.

2.4. Method

2.4.1. Entropy Weight Method

Each indicator has a different unit, and its importance in assessing drought vulnerability also varies. Therefore, it is necessary to standardize each indicator and assign appropriate weights. The Entropy Weight Method, the Analytic Hierarchy Process (AHP), and the Equal Weight Method were usually used for assigning indicator weights [54]. Different weight assignment methods can cause variations in the rankings among different regions, thereby increasing the uncertainty of the drought vulnerability index for SES. Robustness analysis helps enhance the transparency and reliability of the drought vulnerability [54,55]. In this study, three weight assignment methods were used to calculate the drought vulnerability index of the study area from 1990 to 2023, and rankings for different regions were produced. Based on the range of ranking variation, three box plots were constructed to visualize the drought vulnerability index rankings (see Section 3.1 for details). These plots were further used to evaluate the robustness of the drought vulnerability results. A smaller variation range in rankings indicates stronger robustness [54,55]. Among the three methods, the Entropy Weight Method demonstrated the highest robustness in ranking stability and was therefore selected for calculating the drought vulnerability index (Figure 2).

The Entropy weight method, which belongs to the category of objective weighting methods, does not rely on subjective human judgment. The processes of assigning weights to each indicator using the Entropy weight method encompass four steps [55].

Firstly, a judgment matrix is constructed. Assuming there are i samples and j dimensions, the values of the random variables are expressed as follows.

$$X_{ij}=\left(\begin{array}{ccc}{X}_{11}& \cdots & X_{1j}\\ ⋮& \ddots & ⋮\\ X_{i1}& \cdots & X_{ij}\end{array}\right)$$

(1)

Secondly, the data are standardized to eliminate the influence of units and magnitudes.

$$Y_{ij}={\displaystyle \frac{X_{ij}-\mathrm{m}\mathrm{i}\mathrm{n}\left(X_i\right)}{\max \left(X_i\right)-\mathrm{m}\mathrm{i}\mathrm{n}\left(X_i\right)}}$$

(2)

In the formula, $\mathrm{m}\mathrm{a}\mathrm{x}\left(X_i\right)$ and $\mathrm{m}\mathrm{i}\mathrm{n}\left(X_i\right)$ represent the maximum and minimum values of the indicator i, respectively.

Thirdly, the information entropy of each indicator is calculated.

$$H_i=-{\displaystyle \frac{1}{\mathit{ln}m}}{\sum }_{i=1}^m{P}_{ij}ln{P}_{ij}$$

(3)

In the formula, $H_i$ is the information entropy of indicator i. $P_{ij}={\displaystyle \frac{Y_{ij}}{\sum _{i=1}^m{Y}_{ij}}}$, and it is defined as when $P_{ij}$ = 0, $\underset{P_{\mathit{ij}}\to 0}{\lim }{P}_{ij}{\mathit{ln}P}_{ij}$ = 0.

Finally, the weight of each indicator is determined based on the calculated information entropy.

$$w_i={\displaystyle \frac{1-H_i}{n-{\sum }_1^n{H}_i}}$$

(4)

In the formula, $w_i$ is the weight of indicator i.

2.4.2. Calculation of Drought Vulnerability Index

The calculation of the drought vulnerability index is as follows:

$$DVI=E+S+(1-AC)$$

(5)

where DVI is the value of the drought vulnerability index, E is the value of exposure, S is the value of sensitivity, and AC is the value of adaptive capacity. The larger the DVI value, the higher the drought vulnerability. In addition, the levels of drought vulnerability are shown in

Table 2. The levels of drought vulnerability.

Level	Low	Relatively Low	Medium	Relatively High	High
Drought vulnerability value	[0.0, 0.2)	[0.2, 0.4)	[0.4, 0.6)	[0.6, 0.8)	[0.8, 1.0]

.

2.4.3. Getis-Ord Gi* Model

The Getis-Ord Gi* model identifies spatial hot spots and cold spots of a variable within a set of geographical elements, enabling analysis of high-value and low-value clusters in local regions [1,55]. The Getis-Ord Gi* model not only determines the boundaries of high-value and low-value clusters in local areas but also measures clustering degrees within a defined confidence interval. The specific formulas are as follows:

$$G_i^{*}={\displaystyle \frac{\sum _{j=1}^n{w}_{(i,j)}{x}_j-\overline{x}{\sum }_{j=1}^{n}w(i,j)}{S\sqrt{\frac{[n\sum _{j=1}^n{w}_{\left(i,j\right)}^2-{\left(\sum _{j=1}^n{w}_{(i,j)}\right)}^2]}{n-1}}}}$$

(6)

$$\overline{x}={\displaystyle \frac{{\sum }_{j=1}^n{x}_j}{n}}$$

(7)

$$S=\sqrt{{\displaystyle \frac{{\sum }_{j=1}^n{x}_j^2}{n}}-{\left(\overline{x}\right)}^2}$$

(8)

where $G_i^{*}$ denotes the local $G_i^{*}$ index, x_j is the attribute value of element j, w_(i,j) is the spatial weight between elements i and j, and n is the total number of elements.

2.4.4. Analysis of Obstacle Factors

After assessing drought vulnerability, this study introduced the obstacle degree model to identify the key factors influencing drought vulnerability. This model efficiently highlights critical geographical elements [56]. The indicators with an obstacle degree exceeding 10% are regarded as the key influencing factors [1]. The specific calculation formula is as follows:

$$O_j={\displaystyle \frac{I_j\times W_j}{\sum _{j=1}^n{I}_j\times W_j}}$$

(9)

In the formula, O_j denotes the obstacle degree of the indicator j, quantifying each indicator’s impact on drought vulnerability. I_j represents the deviation degree of indicator j, typically calculated as 1 minus the standardized value of indicator j, representing the gap between the actual and optimal target values. W_j denotes the weight of indicator j, and n is the total number of indicators.

2.4.5. MCE-CA-Markov Model

The Cellular Automata (CA) model is a discrete model in time and space, capable of simulating complex spatial processes and serving as an effective tool for such tasks. The Markov model, with its advantage in temporal predictions, analyzes the probability of random events over time using a transition probability matrix. The CA-Markov model is a combination of the Markov and CA models [57]. The Multi-Criteria Evaluation (MCE) model can comprehensively analyze numerous factors affecting the target and provide decision-making assistance for predicting future changes in the target [57,58].

In this study, the MCE-CA-Markov model was employed to predict drought vulnerability in the THTR by 2050. The calculations were performed on the IDRISI Selva 17.0 software platform, following three key steps [57,58,59]. (1) Markovian transition estimator. The drought vulnerability distribution maps of the TRHR in 1990 and 2020 were, respectively, input into the Markov model as the earlier image and the later image. Running the model produced the transition probability matrix and the conditional probability image. (2) Creation of transition suitability image. Two scenarios were used to generate the transition suitability image. In the first scenario, the conditional probability image obtained from Step (1) was directly utilized as the transition suitability image. In the second scenario, an MCE model was applied, incorporating stakeholder input. During the fieldwork conducted from 2021 to 2024, discussions were held with a social network of stakeholders, including farmers, herdsmen, enterprises, local governments, NGOs, research institutions, and national park authorities. From the adaptive governance perspective, stakeholders proposed governance rules aimed at sustainable development. Additionally, they assigned weights to each governance rule and its corresponding factors influencing drought vulnerability. Subsequently, our research team systematically summarized and synthesized the stakeholder discussions, governance rules, and assigned weights. The Weighted linear combination and the Collection Editor in IDRISI were then used to generate the transition suitability image. (3) CA-Markov Model Prediction. Using the drought vulnerability distribution map of the TRHR in 2020 as the basis image and the transition probability matrix from Step (1) as the Markov transition areas file. Then, the transition suitability images of the two scenarios obtained in Step (2) were, respectively, input into the box of Transition suitability image collection, and thus the predictions of the drought vulnerability of the TRHR in 2050 for the two scenarios were obtained. One scenario was the prediction obtained relied solely on historical patterns of drought vulnerability of the TRHR from 1990 to 2020. The other scenario was the prediction obtained from the perspective of adaptive governance, based on the variation in drought vulnerability during 1990–2020 and the governance rules oriented towards sustainable development.

3. Results

3.1. Results of the Robustness Analysis

This study constructed three box plots to visualize the drought vulnerability index rankings of the TRHR based on the Entropy Weight Method, the AHP, and the Equal Weight Method, respectively (Figure 2). These box plots depict the variation range in drought vulnerability rankings for each county in the TRHR, indicated by the highest and lowest ranks. The most stable ranking sequence corresponds to the one with the highest frequency of occurrence.

Figure 2 demonstrates that the variation range for county rankings is the smallest based on the Entropy Weight Method. Since a smaller variation range reflects greater robustness, the Entropy Weight Method was selected among the three weight assignment methods for calculating the drought vulnerability index of the SES.

3.2. Spatiotemporal Evolution Characteristics of Drought Vulnerability at County-Level

This paper analyzes the spatiotemporal evolution characteristics of drought vulnerability in the TRHR from 1990 to 2023 and examines the corresponding changes in drought vulnerability hot spots during the same period (Figure 3 and Figure 4). From 1990 to 2023, drought vulnerability hot spots in the TRHR were primarily concentrated in the southeast, while cold spots were mainly found in the northwest (Figure 3 and Figure 4). From 1990 to 2000, drought vulnerability exhibited an increasing trend across all counties, followed by a decreasing trend from 2000 to 2023 (Figure 3).

In 1990, the average drought vulnerability in the TRHR was 0.526, indicating a medium level. Nine counties in the east, south, and southeast exhibited relatively high drought vulnerability, while six counties in the northwest and northeast, along with Tanggula Town, showed relatively low vulnerability (Figure 3). The highest drought vulnerability, 0.742, was recorded in Henan County, while the lowest, 0.308, was in Tanggula Town. In the same year, hot spots were identified in five counties in the southeast and one county in the south, whereas cold spots were mainly concentrated in Zhidoi County and Tanggula Town. Notably, Machin County, Gander County, Dari County, and Banma County were identified as hot spots with a 95% confidence level, Jiuzhi County with 99% confidence, and Nangqian County with 90% confidence (Figure 4).

In 2000, the average drought vulnerability in the TRHR rose to 0.637, indicating a relatively high level. Except for Tanggula Town, where drought vulnerability remained relatively low, other areas demonstrated vulnerability levels above medium (Figure 3). Zeku County, Henan County, Gander County, and Jiuzhi County exhibited high vulnerability. The highest value was recorded in Henan County at 0.846, whereas the lowest value was observed in Tanggula Town at 0.334. Hot spots in 2000 were concentrated in five counties in the southeast, with cold spots in Zhidoi County and Tanggula Town. Specifically, Dari County and Banma County were identified as hot spots with a 90% confidence level, Machin County and Gander County with 95% confidence, and Jiuzhi County with 99% confidence (Figure 4).

In 2010, the average drought vulnerability in the TRHR decreased to 0.523, indicating a medium level. Six counties in the east and southeast displayed relatively high vulnerability, while Tongren County, Qumarlai County, Chindu County, and Tanggula Town showed relatively low levels (Figure 3). The maximum value of 0.774 was observed in Jiuzhi County, whilst the minimum value of 0.313 was recorded in Tanggula Town. Hot spots were mainly distributed in seven counties in the southeast and east, while cold spots appeared in Zhidoi County, Qumarlai County, and Tanggula Town. Notably, Dari County, Zeku County, and Tongde County were identified as hot spots with a 90% confidence level, Banma County with 95% confidence, and Machin County, Gander County, and Jiuzhi County with 99% confidence (Figure 4).

In 2015, the average drought vulnerability in the TRHR slightly declined to 0.511, still indicating a medium level. Five counties in the east and southeast showed relatively high vulnerability, while four counties in the northeast, Qumarlai County, and Tanggula Town demonstrated relatively low levels (Figure 3). The highest value, at 0.753, was observed in Jiuzhi County, and the lowest value, at 0.316, was recorded in Tanggula Town. Hot spots were primarily concentrated in five counties in the southeast. Among them, Machin County, Gander County, Dari County, and Banma County were identified as hotspots with a 95% confidence level, while Jiuzhi County was identified with 99% confidence (Figure 4).

In 2020, the average drought vulnerability in the TRHR was 0.481, remaining at a medium level. Six counties in the east and southeast exhibited relatively high vulnerability, whereas four counties in the northeast, Machin County, Qumarlai County, and Tanggula Town showed relatively low levels (Figure 3). The peak value, at 0.687, was found in Jiuzhi County, whereas the lowest value, at 0.310, was seen in Tanggula Town. In 2020, hot spots were concentrated in five counties in the southeast. Specifically, Machin County, Gander County, Dari County, Banma County, and Jiuzhi County were identified as hot spots with a 99% confidence level (Figure 4).

In 2023, the average drought vulnerability in the TRHR further declined to 0.444, remaining at a medium level. Three counties in the southeast displayed relatively high vulnerability, while four counties in the northeast, three counties in the northwest, Tanggula Town, and Machin County exhibited relatively low levels (Figure 3). Jiuzhi County recorded the highest value at 0.656, while Tanggula Town had the lowest at 0.303. Hot spots in 2023 were concentrated in five counties in the southeast, with Machin County, Gander County, Dari County, Banma County, and Jiuzhi County all identified as hot spots with a 99% confidence level (Figure 4).

3.3. Spatiotemporal Evolution Characteristics of Drought Vulnerability at Town-Level

Due to the availability of the data, this paper analyzes the spatiotemporal variation in drought vulnerability in 82 towns within the TRHR from 1990 to 2023, and it also examines changes in hot spots of drought vulnerability during the same period (Figure 5 and Figure 6). Between 1990 and 2023, 85.37% of the 82 towns showed a decline in drought vulnerability (Figure 5). During this period, hot spots of drought vulnerability were mainly concentrated in the southeast and south, while cold spots were primarily located in the northwest and north (Figure 5 and Figure 6). From 1990 to 2000, the drought vulnerability of the 82 towns increased, but from 2000 to 2023, it declined (Figure 5).

In 1990, 38 towns in the east and south of the TRHR had high or relatively high levels of drought vulnerability, while 15 in the west and central parts exhibited relatively low levels (Figure 5). The highest drought vulnerability was recorded in Waeryi Town, Jiuzhi County (0.902), and the lowest was in Suojia Town, Zhidoi County (0.205). Hot spots in 1990 were concentrated in 45 towns in the southeast and south of the TRHR, while cold spots were located in 19 towns in the northwest and north-central regions (Figure 6).

In 2000, 58 towns in the east, south, and central regions of the TRHR exhibited high or relatively high levels of drought vulnerability, while 6 towns in the west and south-central regions had relatively low values (Figure 5). Waeryi Town, Jiuzhi County, again recorded the highest drought vulnerability (0.995), and Suojia Town, Zhidoi County, had the lowest (0.311). Hot spots were concentrated in 40 towns in the southeast and south of the TRHR, while cold spots appeared in 13 towns in the northwest and north-central regions (Figure 6).

In 2010, 31 towns in the southeast and south of the TRHR had high or relatively high levels of drought vulnerability, while 16 towns in the west and south-central regions showed relatively low levels (Figure 5). The peak value for drought vulnerability, at 0.991, was noted in Waeryi Town, Jiuzhi County, whereas the lowest value was 0.250 in Qingshuihe Town, Chindu County. Hot spots were observed in 37 towns in the southeast of the TRHR, while cold spots were found in 21 towns in the northwest and north-central regions (Figure 6).

In 2015, 28 towns in the southeast and south had high or relatively high levels of drought vulnerability, while 18 towns in the west and south-central regions exhibited relatively low levels (Figure 5). Drought vulnerability reached its highest level, at 0.985, in Baiyu Town, Jiuzhi County, with the lowest level, at 0.231, being found in Qumahe Town, Qumarlai County. Hot spots were concentrated in 35 towns in the southeast of the TRHR, while cold spots appeared in 17 towns in the northwest and north-central regions (Figure 6).

In 2020, 22 towns in the southeast and south showed high or relatively high levels of drought vulnerability, while 20 towns in the west and south-central regions had relatively low levels (Figure 5). The maximum drought vulnerability, at 0.898, was identified in Baiyu Town, Jiuzhi County, and the minimum vulnerability, at 0.210, was observed in Qumahe Town, Qumarlai County. Hot spots were concentrated in 33 towns in the southeast of the TRHR, while cold spots appeared in 17 towns in the northwest and north-central regions (Figure 6).

In 2023, 15 towns in the southeast and south-central regions of the TRHR exhibited high or relatively high levels of drought vulnerability, while 28 towns in the west, central, and south showed relatively low levels (Figure 5). The highest value of drought vulnerability, at 0.857, was observed in Baiyu Town, Jiuzhi County, while the lowest value, at 0.198, was recorded in Suojia Town, Zhidoi County. Hot spots in 2023 were concentrated in 33 towns in the southeast of the TRHR, while cold spots were observed in 15 towns in the northwest, north-central, and south regions (Figure 6).

3.4. Spatiotemporal Evolution Characteristics of Drought Vulnerability at Village-Level

Due to the availability of the data, this study analyzed the changes in drought vulnerability across 152 villages in four counties within the TRHR. Between 1990 and 2023, among 152 villages across four counties—Chindu County, Madoi County, Dari County, and Banma County—95.39% of the villages exhibited a declining trend in drought vulnerability (Figure 7). In 1990, 41.45% of the 152 villages exhibited relatively high or high levels of drought vulnerability, while 23.68% showed relatively low levels. By 2023, 22.37% of the 152 villages displayed relatively high or high levels of drought vulnerability, while 42.11% showed relatively low or low levels (Figure 7).

In Chindu County, the drought vulnerability of 57 villages in 1990 ranged from 0.210 to 0.950. Among these, 31.58% of the 57 villages experienced high or relatively high drought vulnerability, and 38.60% fell into the relatively low category. By 2023, vulnerability ranged from 0.165 to 0.746, with 15.79% of the 57 villages remaining in the relatively high category, while 54.39% experienced low or relatively low vulnerability (Figure 7).

In Madoi County, the drought vulnerability of 30 villages in 1990 ranged between 0.372 and 0.871. A total of 46.67% of the 30 villages were categorized as having high or relatively high drought vulnerability, while only 6.67% exhibited relatively low vulnerability. By 2023, the range shifted to 0.291–0.682, with only 3.33% of the 30 villages remaining in the relatively high category and 40.00% showing relatively low vulnerability (Figure 7).

In Dari County, the drought vulnerability of 33 villages in 1990 spanned from 0.222 to 0.944, with 45.45% experiencing high or relatively high vulnerability, and 21.21% classified as relatively low vulnerability. By 2023, drought vulnerability values ranged from 0.181 to 0.815, with 24.24% of villages in the high or relatively high category and 36.36% in the low or relatively low category (Figure 7).

In Banma County, the drought vulnerability of 32 villages in 1990 varied between 0.349 and 0.989. Half of the villages experienced high or relatively high vulnerability, while 15.63% exhibited relatively low vulnerability. By 2023, drought vulnerability ranged from 0.322 to 0.906, with 50.00% of villages still in the high or relatively high category, and 28.13% in the relatively low category (Figure 7).

3.5. Influencing Factors of Drought Vulnerability

In this paper, an analysis based on the obstacle degree model was conducted to identify the main factors influencing the drought vulnerability of the TRHR from 1990 to 2023. The results indicate that the obstacle degree values of five indicators (

Table 3. The obstacle degrees of the key influencing factors of drought vulnerability in the TRHR.

Domains	Variable	Oj	r	p
Exposure	a1	0.1086	0.770	0.037
Exposure	a8	0.1267	0.943	0.002
Sensitivity	b2	0.1008	0.538	0.054
Sensitivity	b5	0.1234	0.827	0.000
Adaptive capacity	c1	0.1229	−0.809	0.002

) to drought vulnerability exceeded 10%, highlighting their role as the key factors shaping the drought vulnerability of TRHR during this period.

Foremost, within the Exposure domain, drought disaster record (a1) and the proportion of agricultural and animal husbandry output value (a8) were identified as critical factors influencing drought vulnerability. These two factors (a1 and a8) underscore the extent of the impact of external drought events and socio-economic pressures on the SES. Secondly, within the Sensitivity domain, the proportion of grassland (b2) and the proportion of large livestock (b5) emerged as significant factors influencing drought vulnerability. These two factors (b2 and b5) reflect the intrinsic attributes of the SES before exposure to external drought-related disturbances, highlighting its sensitivity level. Finally, within the Adaptive capacity domain, per capita disposable income (c1) was found to be a pivotal factor, representing the SES’s capacity to respond to drought pressures. This factor (c1) indicates the economic foundation enabling the system to accommodate and recover from drought disturbances in a socio-ecological context. As for the remaining 25 indicators, while their obstacle degrees were relatively low, their presence nevertheless highlights their importance in the overall evaluation of drought vulnerability in TRHR.

3.6. Modeling Characteristics of Drought Vulnerability in 2050

In Figure 8, this study predicts the drought vulnerability of the TRHR in 2050 under two scenarios and analyzes the corresponding hotspots of drought vulnerability for each scenario. Figure 8a illustrates the projected drought vulnerability of the TRHR in 2050, derived from historical trends observed from 1990 to 2020. Under this scenario, the average drought vulnerability of the TRHR is 0.428 (Figure 8a), indicating a medium level. Additionally, 12 counties in the eastern and southern parts of the TRHR exhibit medium to relatively high drought vulnerability, with Jiuzhi County showing the highest value at 0.639. In contrast, the drought vulnerability of 9 counties in the northwest, central, and northeast TRHR, as well as Tanggula Town, is at a relatively low level, with the lowest value of 0.302 observed in Tanggula Town. In Figure 8b, the drought vulnerability of TRHR in 2050 is predicted based on the variation in drought vulnerability during 1990–2020 and adaptive governance rules oriented towards sustainable development. Under this scenario, the average drought vulnerability of the TRHR decreases to 0.350 (Figure 8b), indicating a relatively low level. Except for Zeku County and 4 counties in the southeastern TRHR, where drought vulnerability remains at a medium level, the remaining 16 counties and Tanggula Town demonstrate relatively low drought vulnerability. The lowest value of 0.247 is found in Tanggula Town, while Jiuzhi County again shows the highest value, at 0.522. In both scenarios, the drought vulnerability hot spots in the TRHR are concentrated in 5 counties in the southeastern region (Figure 8c,d). Among these, Machin County and Gander County are identified as hot spots with 95% confidence, while Dari County, Jiuzhi County, and Banma County are identified as hot spots with 99% confidence.

4. Discussion

4.1. Policy Recommendations

Drawing on the analysis of influencing factors in Section 3, stakeholder interview data from social networks, and findings from previous studies, this paper proposes a set of adaptive governance strategies to mitigate drought vulnerability in the TRHR: (a) Develop an integrated drought dynamic monitoring, early warning, and emergency management system. (b) Establish an ecological management, protection, and financial support model that involves the joint participation of a social network comprising farmers, herders, enterprises, local governments, NGOs, universities, research institutions, and national park management departments. (c) Build a collaborative mechanism for multiple stakeholders to enhance the livelihood security of ecological migrants and improve ecological protection subsidies. (d) Accelerate the implementation of a horizontal transfer payment system for cross-regional ecological compensation. (e) Strengthen the intelligent management of nature reserves and national parks. (f) Foster multi-stakeholder participation to promote the coordinated development of scientific research, the application of new technological advancements, and ecological and environmental protection education.

4.2. Comparison with Previous Research Findings

Drought is one of the costliest and most complex disasters, with increasingly prominent impacts on SES [44,49]. The intensification of climate change and human activities introduces uncertainty into drought dynamics in the TRHR [49]. Research on drought simulation and prediction is crucial to minimizing the economic and ecological damage caused by drought hazards [60]. This study on drought vulnerability in the TRHR contributes to improving the capacity of SES to respond to drought disturbances. Furthermore, this paper forecasts the drought vulnerability of the TRHR under two future scenarios, representing a meaningful exploration for mitigating the losses caused by drought hazards.

Previous research indicated that drought conditions in the South Qinghai Plateau generally showed a mitigation trend between 1982 and 2015 [61]. Consistently, this study found that drought vulnerability in the TRHR, located on the South Qinghai Plateau, decreased from 0.526 in 1990 to 0.444 in 2023. Regional differences in drought dynamics within the TRHR have also been identified in prior studies, with more severe drought conditions in the eastern and southern parts compared to other regions [44,45,46,47]. Similarly, this study revealed that the values of drought vulnerability in the eastern and southern parts are higher than in other regions within the TRHR. This paper found that the hot spots of drought vulnerability in the TRHR are concentrated in the southeastern and southern regions, while cold spots are predominantly in the northwest. The drought trend in the Yellow River source region has been highlighted as particularly pronounced in previous research [44,48,49]. This study corroborates this, identifying Machin County, Gander County, Dari County, and Jiuzhi County—areas within the Yellow River source region—as drought vulnerability hotspots. Gander County and Jiuzhi County consistently exhibited relatively high or high levels of drought vulnerability during 1990–2023. In addition, previous studies observed a tendency toward wetter conditions and drought alleviation in the central-northern and northeastern parts of the TRHR, whereas drought frequency increased in Banma County in the southeastern part [44,45,46,47,48,49]. This study similarly found that drought vulnerability cold spots were concentrated in the central-northern TRHR at the town scale from 1990 to 2023. Meanwhile, four counties in the northeastern TRHR exhibited relatively low levels of drought vulnerability from 2015 to 2023. Notably, Banma County maintained relatively high levels of drought vulnerability throughout 1990–2023.

Previous studies have emphasized that drought is a significant factor affecting ecosystem vulnerability [62]. This study aligns with this finding, identifying the drought disaster record as a key factor influencing drought vulnerability in the TRHR. Grasslands on the Tibetan Plateau have been recognized as the most drought-prone ecosystems in western China [62]. Consistent with this, this study found that the proportion of grassland is a critical factor influencing drought vulnerability in the TRHR. Moreover, diversifying livelihoods to increase income is one of the most commonly practiced adaptive strategies for pastoralists on the Tibetan Plateau in response to climate change and disasters [63]. Correspondingly, this study identified per capita disposable income as a key factor influencing drought vulnerability in the TRHR. Relevant scholars have identified the proportion of agricultural and animal husbandry output value and the proportion of large livestock as key factors influencing the vulnerability to drought disasters in Tibet [19]. Given that the TRHR is geographically adjacent to the Tibet Autonomous Region, our study has reached similar conclusions. Specifically, the southeastern part of the TRHR has consistently been a hotspot for drought vulnerability due to its high proportion of agricultural and animal husbandry output value and a significant presence of large livestock (such as yaks).

4.3. Uncertainties

Studies concerning multi-scale assessments of the drought vulnerability of SES in the TRHR are rare. Therefore, we investigated the drought vulnerability of SES at the county-town-village spatial scales over the period 1990–2023. The uncertainties of this work have also been considered. First, due to limitations in data availability or accuracy, factors such as the internal structure and function of ecosystems, species diversity, and transportation modes such as air transport and rail transport have not been incorporated into the indicator system. Second, there are uncertainties or limitations associated with the drought vulnerability assessment model, such as the impact of complex associations between the factors of the SES on the assessment model, and Limitations of model validation. Third, Limitations of the Scenario Settings. Due to data availability constraints, scenarios involving extreme climate change and technological breakthroughs have not been considered.

5. Conclusions

This study established a drought vulnerability indicator system for the TRHR, analyzed the spatiotemporal evolution of drought vulnerability at the county, town, and village levels from 1990 to 2023, examined key driving factors influencing these changes, and predicted the distribution of drought vulnerability in the TRHR by 2050. The main conclusions are as follows:

(1) From 1990 to 2023, the average drought vulnerability of the TRHR decreased from 0.526 to 0.444. Drought vulnerability declined in 17 out of 21 counties within the TRHR. Among the 82 selected towns, 85.37% experienced a decline, and among the 152 selected villages, 95.39% showed a reduction. From 1990 to 2000, the drought vulnerability of counties and towns in the TRHR increased, but it decreased between 2000 and 2023. At the county level, hot spots were concentrated in the southeast of the TRHR, while cold spots were in the northwest. At the town level, hot spots were mainly in the southeast and south of the TRHR, and cold spots were primarily in the northwest and north-central regions.

(2) In 1990, the county with the highest drought vulnerability was Henan County. Waeryi Town in Jiuzhi County had the highest vulnerability among towns, while Suojia Town in Zhidoi County had the lowest. Of the 152 selected villages, 41.45% exhibited relatively high or high levels of drought vulnerability, while 23.68% showed relatively low levels. In 2023, the county with the highest drought vulnerability was Jiuzhi County. Baiyu Town in Jiuzhi County ranks highest among towns and Suojia Town in Zhidoi County remains the lowest. 22.37% of the selected 152 villages displayed relatively high or high levels of drought vulnerability, while 42.11% showed relatively low or low levels.

(3) Key factors influencing drought vulnerability in the TRHR include drought disaster records, the proportion of agricultural and animal husbandry output value, the proportion of grassland, the proportion of large livestock, and per capita disposable income.

(4) By 2050, under the first scenario, the average drought vulnerability of the TRHR is projected to be 0.428, indicating a medium level. Under the second scenario, the average vulnerability decreases to 0.350, representing a relatively low level. In both scenarios, the highest drought vulnerability is projected in Jiuzhi County, and the lowest in Tanggula Town. Hot spots are concentrated in five counties in the southeast of the TRHR: Machin County, Gander County, Dari County, Jiuzhi County, and Banma County.

(5) The adaptive governance strategies to mitigate drought vulnerability in the TRHR include developing an integrated drought management system; establishing an ecological management, protection, and financial support model; building a collaborative mechanism for multiple stakeholders; accelerating the implementation of horizontal transfer payment system; strengthening the intelligent management of nature reserves and national parks; and so on.