Suitability Evaluation of Human Settlements Using a Global Sensitivity Analysis Method: A Case Study in China

Abstract

The suitability evaluation of human settlements over time and space is essential to track potential challenges towards suitable human settlements and provide references for policy-makers. This study established a theoretical framework of human settlements based on the nature, human, economy, society, and residence subsystems. Evaluation indicators were determined with the consideration of the coupling effect among subsystems. Based on the extended Fourier amplitude sensitivity test algorithm, the global sensitivity analysis was used to determine the weights of indicators. The human settlement suitability was evaluated in 30 provinces of China between 2000 and 2016. The findings were as follows: (1) human settlement suitability index (HSSI) values increased significantly in all 30 provinces from 2000 to 2016. The suitability index of the residence subsystem in China exhibited the fastest growth, followed by the society and economy subsystems. (2) HSSI in eastern provinces with a developed economy was higher than that in western provinces with an underdeveloped economy. In contrast, the growth rate of HSSI in eastern provinces was significantly higher than that in western provinces. (3) The inter-provincial difference in HSSI narrowed down from 2000 to 2016. For subsystems, the difference narrowed down for the residence system, whereas it widened for the economy system. (4) The suitability of the nature subsystem has become a limiting factor for the improvement of human settlement suitability, especially in economically developed provinces, such as Beijing, Shanghai, and Guangdong. The results can be helpful to support decision-making and policy for improving the quality of human settlements in a broad nature, human, economy, society, and residence context.

1. Introduction

A human settlement is a fundamental environment where people gather and live [1]. It comprises five systems, including the natural, human, social, residential, and support systems [2]. The 2030 Agenda for Sustainable Development, proposed by the United Nations General Assembly, demonstrated the economic, social, and environmental pillars of sustainable human settlements [3]. In addition, United Nations Human Settlements Programme promoted human settlements development with adequate housing, infrastructure, and basic services, such as water, land, and energy. Several scholars advocated that suitable human settlements must be established in an integrated framework [4,5]. The suitability of human settlements directly affects human comfort and livability, and it also affects the development of the economy and society [6]. Over the last decade, there has been an interest and increasing concern by residents, scholars, and governments over the suitability evaluation of human settlements all over the world [7,8,9].

With global industrialization and urbanization, human settlements have undergone dramatic changes in most regions of the world, especially in developing countries, such as China [10,11]. For instance, the global urban population outnumbered its rural counterpart in 2008, and it was predicted to reach 70% of the global population by 2050 [12]. As one of the world’s largest developing countries, the urbanization rate in China increased from 17.9% in 1978 to 60.6% in 2019 [13]. In addition, the built-up ratio enlarged three-fold between 1975 and 2014 [14]. Over the past few decades, while rapid economic growth has greatly improved the quality of life for human beings, it has also caused some problems in human settlements [15]. For instance, previous studies found that natural habitat loss was closely associated with the expansion of human settlements [11]. Specifically, dramatic changes in the natural environment, such as the depletion of natural resources and air pollution, have reduced living suitability. As there are significant regional differences in climatic, natural, economic, social, and residential conditions, the suitability level of human settlements varies widely from region to region [16]. Therefore, tracking the suitability of human settlements over time and space can help unveil potential challenges in improving the quality of human settlements and provide references for regional policy-makers.

To attain suitable human settlements, it is of great importance to protect natural resources and the ecological environment, especially in a densely populated region. Moreover, the coordinated development of the natural, economic, social, and residential conditions is the key aspect of a suitable human settlement [17,18]. Yamasaki and Yamada [19] determined the causal relationships among the Sustainable Development Goal (SDG) 11 indicators. They suggested that some evaluation indicators belonging to different subsystems of human settlements which interacted closely. This coupling relationship among subsystems has complex effects on the suitability of human settlements. For instance, China’s high-speed economic growth is natural resources-driven [20]. Rapid economic growth improves the suitability of the economic subsystem, whereas it also reduces the suitability of the natural subsystem. However, previous studies on suitability evaluation mainly concentrated on each subsystem of human settlements. The coupling effects among subsystems of human settlements were rarely considered during the process of suitability evaluation. In terms of evaluation methods, the traditional methods, such as the entropy method [21], analytic hierarchy process [22], and fuzzy comprehensive evaluation [23], failed to quantify the coupling effect among indicators. As a global sensitivity analysis method, the extended Fourier amplitude sensitivity test (EFAST) algorithm can effectively quantify the parameter sensitivity based on independent and coupling effects between parameters. It has been widely used in parameter sensitivity analysis in ecology, hydrology, and remote sensing models [24,25,26]. To overcome the limitation of traditional evaluation methods, the EFAST algorithm was introduced to analyze the sensitivity of the indicator and then determine its weight in this study. Given the consideration of uncertainty and coupling effects among indicators, it makes the evaluation results more objective.

In this study, we considered coupling effects among subsystems of human settlements as decisive means to evaluate the suitability of human settlements. The objective of this study was to quantify spatial and temporal changes in human settlement suitability in China. First, evaluation indicators of human settlement suitability were selected based on coupling effects between subsystems. Second, the weights of indicators were determined using the EFAST algorithm that considered coupling effects among indicators. Third, the suitability change in human settlements over time and space was analyzed for each subsystem and comprehensive system in 30 provinces of China between 2000 and 2016. The results can be helpful to support decision-making and policy for improving the quality of human settlements in a broad nature, human, economy, society, and residence context.

2. Literature Review

Although many studies of human settlements have been conducted all over the world, the definition of human settlements has not yet reached a consensus. The concept of human settlements was first proposed by Doxiadis, who stressed that five basic elements (i.e., nature, human, society, architecture and city, transportation, and communication network) should be systematically studied [27]. Wu [2] proposed the theory of human settlements science and classified the human settlement system into five subsystems, namely the nature, human, society, residence, and support subsystems. Due to the complexity of human settlements, studies on the suitability evaluation of human settlements were conducted from different perspectives. For instance, Mori and Yamashita [28] constructed a sustainable urban development index from three aspects, namely environmental sustainability, and economic and social equity. Proposed in 2012 and adjusted in 2016 by UN-Habitat, the City Prosperity Index was constructed from six aspects, namely productivity, environmental sustainability, quality of life, infrastructure development, equity and social inclusion, and urban governance and regulations [29]. According to Sustainable Development Goal 11, human settlements should be more inclusive, safe, resilient, and sustainable, and positive economic, social, and environmental links should be established between urban and rural areas [10]. Tang et al. [16] assessed the suitability of human settlements in 35 cities in China focusing on the natural environment, economic, social, housing, and infrastructure conditions. Hu and Wang [30] established a theoretical framework of rural human settlements from five dimensions, namely infrastructure, public services, quality of life, living standards, and environmental health. Xue et al. [15] evaluated the suitability of human settlements in China with a real and pseudo human settlements index. To sum up, while the elements or subsystems of human settlements differed widely in the aforementioned studies, the suitability evaluation mainly focused on the natural environment, economic and social development, as well as human and housing conditions. Moreover, existing studies found that human settlement subsystems interacted closely [10,31]. However, the coupling effects among subsystems of human settlements were rarely considered during the process of suitability evaluation and indicator determination.

It is an important step to determine indicator weights and aggregate them to form a composite index to evaluate the suitability of human settlements. Existing evaluation studies used objective and subjective methods to determine indicator weights, such as the entropy method [21], analytic hierarchy process [22], fuzzy comprehensive evaluation [23], principal component analysis [32], the Delphi method, and the questionnaire survey method [7]. However, these traditional methods have limitations in indicator determination. In particular, the subjective evaluation methods (i.e., the Delphi method) are significantly influenced by individual knowledge and experience, which can affect the objectivity of the results to a certain extent [33]. In contrast, the objective evaluation methods (i.e., the entropy method) focus on the discrete data analysis and do not fully consider the coupling effects among indicators, thereby causing deviations in the evaluation results [34]. Researchers have revealed that human settlement subsystems closely interacted and, hence, strong coupling effects were obtained between indicators. Chai et al. [19] used population, GDP, annual precipitation, total water withdrawal, and total energy production as indicators to analyze the causal relationship among water, food, energy, social, economic, and environmental systems. The influence of interactions among indicators cannot be neglected for an accurate evaluation of human settlement suitability. However, both the objective and subjective methods mentioned above failed to quantify the coupling effects among indicators.

The EFAST algorithm can quantify the sensitivity of the independent effect of parameters and the coupling effect between parameters to the results. As a global sensitivity analysis method, it has been widely used in parameter sensitivity analyses in ecology, hydrology, and remote sensing models [24,25,26]. To address the limitation of traditional methods, some scholars used the EFAST algorithm that considered the coupling effect among parameters to determine indicator weights. Luan et al. [33] used the entropy method and EFAST algorithm to weigh indicators in sustainable development evaluation. The comparison revealed that the EFAST algorithm was more efficient and robust to quantify the sensitivity and uncertainty among indicators. Zhao et al. [26] used the EFAST algorithm to determine the weights of water resource carrying capacity evaluation indicators. This evaluation further verified the feasibility and advantages of this method in efficiently separating the importance of the indicators. For the human settlements indicator system, although the coupling relationship among indicators was uncertain, the interaction was sufficient to generate high sensitivity. In consideration of the uncertainty and coupling effect among indicators, the EFAST algorithm was introduced to analyze the indicator sensitivity and determine indicator weights in this study.

3. Data and Methodology

3.1. Data Description

Human settlement suitability evaluation required extensive amounts of data concerning the nature, human, economy, society, and residence factors in the specific region. These data were mainly obtained from national and provincial statistical yearbooks, such as the China Statistical Yearbook, China Industry Statistical Yearbook, China Energy Statistical Yearbook, China Water Resources Bulletin, China Urban and Rural Construction Statistical Yearbook, and so on. To ensure data continuity and integrity, provincial administrative regions with >1% of missing data were excluded from this study [35]. Due to a lack of available data, the regions of Tibet, Hong Kong, Macao, and Taiwan in China were excluded from the analysis. Ultimately, 30 provincial administrative regions of China were included (Figure 1), and their yearly data from 2000 to 2016 were used to evaluate the suitability of human settlements.

3.2. Establishment of the Indicator System

3.2.1. Theoretical Framework of Human Settlements

Human settlements are a multi-dimensional concept, and the evaluation of human settlements needs to be conducted in a unified framework. Human settlements, first proposed by Doxiadis, were defined as a comprehensive system that combined five basic elements, including nature, human, society, architecture and city, transportation, and communication networks [27]. Enlightened by Doxiadis, the Chinese scholar Wu [2] put forward the science of human settlements at the beginning of the 1990s. The core of human settlements science was the relationship between humans and the environment. The concept of the environment referred to the natural, economic, social, housing, and other environments for supporting human survival and development. Based on the theory of human settlements and “society-economy-natural” compound ecosystems [36], human settlements were defined as the fundamental environment consisting of the nature, population, economy, society, and housing subsystems. Many scholars evaluated human settlements from different perspectives, such as nature sustainability [37], economic and social equity [28], population, housing, infrastructure, public services, quality of life, living standards, and so on [10,16,30]. On this basis, the theoretical framework of human settlements in this study was constructed to include five perspectives, namely nature, human, economy, society, and residence (Figure 2).

3.2.2. Selection of Evaluation Indicators

Based on SDG 11 indicators [38] and the coupling relationship analysis framework of the environmental livelihood security subsystem [39], we proposed an integrated framework (Figure 3) to analyze the coupling effect of human settlements. From Figure 2, there are connections between all five attributes. In combination with Figure 2, we further analyzed the mutually dependent relationship among subsystems to identify the main factors affecting suitability. Figure 3 shows the influencing factors on human settlement suitability evaluation.

We take the relationship between the natural subsystem and other subsystems as an example. The nature subsystem is the basis of human settlements. Resources and the environment are important for human survival and development [31]. As shown in Figure 3, the nature subsystem supplies natural resources and the environment to support other subsystems. For the human subsystem, the climate, land, food, water, and energy supported by the nature subsystem are the essential resources for human survival and development. For the economy subsystem, the natural resource endowment has a relationship to the industrial structure. In addition, natural resource utilization efficiency, such as water and energy, is an important factor related to the economic development level. For the social subsystem, the traffic road is supported by the nature subsystem. For the residential subsystem, public green spaces and domestic water and natural gas penetration rates are essential factors in the suitability, which are also closely related to the nature subsystem.

It is necessary to consider the coupling relationship in indicator determination. We also take the nature subsystem as an example. Based on mutual relationship between nature and the other subsystems shown in Figure 3, these factors need to meet human demand and decrease the pressure on nature. On the basis of existing studies [1,28,30] and the representativeness of indicators, five indicators were set up to evaluate the suitability of the nature subsystem. The Universal Thermal Climate Index (UTCI) represents the climate suitability. Four indicators, namely the number of days when the air quality reaches or is better than Grade II, per capita water resources, cultivated land area, and primary energy production, can measure the natural resource endowment for supporting human survival and development. Finally, 30 indicators in the dimensions of nature, human, economy, society, and residence were determined.

Table 1. An evaluation index system for human settlement suitability.

Level 1 Indicator	Level 2 Indicators	Level 3 Indicators	Indicator Type	Range of Indicators
Human settlement suitability A	Nature subsystem B1	UTCI (°C) H1	±	−2.4~27.1
		Number of days when the air quality reaches or is better than Grade II (Day) H2	+	49~365
		Per capita water resources (m3/person) H3	+	72.8~16,176.9
		Per capita cultivated land area (hectares/person) H4	+	0.01~0.44
		Per capita primary energy production (tons of standard coal/person) H5	+	0~29.9
	Human subsystem B2	Population density (person/km2) H6	−	7~3851
		Proportion of labor force (%) H7	+	63.5~83.8
		Sex ratio (female = 100) H8	±	94.9~120.4
		Proportion of urban population (%) H9	+	14.5~89.6
		Level of education (%) H10	+	1.8~45.5
	Economy subsystem B3	Per capita GDP (yuan) H11	+	2759~118,198
		Per capita local fiscal revenue (yuan) H12	+	227~26,472
		Proportion of primary industry’s added value (%) H13	−	0.4~36.4
		Energy consumption per 10,000 yuan GDP (tons of standard coal) H14	−	0.15~5.92
		Water consumption per 10,000 yuan GDP (m3) H15	−	15.1~3519.8
		Proportion of investment in environmental pollution control GDP (%) H16	+	0.3~4.2
		Residential investment as a share of GDP (%) H17	±	4.3~37.4
	Society subsystem B4	Per capita disposable income of urban residents (Yuan) H18	+	4724~57,692
		Urban registered unemployment rate (%) H19	−	0.8~6.5
		Urban road area per capita (m2/person) H20	+	3.9~25.8
		Public transport vehicles per 10,000 people (vehicle) H21	+	3.0~26.4
		Number of health technicians per 10,000 people (person) H22	+	20~108
		Urban sewage treatment rate (%) H23	+	8.7~97.4
	Residence subsystem B5	Residential building area per capita (m2/person) H24	+	13.7~76.7
		Per capita public green space (m2/person) H25	+	2.2~19.8
		Daily domestic water consumption per capita (liter/person) H26	+	66.8~292.0
		Annual electricity consumption per capita (kWh) H27	+	56.3~988.1
		Penetration rate of urban tap water (%) H28	+	47.2~100.0
		Penetration rate of city gas (%) H29	+	23.5~100.0
		Harmless treatment rate of domestic garbage (%) H30	+	13.1~100.0

Note: “+” represents a benefit in the index; “−” represents a cost in the index and “±” represents the moderate indicator.

shows the evaluation index system for human settlement suitability.

3.3. Construction of the Suitability Index

It is critical to determine the indicator weights and aggregate them to form a composite index. To better reflect the characteristics and variations of human settlement suitability, the suitability index was established for each subsystem and comprehensive system based on the evaluation index system in Table 1.

3.3.1. Data Standardization

The initial indicator data need to be normalized to eliminate the influence of different magnitudes and units across different indicators [21]. In this study, the min–max normalization was used to standardize indicators to a uniform scale with the mean value of 0 and the standard variation of 1.

Construct the original data matrix $V={\left\{x_{ij}\right\}}_{m\times n}$, which contains i indicators for j samples. As shown in Table 1, the indicators were divided into three types. The positive indicator (+) represents the higher suitability with a larger value. The negative indicator (−) represents the higher suitability with a lower value. The min–max normalization equations to positive and negative indicators can refer to [15]. The moderate indicator (±) represents the high suitability within a certain interval and can be normalized using the following Equation (1):

$$X_{ij}=\left\{\begin{array}{c}\begin{array}{c}1-\frac{q_1-x_{ij}}{\max \left(q_1-\underset{1\le j\le n}{\min }{(x}_{ij}),\underset{1\le j\le n}{\max }{(x}_{ij})-q_1\right)}{x}_{ij}<q_1\\ 1-\frac{q_1-x_{ij}}{\max \left(q_2-\underset{1\le j\le n}{\min }{(x}_{ij}),\underset{1\le j\le n}{\max }{(x}_{ij})-q_2\right)}{x}_{ij}>q_2\end{array}\\ 1\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{q}_2\le x_{ij}{\le q}_1\end{array}\right\}$$

(1)

where X_ij is the normalized value, and q₁ and q₂ are the low and high values of the optimal interval, respectively. In this study, the suitability of 30 provinces between 2000 and 2016 was evaluated, thus, m = 30 and n = 510.

3.3.2. Weight Determination

The EFAST algorithm is a global sensitivity analysis method based on the variance decomposition, which incorporates the advantage of the FAST and Sobol method [40]. In this study, 30 indicators were selected to evaluate human settlement suitability. The influence of different indicators on evaluation results varied. Moreover, the coupling effect among indicators produced uncertainty in the evaluation results. In consideration of the uncertainty and the coupling effects between indicators, the EFAST algorithm was used to analyze the indicator sensitivity and determine indicator weights in this study.

The indicator sensitivity analysis model with n indicators can be written as follows: y = f(x₁, x₂,…, x_n). According to the variance decomposition method of Sobol, the model output variance can be decomposed into the following components:

$$V={\sum }_i{V}_i+{\sum }_{i\ne j}{V}_{ij}+{\sum }_{i\ne j\ne k}{V}_{ijk}+\dots +\sum V_{12\dots n}$$

(2)

where V_i is the variance caused by indicator x_i, V_ij is the variance caused by the coupling effect associated with indicator x_j, and V_12…n is the variance caused by the coupling effect associated with indicators x₁, x₂, ..., x_n.

In the EFAST model, the first-order sensitivity index (S_i) represents the contribution of a single parameter to the output, and the global sensitivity index (S_ti) represents the contribution of both the parameter itself and the coupling effect among all parameters to the model output [24,26]. In this study, the indicator can be taken as a parameter. Thus, the first-order, second-order, third-order, and other high-order sensitivity indexes of indicator x_i can be defined as the following Equation (3):

$$S_i=\frac{V_i}{V},S_{ij}=\frac{V_{ij}}{V},S_{ijk}=\frac{V_{ijk}}{V},S_{ij,\dots n}=\frac{V_{ij,\dots n}}{V}$$

(3)

where S_i, S_ij, S_ijk, and S_ij...n are the first-order, second-order, third-order, and n-order sensitivity indexes, respectively.

The global sensitivity index of indicator x_i is the sum of each order of sensitivity index and can be defined as follows in Equation (4):

$$S_{ti}=S_i+S_{ij}+S_{ijk}+\dots +S_{ij\dots n}$$

(4)

The coupling effect need to be considered in the weight determination. Thus, the global sensitivity index (S_ti) was used to calculate the indicator weight. The weight of the ith indicator (W_i) can be defined as the following Equation (5):

$$W_i=S_{ti}/\sum _{i=1}^n{S}_{ti}$$

(5)

where w_i is the indicator weight.

The EFAST algorithm is embedded in Simlab 2.2, which is the uncertainty and sensitivity analysis software published by the European Commission Joint Research Center. Simlab 2.2 includes data pre-processing, model calculation, and data post-processing modules. In this study, the indicators were taken as input parameters. The process of indicator weight calculation can be shown as follows:

• Step 1: Establish the data matrix X = {x_mn} for 30 evaluation indicators during 2000–2016, thus, m = 30 and n = 17.
• Step 2: Define the value range and distribution form of the input indicator. The range of the indicator refers to the maximum and minimum values for each indicator during 2000–2016. Because the probability distribution of the indicator is unpredictable, it is appropriate to assume a uniform distribution function.
• Step 3: Obtain input samples after sampling using the pre-processing module. The EFAST algorithm is effective when the sampling size is 65 times greater than the number of parameters. Considering the computational efficiency and the number of representative samples, the sampling size of each indicator weight is set to 260. Thus, the total sampling size is 7800.
• Step 4: Calculate the sensitivity for each indicator. Take the sampling data as the input variable and take the human settlement suitability index as the output variable. Calculate the first-order sensitivity and global sensitivity index for each indicator. Analyze the impact of each indicator on the output result.
• Step 5: Calculate the indicator weight based on the sensitivity index. The first-order sensitivity index represents the contribution of a single indicator to the output, and the global sensitivity index represents the contribution of both the indicator itself and the coupling effect among all indicators to the model output [24,41]. Calculate and compare the first-order and global sensitivity weights based on the first-order sensitivity index and global sensitivity index of each indicator.

3.3.3. Establishment of Human Settlement Suitability Index

The human settlement system is comprised of five subsystems, including the natural, human, economic, social, and residential subsystems. The suitability index for each subsystem and comprehensive system were calculated using the global sensitivity weight (W_i) that considered the coupling effect between indicators. NSI, HSI, ESI, SSI, and RSI were proposed to represent the suitability of the natural, human, economic, social, and residential subsystems, respectively. Taking NSI as an example, NSI can be defined as in Equation (6) based on the evaluation index system in Table 1. The human settlement suitability index (HSSI) denoted the suitability of the comprehensive system. The formula is shown in Equation (7), as follows:

$${NSI}_j={\sum }_{i=1}^5{x}_{ij}\times \frac{W_i}{{\sum }_{i=1}^5{W}_i}\times 100$$

(6)

$${HSSI}_j={\sum }_1^{30}{x}_{ij}\times W_i\times 100$$

(7)

where NSI_j denotes the suitability index of the natural subsystem for the jth sample, HSSI_j denotes the suitability index of human settlement system for the jth sample, and $W_i$ denotes the global sensitivity weight.

4. Results

4.1. Indicator Weights Calculation

Table 2. Comparison of first-order and global sensitivity weights.

Level 1 Indicator	Level 2 Indicators	Level 3 Indicators	First-Order Sensitivity Weight		Global Sensitivity Weight
			Weight	Ranking	Weight	Ranking
Human settlement suitability A	Nature subsystem B1	UTCI (°C) H1	0.016	30	0.018	30
		Number of days when the air quality reaches or is better than Grade II (Day) H2	0.024	22	0.026	22
		Per capita water resources (m3/person) H3	0.018	29	0.026	24
		Per capita cultivated land area (hectares/person) H4	0.026	20	0.028	20
		Per capita primary energy production (tons of standard coal/person) H5	0.026	21	0.029	17
	Human subsystem B2	Population density (person/km2) H6	0.028	17	0.029	16
		Proportion of labor force (%) H7	0.033	12	0.031	14
		Sex ratio (female = 100) H8	0.024	23	0.026	23
		Proportion of urban population (%) H9	0.039	7	0.038	8
		Level of education (%) H10	0.086	1	0.093	1
	Economy subsystem B3	Per capita GDP (yuan) H11	0.051	3	0.044	4
		Per capita local fiscal revenue (yuan) H12	0.035	11	0.034	11
		Proportion of primary industry’s added value (%) H13	0.059	2	0.050	2
		Energy consumption per 10,000 yuan GDP (tons of standard coal) H14	0.020	27	0.020	29
		Water consumption per 10,000 yuan GDP (m3) H15	0.027	18	0.028	19
		Proportion of investment in environmental pollution control GDP (%) H16	0.020	28	0.021	28
		Residential investment as a share of GDP (%) H17	0.045	5	0.041	5
	Society subsystem B4	Per capita disposable income of urban residents (Yuan) H18	0.021	26	0.021	27
		Urban registered unemployment rate (%) H19	0.036	10	0.037	10
		Urban road area per capita (m2/person) H20	0.047	4	0.049	3
		Public transport vehicles per 10,000 people (vehicle) H21	0.039	8	0.038	9
		Number of health technicians per 10,000 people (person) H22	0.027	19	0.026	21
		Urban sewage treatment rate (%) H23	0.042	6	0.039	7
	Residence subsystem B5	Residential building area per capita (m2/person) H24	0.024	24	0.023	25
		Per capita public green space (m2/person) H25	0.032	14	0.030	15
		Daily domestic water consumption per capita (liter/person) H26	0.032	15	0.032	13
		Annual electricity consumption per capita (kWh) H27	0.038	9	0.040	6
		Penetration rate of urban tap water (%) H28	0.033	13	0.033	12
		Penetration rate of city gas (%) H29	0.030	16	0.028	18
		Harmless treatment rate of domestic garbage (%) H30	0.022	25	0.022	26

shows the comparison of first-order and global sensitivity weights. The indicators with the top five weights were consistent for first-order and global sensitivity weights, including the level of education, proportion of the primary industry’s added value, per capita GDP, urban road area per capita, and residential investment as a share of GDP. Indicator weights and their rankings with global sensitivity weight changed for some indicators. For example, the weights of per capita GDP and urban road area per capita with first-order sensitivity weight ranked third and fourth, respectively. In comparison, they ranked fourth and third with global sensitivity weight. Based on the coupling effect, the weight of per capita GDP decreased from 0.051 to 0.044, whereas the weight of urban road area per capita increased from 0.047 to 0.049. The global sensitivity weight was adopted to evaluate the human settlement suitability.

4.2. Temporal and Spatial Changes in Suitability for Each Subsystem

To reveal the suitability change over time and space, the suitability indexes were calculated for each subsystem in 30 provinces of China from 2000 to 2016, using global sensitivity weights in Table 2. In addition, the mean and standard deviation of yearly provincial suitability index values were calculated for each subsystem and are shown in Figure 4a,b, respectively.

From Figure 4a, the suitability index of China’s human, economy, society, and residence subsystems exhibited a significant increasing trend from 2000 to 2016. RSI grew the fastest from 27.9 in 2000 to 68.0 in 2016. NSI increased slightly from 37.3 in 2000 to 37.7 in 2016. Among the five subsystems, the suitability index was the lowest for the nature subsystem since 2008. Notably, NSI exhibited a significant drop from 2012 to 2013. The standard deviation of the provincial suitability index represented the provincial difference. From Figure 4b, the inter-provincial difference in ESI increased from 2000 to 2016, whereas that in RSI decreased. In 2016, the inter-provincial differences in each subsystem were as follows: economy subsystem (11.2) > population subsystem (9.9) > residence subsystem (9.7) > nature subsystem (9.5) > society subsystem (8.5).

The histogram in Figure 5 shows the suitability index of each subsystem in 2000 and 2016, as well as the average value between 2000 and 2016. Meanwhile, Figure 5 shows the suitability ranking for each subsystem in 2016 from the highest suitability to the lowest suitability.

From Figure 5a, the suitability index of the natural subsystem in 24 provinces was lower than that of other subsystems in 2016. In addition, NSI in 15 provinces decreased from 2000 to 2016, especially in Beijing, Tianjin, and Guangdong. As shown in Figure 5d, the SSI value in Shanghai dropped by 0.6 from 2000 to 2016. Except for this, the suitability of the human, economy, society, and residence subsystems improved in each province from 2000 to 2016. This was consistent with the trend of the suitability index of human settlement subsystems in China in Figure 4a.

According to the suitability ranking in Figure 5, it revealed that the province with a higher suitability in the economic subsystem had a relatively lower suitability in the natural subsystem. For example, Beijing, Shanghai, and Tianjin were in the top three in economic subsystem suitability, respectively. ESI values in these provinces were higher than 80. By comparison, the natural subsystem suitability in Beijing, Shanghai, and Tianjin ranked 30th, 25th, and 28th, respectively. NSI values in these provinces were less than 30. In addition, the economic subsystem suitability in Jiangsu, Zhejiang, and Guangdong ranked from 4 to 6, with suitability indexes ranging from 72 to 77. The natural subsystem suitability in Jiangsu, Zhejiang, and Guangdong was ranked at 23, 21, and 16, respectively. The NSI values in these provinces were between 33 and 38.

4.3. Temporal and Spatial Changes in Suitability for Comprehensive System

To evaluate HSSI changes over time and space, HSSI values in 30 provinces from 2000 to 2016 were calculated using the global sensitivity weight. In addition, the mean value and standard deviation of HSSI in 30 provinces over time were calculated and shown in Figure 6. Over the period from 2000 to 2016, the minimum value of HSSI at the provincial scale was 25.5, which appeared in Guizhou Province in 2000. In contrast, the maximum value of HSSI was 75.9, which appeared in Beijing in 2016. Based on the maximum and minimum values of HSSI, the level of human settlement suitability was divided into five levels, including low suitability (HSSI ≤ 35), relatively low suitability (35 < HSSI ≤ 45), medium suitability (45 < HSSI ≤ 55), relatively high suitability (55 < HSSI ≤ 65), and high suitability (HSSI > 65). The spatial distribution of suitability level at the provincial scale from 2000 to 2016 was shown in Figure 7.

As shown in Figure 6, the suitability of China’s human settlements increased significantly from 2000 to 2016, indicated by an increasing HSSI value from 34.5 in 2000 to 58.0 in 2016. Meanwhile, the inter-provincial difference in HSSI showed a downward trend. As shown in Figure 7, the suitability of human settlements improved in every province. Regarding the suitability level, 19 provinces (including Guizhou and Ningxia) exhibited low suitability, and 9 provinces (including Guangdong and Jiangsu) exhibited medium suitability in 2000. In contrast, 13 provinces (including Henan and Guizhou) had a moderate suitability, 12 provinces (including Tianjin and Guangdong) had a relatively high suitability, and 5 provinces (including Beijing and Jiangsu) had a high suitability in 2016. Based on the spatial distribution of HSSI in 2016, it was notable that the suitability level in the eastern provinces with a developed economy was significantly higher than that in the western provinces with an underdeveloped economy. For the growth rate of HSSI, it was higher in the economically underdeveloped regions in western China than that in the economically developed provinces in eastern China. For example, the top three provinces with the highest HSSI growth rate were Ningxia, Inner Mongolia, and Anhui. In contrast, the provinces with the lowest growth rate were Shanghai, Tianjin, Hainan, and Beijing.

5. Discussion

5.1. Sensitivity Analysis of Indicator

Using the EFAST algorithm, the indicator sensitivity was analyzed based on the data from 30 provinces in China from 2000 to 2016. Figure 8 shows the first-order sensitivity index (S_i) and global sensitivity index (S_ti) of each indicator. The black bar represented the first-order sensitivity index, and the red bar represented the coupling effect index among indicators, with the value of the difference between S_ti and S_i (S_ti − S_i). From Figure 8, the first-order sensitivity index of each indicator was slightly lower than the global sensitivity index. The coupling effect index varied from 0.001 to 0.028. It suggested that a certain degree of coupling relationship existed among human settlement suitability indicators. The first-order sensitivity index varied from 0.017 to 0.094, which was higher than the coupling effect index for each indicator. It indicated that the direct contribution of each indicator to the HSSI score was significantly higher than the indirect contribution. It also suggested that the selected indicators met the indicator selection principle of representativeness and independence. In consideration of the coupling effect among indicators, the sensitivity ranking sorted by the first-order sensitivity index had changed for two-thirds of indicators in comparison with that sorted by the global sensitivity index. The sensitivity ranking increased for 13 indicators, whereas it decreased for 7 indicators. It did not change for only 10 indicators. This suggested that although the coupling index was small, it should not be neglected when objectively calculating the weights of indicators for the HSSI score.

Based on Figure 8 and Table 2, it suggested that the most sensitive indicators calculated by the first-order and global sensitivity index exhibited high similarities. In consideration of the coupling effect among indicators, the global sensitivity weights can modify the first-order sensitivity weights, which can more effectively identify the sensitivity of indicators. From Table 2, the global sensitivity weight of per capita GDP was lower than that of urban road area per capita. It was consistent with the fact that except for the economic subsystem, other subsystems also played an important role in human settlement suitability improvement. It also indicated that the evaluation indicators interacted, and, thus, the coupling effect between indicators should not be neglected during the process of indicator determination [19,34]. Considering the coupling effect was a more objective and comprehensive idea when fixing weights for the indicators [33].

5.2. Factors Affecting Changes in Suitability for Each Subsystem

It was notable that NSI exhibited a significant drop from 2012 to 2013 (Figure 4). This was because the State Council of China passed the new “Ambient Air Quality Standards” in 2012, and it introduced PM2.5 as the new indicator for air quality assessment. Air pollution was a serious problem that seriously influenced human settlement suitability. After adopting the new air quality standard, the number of days when the air quality reaches or is better than Grade II (Day) in China was the lowest in 2012 and gradually increased from 2013. The gradual increase in NSI from 2013 to 2016 was attributed to the improvement in air quality and energy production capacity. It was also possible to improve the suitability by policies and technical interventions, such as environmental governance and technological advancements. Under the influence of multiple factors, such as rapid population growth and expansion of urban land use, per capita cultivated land area and water resources continued to decrease, which became the main factors restricting the suitability improvement of the nature subsystem.

From the perspective of spatial variations, NSI decreased in NSI in 15 provinces. It was mainly attributed to the decline in air quality, per capita cultivated land area and water resources [42]. For Guangdong, Hainan, Yunnan and Xinjiang, the decline in the nature subsystem suitability was due to the decrease in per capita cultivated land area. For Beijing, Shanghai, and the other 9 provinces, the declining suitability of the nature subsystem was due to the decline in air quality. With the rapid population growth, per capita cultivated land area decreased in 29 provinces from 2000 to 2016. Moreover, the number of days when the air quality reached or was better than Grade II decreased in 22 provinces. Although the national average precipitation was the highest in 2016 since 1951, per capita water resources in 10 provinces in 2016 were lower in comparison with those in 2010. This suggested that the water shortage directly affected the nature subsystem suitability. The SSI value in Shanghai dropped by 0.6 from 2000 to 2016. This was mainly due to the fast population agglomeration under limited land resources in Shanghai. Although the urban infrastructure and public services developed well in Shanghai, they have entered an overloaded stage at the per capita level [16]. In particular, the population in Shanghai grew rapidly from 16.09 million in 2000 to 24.2 million in 2016, but per capita urban road area and public transport vehicles per 10,000 people decreased in 2016 compared with those in 2000. From Figure 5, it is suggested that NSI has become the primary limiting factor for the coordinated development of human settlement subsystems, especially in economically developed provinces. It was essential to focus on the natural subsystem to promote the coordinated development of integrated human settlement systems.

5.3. Factors Affecting Changes in Suitability for Comprehensive System

For the human settlement system, HSSI in the eastern province featured a developed economy which was significantly higher than that of the western province with an underdeveloped economy. The main reasons are that, firstly, the eastern region was relatively flat and the climate was favorable. It was suitable for human habitation as well as industrial and agricultural development. Secondly, compared with the inland provinces, China attached importance to the economic development in eastern coastal provinces since the reform and opening up of the country. It promoted the development of education, transportation, and residential facilities, which further the suitability development of human, social, and residential subsystems [43]. By comparison, the growth rate of HSSI was higher in the economically underdeveloped region in western China than in the economically developed province in eastern China. This was consistent with spatial and temporal changes in China’s provincial sustainable development goal index from 2000 to 2015 [44].

In combination with changes in the suitability index of subsystems, the residence subsystem contributed the most to the improvement of HSSI, followed by the society and economy subsystem. The residence subsystem accounted for more than 20% of the HSSI growth value in 30 provinces, with the highest value of 46.7% in Shanghai. Except for Beijing and Shanghai, the contribution rate of the social subsystem to HSSI growth was higher than 20%, and the highest is 36.2% (Hebei). Except for Beijing and Shanghai, the total contribution rate of the social and residential subsystems to HSSI growth exceeds 50%, and the highest is 74.7% (Hainan). In addition, the contribution rate of the economic subsystem to HSSI growth is higher than 20% in 10 provinces, including Beijing and Shanghai, and the highest is 37.6% (Shanghai). Compared with other subsystems, the natural subsystem had the lowest contribution rate to HSSI growth. NSI values in 15 provinces declined, which became a limiting factor for the improvement of the comprehensive suitability of human settlements and the coordinated development of the subsystems. Human settlements are the geographic space where humans gather for activities, and the natural subsystem provides the space and resources foundation for the development of human settlements [11]. The excessive development of population, economy, society, and residence subsystems negatively affects the natural subsystem, such as the decline in air quality, drastic reduction in arable land, and the increase in water shortages. Therefore, it is urgent to optimize the land-use space and the allocation of restricted resources according to the development demand of human settlements. Furthermore, it is essential to promote the coordinated development of human, economy, society, residence, and nature subsystems, and, hence, improve the suitability of the integrated human settlements.

6. Conclusions

A set of human settlement suitability indexes (i.e., NSI, HSI, ESI, SSI, RSI, and HSSI) were developed to evaluate the suitability of human settlements. The coupling effect was taken into consideration during the process of indicator selection and weight determination. Although the coupling relationship among indicators was uncertain, it was sufficient to generate high sensitivity. Based on the EFAST algorithm, the weights of indicators were calculated using the global sensitivity analysis, which combined the influence of main effects and coupling effects among indicators. The suitability change in human settlements over time and space was analyzed at both subsystems and comprehensive system levels in 30 provinces of China from 2000 to 2016. The major conclusions were as follows:

(1)

HSSI values increased significantly in all 30 provinces from 2000 to 2016. Among the five subsystems, the suitability index of the residence subsystem in China exhibited the fastest growth, followed by the society and economy subsystems. The suitability index in eastern provinces with a developed economy was higher than in western provinces with an underdeveloped economy. In contrast, the growth rate of HSSI in eastern provinces was significantly higher than that in western provinces.

(2)

The inter-provincial difference in HSSI decreased from 2000 to 2016, indicated by a decreasing difference value between the maximum and minimum provincial HSSI. The inter-provincial difference decreased in RSI, whereas it increased in ESI. The inter-provincial difference in NSI, HSI, and SSI fluctuated and increased slowly.

(3)

NSI in 24 provinces was the lowest compared with other subsystems in 2016. In addition, NSI in 15 provinces decreased from 2000 to 2016. The declining air quality and decreasing per capita cultivated land area were the primary reasons, and the decrease in per capita water resources was the secondary reason. Under increasingly severe constraints of resources and the environment, the suitability of the natural subsystem has become a limiting factor for the improvement of human settlements, especially in economically developed provinces, such as Beijing, Shanghai, and Guangdong.