Multi-Scale Spatiotemporal Characteristics Assessment of Water and Land Resources Ecological Security in China’s Main Grain-Producing Areas

Abstract

Water and land resources, as the material foundation of food production, are essential for national food security. Current research has not yet explored the spatiotemporal features of water and land resources ecological security (WLRES) at the urban scale. To fill this gap, this study evaluated WLRES across 180 cities in China’s main grain-producing areas (MGPAs) from 2005 to 2020. A WLRES evaluation system was developed based on the DPSIR framework and the CRITIC method. The Moran’s I and kernel density estimation were utilized to analyze the spatial distribution, variation trends, and spatial autocorrelation of WLRES from different scales. The results demonstrate the following: (1) WLRES in the MGPAs exhibited a fluctuating upward trend, transitioning from “relatively low ecological security” to “moderate ecological security.” (2) The spatial distribution of WLRES was characterized by higher values in the northeast and southwest regions and lower values in the central region, with spatial heterogeneity gradually intensifying. (3) From 2005 to 2016, WLRES exhibited significant positive spatial autocorrelation: cities with high ecological-security levels were concentrated in the northern region, whereas those with low ecological-security levels were clustered in the central and southern of Huang-Huai-Hai Basin. Over time, this positive spatial autocorrelation weakened and eventually vanished. Our research can provide feasible policy references for improving the sustainable development of WLRES in the MGPAs.

1. Introduction

Water and land resources constitute the fundamental basis for food production and human survival [1,2]. However, in recent years, rapid population growth, agricultural overexploitation, urbanization, and climate change have caused ecological issues such as soil erosion, land desertification, water shortages, soil and water pollution, and spatial-temporal mismatches of water and land resources, threatening the sustainability of soil and water resources [3,4,5]. Therefore, assessing the stability and sustainability of water and land resources’ ecological environment is essential for ensuring food security [6].

Introduced by the International Institute for Applied Systems Analysis (IIASA) in 1989, ecological security denotes the condition where human life, health, well-being, and adaptive capacity remain unthreatened [7]. During this period, ecological security was mainly judged by assessing chemical indicators related to pollution, but the overall health of the ecosystem was neglected [8]. As understanding of ecosystem complexity deepened, the concept of ecological security expanded to encompass natural, economic, and social dimensions. Currently, ecological security refers to the structural and functional stability of ecosystems that can meet the socio-economic development needs of both present and future generations [9].

Domestic and international studies have provided thorough evaluations of the ecological security of both water and land resources. Methodologically, the literature relies on indicator-based integrated assessment, ecological footprint accounting, and landscape-pattern analysis. Thematically, research addresses ecological risk, ecological health, ecosystem-service valuation, ecological pressure, and landscape pattern dynamics [9,10]. For the ecological security of water resources, Qiu et al. used spatial autocorrelation analysis to explore the regional clustering characteristics of ecological security of water resources and to assess the impacts of ecosystem changes on the neighboring regions [11]. Yang et al. employed an enhanced vulnerability-exposure-disaster-loss model to assess water-resource ecological risk, identify its drivers, forecast trends, and quantify impacts [12]. Hu et al. integrated ecosystem-service indices and land-use change into water-resource security assessment to identify key determinants more comprehensively [13]. For the ecological security of land resources, Su et al. applied catastrophe theory to weight indicators, thereby capturing mutation characteristics in ecological security overlooked by conventional methods [14]. Wu and Xie introduced the improved B-P neural network into ecological security assessment, which can effectively integrate different types of data and mine the potential laws in the data to improve the accuracy of the assessment [15].

Water and land resources are highly interdependent; the shortage of either can limit the effective use of the other, reducing overall resource utilization efficiency. Thus, joint assessments of these resources are essential [16,17]. Previous research has focused on analyzing either the structure or the function of the water and land resources ecological security (WLRES) system, particularly in evaluating the matching status and carrying capacity of water and land resources [18,19]. Matching status assessments, employing matching coefficients, coupling coordination models, Gini coefficients, and data envelopment analysis (DEA), quantify resource coupling and balance, thereby revealing spatial equilibrium and utilization efficiency [20,21,22,23]. In recent years, the concept of blue-green water has provided a theoretical basis for accurately identifying the differences in the matching status of water and land resources between irrigated and rain-fed cropland, exploring the utilization efficiency of irrigation water [17]. The carrying capacity of water and land resources, which indicates their ability to sustain socio-economic development given specific resource endowments and utilization efficiency, is crucial for resource planning and utilization [24]. Ecological footprint and comprehensive evaluation are commonly used to quantitatively analyze the interactions and feedback mechanisms among water and land resources, human activities, and the ecological environment to determine the carrying capacity [25,26].

However, current ecological security assessments predominantly focus on individual water or land resources. They overlook the combined role of these resources in agriculture, ecology, and socio-economic development, as well as the influence of social, economic, and ecological factors on WLRES [27]. Meanwhile, the existing studies on WLRES are mostly focused on watersheds, provinces, or specific regions. There is a lack of unified and comprehensive ecological security assessments at the urban scale, which hinders the formulation of relevant policies [28,29]. Additionally, while many studies concentrate on environmental pollution and landscape patterns, there is a shortage of WLRES assessments specifically targeting the agricultural sector [22].

Aiming at the deficiencies in existing studies, this paper selects 180 cities in China’s main grain-producing areas (MGPAs) as research objects. It evaluates the WLRES level by constructing an assessment index system, explores the spatial distribution characteristics of WLRES from multiple scales, and proposes suggestions for the sustainable development of water and land resources based on the research results (Figure 1). The study includes the following contributions: (1) In terms of research scale, it uses cities as the basic evaluation units and assesses the WLRES in MGPAs from the perspectives of the overall research region, subregions, and provinces; (2) in terms of research content, it explores the impact of agricultural production on the WLRES in view of the special nature of agriculture; and (3) in terms of indicator selection, it chooses indicators that are sensitive to changes in water and land resources as well as comprehensive indicators that reflect the coupling and overall effects of these resources.

2. Materials and Methods

2.1. Study Area

In 2003, China’s Ministry of Finance designated a total of 13 provincial-level administrative regions as MGPAs, including Inner Mongolia, Heilongjiang, Jilin, Liaoning, Hebei, Shandong, Henan, Jiangxi, Anhui, Jiangsu, Hubei, Hunan, and Sichuan. The MGPAs are China’s top grain production bases, with MGPAs accounting for 78.5% of the country’s total grain production in 2020. From 2003 to 2020, China’s total grain production increased from $4.31\times 10^8$ t to $6.69\times 10^8$ t, of which MGPAs contributed 92.5% of the increment. However, under the pressure of grain production increase, the WLRES situation of MGPAs is not optimistic. In order to increase grain production, a large number of chemical products such as fertilizers, pesticides, and agricultural films are used in agricultural production. In 2020, China’s fertilizer application reached about $5.25\times 10^7$ t, and the intensity of fertilizer application is about three times the world average. The annual use of pesticides is as high as $1.2\times 10^6$ t, and the utilization rate is less than 30%. Additionally, the area of agricultural films used has exceeded $1\times 10^8$ mu, with about $3\times 10^5$ t of agricultural films left in the land every year. The inefficient use of pesticides and fertilizers and the residue of agricultural films pollute water and land resources, destroy land structure, reduce the fertility of arable land, and lead to crop yield reduction [30,31]. Recently, the center of gravity of China’s grain has continued to move northward, and the over-concentration of grain production in some water-scarce provinces has exacerbated the contradiction between the supply and demand of agricultural water, threatening food security [32]. Hence, assessing the ecological security of water and land resources to ensure their stable and sustainable use has become vital for food security.

Combined with the availability of data, this paper selects 180 cities in MGPAs as the study area. In order to facilitate the analysis, the study area is divided into three subregions based on geographical characteristics: the northern region–Inner Mongolia Autonomous Region, Heilongjiang Province, Liaoning Province, Jilin Province; the Huang-Huai-Hai Basin—Hebei, Shandong, Henan, Jiangsu, and Anhui Provinces; and the Yangtze River Basin—Sichuan, Hubei, Hunan, and Jiangxi Provinces, as shown in (Figure 2). The northern region has a temperate monsoon climate, warm and rainy in summer and cold and dry in winter. In 2020, it had 34.32 million hectares of arable land (20.3% of the national total) and 7.6% of the nation’s water resources. The Huang-Huai-Hai Basin has a semi-arid, semi-humid climate zone and had 50.15 million hectares of arable land (30% of the national total) in 2020, with 7.19% of the national water resources. The Yangtze River Basin, with a warm and humid subtropical monsoon climate, had 32.29 million hectares of arable land (19.1% of the national total) in 2020 and 24.76% of the nation’s water resources.

2.2. Research Methodology

2.2.1. Construction of WLRES Indicator System

WLRES is a complex dynamic system that integrates social, economic, and ecological environments and resources. Within the system, rapid socio-economic development intensifies human-induced ecological impacts, altering WLRES structure and function. These changes affect human life and production, prompting ecological improvement measures and fostering shifts in socio-economic development patterns [30]. The DPSIR (driving force-pressure-status-impact-response) model can reflect the interconnection between the factors within the WLRES system and comprehensively reflect the state of the WLRES system. Thus, this paper chooses the DPSIR model to construct the indicator system.

Based on scientific, systematic, and feasibility principles, referring to the previous research results and combining them with the actual situation of MGPAs, the evaluation indexes were selected from the social, economic, environmental, and ecological perspectives, and the ecological security assessment index system of water and land resources consisting of 25 indexes was established based on the DPSIR model [28,33], as shown in

Table 1. Indicator system for WLRES evaluation.

Subsystems	Indicator	Explanation of Indicator	Symbol
Driving force (D)	Population density (D1)	Total population/total area	−
	Natural population growth rate (D2)	Birth rate–death rate	−
	GDP per capita (D3)	GDP/total population	+
	Urbanization rate (D4)	Non-farm population/total population	+
Pressure (P)	Fertilizer application per unit of cultivated area (P1)	Fertilizer application/cultivated land area	−
	Percentage of built-up land (P2)	Urban built-up land area/total urban area	−
	Per capita residential water consumption (P3)	Residential water consumption/total population	−
	Water consumption per 10,000 CNY GDP (P4)	Water consumption/GDP	−
	Irrigation water consumption per unit cultivated land area (P5)	Irrigation water use/cultivated land area	+
Status (S)	Land and water resources match (S1)	Amount of water available for arable land [21]	+
	Vegetation water and land utilization (S2)	Net Primary Productivity (NPP): The net rate of carbon assimilation by vegetation per unit area and time	+
	Effective irrigation rate of cropland (S3)	Effective irrigated area of arable land/area of arable land	+
	Normalized vegetation index (S4)	Normalized vegetation index (NDVI): An index reflecting vegetation coverage	+
	Annual precipitation (S5)	Total precipitation in the region during the year	+
	Modulus of water production (S6)	Total water resources/total area	+
	Land resettlement rate (S7)	Cultivated land area/total area	−
	Water resources per capita (S8)	Total water resources/total population	+
Impact (I)	Food production per capita (I1)	Food production/total population	+
	Disposable income per rural resident (I2)	Primary and redistributed income received by rural households	+
	Level of agricultural mechanization (I3)	Total power of agricultural machinery/cultivated area	+
Response (R)	Non-hazardous treatment rate of domestic waste (R1)	Non-hazardous treatment rate of domestic waste	+
	Sewage treatment rate (R2)	Sewage treatment rate	+
	Value added of tertiary sector as a share of GDP (R3)	Tertiary value added/GDP	+
	Greening coverage of built-up areas (R4)	Green coverage of urban built-up areas/area of built-up areas	+
	Investment in environmental pollution control (R5)	Investment in environmental pollution control	+

Note: “+” denotes a “larger-is-better” indicator, while “−” denotes a “smaller-is-better” indicator.

.

In the driving force subsystem, population density (D1) and natural population growth rate (D2), GDP per capita (D3), and urbanization rate (D4) were chosen to represent the potential driving effects of population size, economic development, and urbanization level on ecological and environmental changes, respectively. The pressure subsystem primarily evaluates the impacts of environmental pollution and resource consumption caused by human activities on WLRES. The state subsystem contains the structure, function, and state indicators of land and water resources as well as the coupled and coordinated efficiency indicators of land and water resources, reflecting the ecological carrying capacity of land and water resources. The impact subsystem mainly includes the impacts of changes in the ecological environment of water and land resources and economic development on crop yields (I1), farmers’ living standards (I2), and agricultural infrastructure construction (I3). The response system is the measures taken to cope with environmental changes and improve ecological security, which mainly include the harmless treatment of pollutants (R1, R2), the development level of non-polluting tertiary industries (R3), the greening coverage rate (R4), and the investment in environmental pollution control (R5). The detailed influence mechanisms of the indicators on WLRES are presented in Table S1.

2.2.2. MGPAs Overall Level Assessment

In this paper, the WLRES index is used to reflect the security level, and the ecological security grade is divided according to the value of the WLRES index. The calculation process includes the following steps: (1) standardization of index data, (2) determination of index weights, (3) calculation of the WLRES index, and (4) division of the WLRES grade. The specific step contents are as follows:

Step 1: Data standardization

In order to eliminate the interference of the raw data of the indicators on the calculation results due to the different scales and attributes, the raw data are standardized for the positive and negative attributes of the indicators calculation.

$$\begin{array}{c}\mathrm{Positive\; indicator}:X_{ij}^{\prime }=\left(X_{ij}-min\left(X_j\right)\right)/\left(max\left(X_j\right)-min\left(X_j\right)\right)\\ \mathrm{Negative\; indicator}:X_{ij}^{\prime }=\left(\mathit{max}(X_j)-X_{ij}\right)/\left(max\left(X_j\right)-min\left(X_j\right)\right)\end{array}$$

(1)

where $X_{ij}$ is the value of the $j$ evaluating value mark of the $i$ evaluating object, ${X^{\prime }}_{ij}$ is the standardized value of $X_{ij}$, $\mathrm{m}\mathrm{a}\mathrm{x}\left(X_j\right)$ is the maximum value of the $jth$ evaluating indicator, and $\mathrm{m}\mathrm{i}\mathrm{n}\left(X_j\right)$ is the minimum value of the $jth$ evaluating indicator.

Step 2: Calculation of indicator weights

The CRITIC assignment method is an objective assignment method proposed by [34] to determine the weights based on the contrast and conflict between the indicators. This method takes into account the degree of difference between the evaluation objects and the correlation between the indicators at the same time, as well as reflecting the importance of the indicators through the indicator’s information-carrying capacity, and therefore has been widely used in the comprehensive evaluation of multiple indicators [35]. In this paper, the CRITIC assignment method is used for weight calculation, and the method is as follows:

Comparative calculations:

$$\begin{array}{c}{B}_j=\sqrt{{\displaystyle \sum _{i=1}^m}\left({X^{\prime }}_{ij}-{{\overline{X}}^{\prime }}_{ij}\right)/(n-1)}\end{array}$$

(2)

Conflicting calculations:

$$\begin{array}{c}{A}_j={\displaystyle \sum _{i=1}^m}\left(1-r_{ij}\right)\end{array}$$

(3)

Calculation of information carrying capacity:

$$\begin{array}{c}{C}_j=B_j\times A_j\end{array}$$

(4)

Calculation of indicator weights:

$$\begin{array}{c}{W}_j=C_j/{\displaystyle \sum _{i=1}^n}{C}_j\end{array}$$

(5)

where $B_j$ is the contrast between indicators, $m$ is the number of evaluation objects, $n$ is the number of evaluated markers, ${{\overline{X}}^{\prime }}_{ij}$ is the mean value of each standardized evaluation indicator, $A_j$ is the degree of conflict between the indicators, $r_{ij}$ is the correlation coefficient between the $i$ indicator and the $j$ indicator, and $W_j$ is the weight of the $j$ indicator.

Step 3: WLRES index calculation

Combining the standardized indicator data obtained from Equation (1) and the weight data obtained from the CIRTIC assignment method in Equation (5), the WLRES index is calculated as follows:

$$\begin{array}{c}{S}_i={\displaystyle \sum _{j=1}^n}{W}_j\times {X^{\prime }}_{ij}\end{array}$$

(6)

where $S_i$ is the WLRES index value of the first $i$ evaluation target, and the value range is $\left[0,1\right]$. The larger $S_i$ is, the safer the water and land resources ecosystem.

Step 4: WLRES Classification

The current methods for determining thresholds include the Equal Interval method, the Cloud Modeling method, the Expert Rating method, and the Natural Breaks (Jenks) method [36,37,38]. The Natural Breaks method classifies the data based on the data itself from the perspective of clustering, which maximizes the similarity within each class and maximizes the difference between different classes [39]. Therefore, we choose the Natural Breaks method to classify the ecological security of water and land resources into five grades based on the calculated $S_i$ results, as shown in

Table 2. Grading Standards for WLRES.

Range of Index	Grade	Security Level	Ecosystem Characterization
${0\le S}_i<0.389$	I	Low ecological security	The structural functioning of the system is seriously undermined, to the detriment of socio-economic development
${0.389\le S}_i<0.481$	II	Relatively low ecological security	Destruction of the system’s structural functions and serious pollution, which to some extent constrains socio-economic development
${0.481\le S}_i<0.560$	III	Moderate ecological security	The system has a basic structure and function and is moderately polluted, and the ecosystem shows a deteriorating trend, which is not conducive to sustainable socio-economic development
${0.560\le S}_i<0.644$	IV	Relatively high ecological security	The system is more structurally and functionally intact, slightly contaminated, and conducive to stable socio-economic development
${0.644\le S}_i\le 1$	V	High ecological security	The system is structurally and functionally intact, uncontaminated, and capable of supporting long-term sustainable socio-economic development

. The spatial distribution maps of WLRES grades for the different cities were generated using ArcMap 10.8.2.

2.2.3. Kernel Density Estimation

Kernel density estimation is a nonparametric probability density function for estimating the distribution of random variables. This method has the advantage of not relying on specific probability distribution assumptions and can be applied to any type of data [40]. In this paper, Gaussian kernel density estimation is used to analyze the distribution characteristics and dynamic trends of WLRES, to reveal the distribution and dynamic characteristics of WLRES within the whole and subregions of MGPAs, and to find outliers. The specific formulas are as follows:

$$\begin{array}{c}f\left(x\right)={\displaystyle \frac{1}{mh}}{\displaystyle \sum _{i=1}^m}K\left({\displaystyle \frac{S_i-\overline{S}}{h}}\right)\end{array}$$

(7)

where $f\left(x\right)$ is the WLRES kernel density function, $h$ is the bandwidth, $\overline{S}$ is the mean value of the WLRES exponent, and $K\left(·\right)$ is the Gaussian kernel function.

2.2.4. Moran’s I

Due to the influence of human activities and the flow of ecological elements, neighboring regions tend to exhibit similar ecological security characteristics, indicating spatial autocorrelation [40]. Therefore, we employ Moran’s I for spatial autocorrelation analysis to access interactions of WLRES and its influencing factors across neighboring cities. Among them, global Moran’s I is a comprehensive evaluation of the spatial data of the whole study area, which reacts to whether the study area as a whole is characterized by aggregation or disaggregation as well as the degree of significance and strength of this trend. Local Moran’s I identifies specific ecological security clusters, highlighting spatial differences and the significance of variations between cities and their neighbors. Local Moran’s I spatial maps for each city were produced using ArcMap 10.8.2. The specific calculations are as follows.

$$\begin{array}{c}I={\displaystyle \frac{m{\sum }_{i=1}^m{\sum }_{j=1}^m{w}_{ij}\left(S_i-\overline{S}\right)\left(S_j-\overline{S}\right)}{{\sum }_{i=1}^m{\sum }_{j=1}^m{w}_{ij}{{\sum }_{i=1}^m\left(S_i-\overline{S}\right)}^2}}\end{array}$$

(8)

$$\begin{array}{c}{I}_i={\displaystyle \frac{{\sum }_{j=1}^m{w}_{ij}\left(S_i-\overline{S}\right)\left(S_j-\overline{S}\right)}{{\sum }_{j=1}^m{w}_{ij}{{\sum }_{k=1}^m{w}_{ik}\left(S_k-\overline{S}\right)}^2}}\end{array}$$

(9)

$$\begin{array}{c}{w}_{ij}=1/d_{ij}^2\end{array}$$

(10)

where $I$ and $I_i$ denote global and local Moran indexes, respectively, ranging from −1 to 1. Values $>0$ indicate positive spatial autocorrelation, i.e., cities with similar levels of WLRES tend to cluster together; values $<0$ indicate negative autocorrelation, i.e., cities with similar levels of WLRES tend to be excluded from each other; values $=0$ indicate there is no spatial autocorrelation, and the WLRES values of adjacent cities are randomly distributed. $w_{ij}$ is the spatial weight between region $i$ and region $j$, and $d_{ij}$ is the distance between region $i$ and region $j$.

2.3. Data Source

The data used in this study include statistical data, remote sensing data, and meteorological data for China’s MGPAs from 2005 to 2020. The data types, names, sources, and explanations are provided in Table S2. Where missing values occurred in the statistical series, linear interpolation was applied to estimate the intermediate data points by weighing their two nearest neighbors in proportion to their temporal distance. Although this method is less precise than more sophisticated algorithms, it is computationally straightforward and efficient, and it adequately preserves the underlying trend of the time-series data.

3. Results and Analysis

3.1. Multi-Scale Spatiotemporal Variation Analysis of WLRES

3.1.1. The Weights Results

The weight calculation result is shown in Figure 3. The results reveal the state subsystem (S) to be the most influential, followed by the response (R) and driving-force (D) subsystems, whereas the pressure subsystem (P) exerts the weakest effect. At the indicator level, land reclamation rate (S7) and population density (D1) exhibit the highest weights, signifying that agricultural expansion and anthropogenic activities constitute the principal determinants of WLRES variability. By contrast, the markedly lower weights of water consumption per 10,000-yuan GDP (P4) and per-capita water resources (S8) suggest that absolute water availability is not a pivotal driver of ecological security issues within the MGPAs.

3.1.2. Spatiotemporal Pattern and Variation Trends of WLRES in MGPAs

For the MGPAs as a whole, the WLRES index shows a fluctuating upward trend, which is divided into three stages: rising, declining, and the disparities in the WLRES levels among different cities progressively increasing, as shown in (Figure 4).

In the first stage, from 2005 to 2011, driven by improvements in the response subsystem, the WLRES index for the MGPAs rose from 0.470 to 0.504 (+7.34%), shifting the ecological security grade from relatively low ecological security to moderate ecological security. During this period, China introduced a series of arable land protection policies, which increased the amount of arable land in the MGPAs and effectively improved the quality of arable land. Water-saving irrigation technologies have been widely promoted. Efficient drip and sprinkler systems are rapidly supplanting traditional flood irrigation. This shift enhances agricultural water-use efficiency, reduces withdrawals, and consequently elevates the WLRES index.

The reduction of the state subsystem and impact subsystem in the second stage from 2011 to 2015 resulted in a decreasing trend of WLRES, which fell back from 0.504 to 0.482 (−4.43%). Under the pressure of long-term food production, MGPAs have adopted a high-intensity land development model at the expense of the ecological environment. This model has led to the over-consumption of agricultural production resources, causing land structure destruction, falling water tables, and biodiversity loss [41]. At the same time, the misuse of chemical inputs such as fertilizers, pesticides, and agricultural films has exacerbated surface-source pollution and led to a serious degradation of the ecological service functions of water and land resources.

In the third stage, from 2015 to 2020, WLRES increased rapidly from 0.482 to 0.518 (+7.39%), with all subsystems except the state subsystem improving. The government introduced policies to establish a “trinity” protection system for water and land resources, optimizing agricultural production and promoting water-efficient crops. Agricultural water pricing reforms also boosted water resource efficiency, aiding the ecological recovery of MGPAs’ water and land resources.

In terms of subsystems, all subsystems except the state subsystem show an upward trend from 2005 to 2020. Specifically, the drive (D) subsystem increases from 0.117 in 2005 to 0.119 in 2020, an increase of 1.7%. The pressure (P) subsystem shows a fluctuating upward trend, improving from 0.0544 to 0.0556, an improvement of 2.20% but still at a low level. The state (S) subsystem fluctuates constantly, with an overall decreasing trend, from 0.148 in 2005 to 0.138 in 2020, a decrease of 6.54%. The impact (I) subsystem improves from 0.0302 to 0.0327, an increase of 8.14%. The response (R) subsystem continues to improve and has the largest improvement, from 0.121 in 2005 to a peak of 0.169 in 2020, an improvement of 38.75%.

To visualize the spatial distribution characteristics of WLRES in each city within the MGPAs, the years 2005, 2010, 2015, and 2020 were selected as typical years in this study, and the WLRES levels of different cities were visualized (Figure 5).

The spatial distribution of WLRES levels varies among cities, with significant spatial heterogeneity, and the distribution of WLRES levels is characterized by “high in the southwest and northeast and low in the middle.” In a typical year, the cities with relatively high ecological security and moderate ecological security levels are mainly concentrated in the northern region and western region in the Yangtze River Basin. The relatively low ecological security and low ecological security cities are concentrated in the central and southern parts of the Huang-Huai-Hai Basin and the eastern part of the Yangtze River Basin.

The overall change in the proportion of cities with different WLRES levels was not significant, but some cities’ own WLRES levels changed significantly. For instance, Ganzhou’s WLRES level improved from “low ecological security” (0.373) in 2005 to “relatively low ecological security” (0.478) in 2010, then to “moderate ecological security” (0.517) in 2015, and finally to “relatively high ecological security” (0.585) in 2020. Yingtan’s WLRES level increased significantly from “relatively low ecological security” (0.444) in 2005 to “relatively high ecological security” (0.562) in 2010. In contrast, Benxi’s WLRES level declined from “high ecological security” in 2010 to “relatively high ecological security” in 2015 and then to “moderate ecological security” in 2020, indicating damage to Benxi’s WLRES during economic development and agricultural upgrading.

3.1.3. Spatiotemporal Pattern of WLRES in Different Sub-Regions

In order to facilitate the analysis of trends in urban WLRES characteristics, the WLRES kernel density map (Figure 6) was used to represent the evolutionary trends in the distributional characteristics of urban WLRES within the MGPAs and the three sub-regions.

From the northern region (Figure 6b), the peak of the WLRES kernel density curve gradually shifted to the left from 2005 to 2020, with a significant increase in the height of the peak and a continuous narrowing of the width of the curve coverage. This indicates that the WLRES in the northern region has decreased during the study period, but the urban WLRES is still concentrated at the moderate ecological security level [42]. Abundant natural resources and low population density initially sustained high WLRES across the northern region. However, intensive agricultural expansion has recently accelerated soil erosion, impaired water quality, and induced black-soil desertification. These processes concurrently reduce terrestrial carbon storage and erode the ecosystem’s capacity for water retention and regulation, thereby driving WLRES downward [43].

From the Yangtze River Basin (Figure 6c), kernel density function curve shows an obvious left trailing pattern, with the peak position slightly shifted to the left but remaining at about 0.5 and the peak height showing the alternating change of “falling-rising-falling.” By 2020, the peak is lower, and the bandwidth has widened. This shows that the WLRES level of cities in the Yangtze River Basin has been reduced but can still be maintained at moderate ecological security level. Additionally, inter-city WLRES disparities are widening, with some cities concentrating their distribution at a low WLRES level; the problem of polarization is serious. The Yangtze River Basin benefits from abundant water, vast wetlands, and high vegetation cover, which collectively sustain elevated WLRES. However, excessive resource extraction and pollutant emissions have impaired the water and land resources system. Moreover, limited economic capacity in several cities constrains ecological investment, thereby widening WLRES disparities across the region [44].

From the Huang-Huai-Hai Basin (Figure 6d), the peak of the kernel density function curve shifts to the right constantly, the peak height fluctuates and rises, and the left trailing tail shifts to the left and elongates. The results show that the WLRES level is gradually increasing, and the absolute difference of ecological security level between cities is decreasing, while there are some cities with WLRES significantly lower than the average level, and the gap is gradually widening [45]. The Huang–Huai–Hai Basin holds only 7.2% of China’s water yet supports 31.6% of its population, 28.8% of the GDP, and 42% of arable-land demand. Rapid demographic and economic growth have encroached on farmland and intensified soil erosion and salinization, impairing land ecological functions and agricultural productivity. In order to cope with the above problems, the local government has taken a series of countermeasures including improving the synergistic mechanism of water and land conservation work and establishing a cross-regional collaboration mechanism. Concurrently, the eastern and middle routes of the South-to-North Water Diversion Project deliver 6–9 billion m³ of water annually to this region, substantially alleviating water scarcity. These coordinated interventions have progressively elevated ecological security, so that by 2020, the level of overall WLRES in the Huang-Huai-Hai Basin has been close to the overall average of the MGPAs.

3.1.4. Spatiotemporal Pattern of WLRES in Different Provinces

Figure 7 presents the WLRES for 13 provinces. Jiangxi, Hebei, Henan, and Anhui experienced rising WLRES, with increases of 0.104, 0.085, 0.069, and 0.028, respectively. Jiangxi saw the most significant growth (26.43%), yet the disparity in WLRES among its cities expanded, exhibiting a “high south, low north” pattern by 2020. In contrast, WLRES in Jiangsu, Jilin, Liaoning, and Heilongjiang showed a marked decline. Unlike the general downward trend in the three northeast provinces, northern Jiangsu cities saw WLRES improvements, while the south decreased, widening the provincial WLRES disparity. Inner Mongolia’s WLRES exhibited irregular fluctuations, with an average decrease of 0.78% every five years. Over the period 2005–2020, both Hubei and Hunan exhibited comparable WLRES distributions and trends, with the index declining by 4.65 and 0.56%, respectively. The larger reduction in Hubei suggests that the province faced relatively higher WLRES risks during this period. Sichuan’s WLRES rose by 3.02% (2005–2010), then dropped by 10.81% over the next decade. Zigong city’s WLRES consistently declined, significantly lagging behind the provincial average and substantially undermining Sichuan’s WLRES. Shandong’s WLRES decreased by 0.063 (2005–2015), then recovered by 0.027. In 2005, its southwestern cities had considerably lower WLRES than other regions. Subsequently, these cities’ WLRES improved continuously, overtaking other areas that displayed a downward trend.

3.2. WLRES Spatial Autocorrelation Analysis

3.2.1. Overall Scale Spatial Autocorrelation Analysis of MGPAs

The results of the global autocorrelation analysis are shown in

Table 3. Global Moran’s I of WLRES.

Year	I	z	p	Year	I	z	p
2005	0.188	8.811	0	2013	0.063	3.086	0.002
2006	0.165	7.767	0	2014	0.049	2.494	0.013
2007	0.14	6.619	0	2015	0.047	2.386	0.017
2008	0.115	5.477	0	2016	0.032	1.721	0.085
2009	0.1	4.782	0	2017	0.012	0.799	0.424
2010	0.086	4.145	0	2018	−0.002	0.17	0.865
2011	0.074	3.611	0	2019	−0.011	−0.24	0.81
2012	0.069	3.36	0.001	2020	−0.018	−0.55	0.582

. From 2005 to 2015, $p\le 0.05$, the I value decreases gradually but is always greater than 0, and z $>1.96$. This result indicates that there is a tendency for cities with similar WLRES levels to be clustered in the MGPAs, but this clustering tendency decreases constantly. In the WLRES system, water and land pollutants from one region can spread via the water cycle, impacting environmental quality across neighboring areas. Meanwhile, densely populated or economically developed regions, driven by resource demands, may shift resource acquisition pressures to other regions through cross-regional allocation. The negative effects of this process can spill over spatially due to factor flow, leading to convergence in the WLRES status of neighboring cities [25]. However, under the background of global warming, the precipitation pattern in the MGPAs has changed, and the non-uniformity of spatial and temporal distribution of precipitation has increased significantly. This phenomenon leads to the coexistence of floods and droughts, which exacerbates the spatial heterogeneity of water and land resources. With the rapid economic development, the increase of regional economic level disparity exacerbates the gap between cities’ ability to develop, utilize, and protect water and land resources, further reducing the overall spatial autocorrelation of the MGPAs. z $<1.96$ and $p>0.05$ in 2016 and later, indicating that the WLRES does not have significant spatial autocorrelation.

3.2.2. Sub-Region Scale Spatial Autocorrelation Analysis

Our study uses cities as basic units to measure the local spatial autocorrelation (LISA) and its significance of WLRES across 180 cities in three sub-regions (Figure 8), analyzing WLRES spatial clustering and its causes. The high-high aggregation area (HH) means that both the city and its neighboring WLRES levels are high. Low-high aggregation area (LH) is where the city has low WLRES levels but is surrounded by areas with high WLRES levels. Low-low aggregation (LL) cities and their surrounding WLRES levels are both low. High-low aggregation (HL) cities have high WLRES levels but are surrounded by areas with low WLRES levels. Non-significant indicates that the evaluated city’s WLRES is not significantly related to that of its neighboring cities.

As shown in Figure 8, the (HH) and (LH) aggregation cities are mainly distributed in the eastern part of the northern region, and the distribution of (LL) and (HL) aggregation cities is mainly concentrated in the central and southern part of the Huang-Huai-Hai Basin. From 2005 to 2015, the number of cities with significant (HH) and (LL) WLRES gradually decreases, while the number of cities with (HL) and (LH) aggregation increases, and the local spatial autocorrelation of some cities changes from positive correlation to negative correlation, which indicates that the spatial positive correlation between neighboring cities gradually decreases, and the difference of WLRES between neighboring cities increases.

The eastern part of the northern region has superior and relatively concentrated water and land resource endowments, thus creating a spatially (HH) aggregation situation. However, as the region’s economy relies on heavy industry and agriculture, and industrial activities are regionally concentrated, this leads to higher resource consumption and pollutant emissions in industrially developed cities, increasing environmental pressure and resulting in lower WLRES, thus forming LH agglomeration. Growing agricultural development in Northeast China, characterized by long-term large-scale farmland development and agricultural activities, has caused the overall regional WLRES level to decline. Some cities, with weak ecosystem protection awareness, have seen a more significant WLRES decline. This has increased the number of cities exhibiting (LH) agglomeration, thereby hindering the regional water and land resources integrated development.

The (LL) aggregation of WLRES in the Huang-Huai-Hai Basin results from severe agricultural water resource supply-demand conflicts. Agricultural over-exploitation causes water–land pollution and soil degradation; coupled with limited environmental investment, these factors collectively depress regional WLRES. Rising ecological awareness has prompted local governments to enhance water and soil management via policies, restoration projects, and innovation, thereby improving WLRES. Some cities, following notable WLRES improvements, have driven neighboring cities’ WLRES development, forming (HL) clusters.

The Yangtze River Basin shows lack of significant spatial autocorrelation of WLRES among cities. The absence of synergistic governance, coupled with stark disparities in economic development, land use, and natural resources among cities, underlies this phenomenon. Economically advanced cities have greater ecological investment and land resource management advantages, while the lack of effective regional collaborative mechanisms further exacerbates urban WLRES gaps [46].

4. Discussion

4.1. Influence of Scale Effects on WLRES Assessment

To reveal the spatiotemporal distribution and evolution of WLRES in China’s main grain-producing areas (MGPAs), we conducted evaluations at four nested scales: the entire MGPAs, sub-regions, provinces, and cities. Three main findings are presented.

First, WLRES outcomes differ across scales. Larger scales capture broad trends but obscure micro-level heterogeneity; multi-scale assessment is therefore essential for pinpointing risk hotspots. Between 2005 and 2020, for example, the Huang-Huai-Hai Basin exhibited an overall upward WLRES trajectory, yet Shandong Province within the basin experienced fluctuating declines. Similarly, Henan Province advanced from WLRES grade II in 2005 to grade III in 2020, while cities such as Shangqiu and Xinxiang, suffering severe groundwater over-extraction, remained stuck at grade I.

Second, multi-scale evaluation effectively unveils spatial heterogeneity. Before 2016, MGPAs displayed significant positive spatial autocorrelation. However, city-level analysis revealed that HH clusters were confined to the eastern north region, whereas LL clusters concentrated in the northern and central Huang-Huai-Hai Basin; other cities lacked significant spatial clustering. This indicates that, within sub-region scales, cities sharing similar resource endowments and ecological conditions tend to exhibit comparable WLRES levels [46]. Conversely, the majority of cities, constrained by administrative borders, economic gradients, and localized governance heterogeneity, fail to form effective linkages, allowing the statistical salience of a few HH/LL clusters to mask city-scale fragmentation.

Third, multi-scale analysis provides a framework for tracing how ecological risks propagate across scales and supplies scientific evidence for conservation policy. In 2005, northern cities exhibited high WLRES and formed HH clusters, whereas a minority of HL outliers arose from intensive agricultural and industrial resource consumption. Owing to the absence of cross-jurisdictional and cross-level coordination, these high–low (HL) anomalies failed to receive timely remediation. Pollutants migrated downstream via river networks, while economically advanced areas shift resource pressures onto surrounding regions through inter-regional allocation. The resulting negative externalities diffused with material flows, transforming localized HL risks into broader LH patches and ultimately eroding the region-wide positive spatial autocorrelation. Therefore, administrative boundaries must be transcended to establish a nested governance framework that integrates basin-level orchestration, provincial coordination, and city-level management. This framework will ensure the sustainable development of WLRES through refined control and cross-regional cooperation.

4.2. Limitations

This study evaluates the spatiotemporal characteristics of WLRES in Chinese MGPAs based on multiple scales, but there are still some limitations.

(1)

Scale resolution remains coarse. Using cities as the minimum unit provides a comprehensive overview of WLRES dynamics, yet it masks significant micro-scale heterogeneity at county or even township levels [43]. Moreover, missing values in city-level statistics necessitate interpolation, introducing additional uncertainty. Future work could integrate high-resolution remote-sensing products and nighttime-light data to construct a 1 km gridded dataset, thereby downscaling spatial resolution to improve accuracy and pinpoint micro-scale risk hotspots.

(2)

Scenario forecasting and policy early-warning are absent. Current findings only back-cast the period 2005–2020 and thus cannot offer prospective insights into WLRES evolution under the “dual-carbon” targets and evolving food-security contexts. Subsequent studies should couple Shared Socioeconomic Pathways (SSPs) and integrate CMIP6 climate projections, land-use simulations (FLUS), and socio-economic parameters to model WLRES trajectories for 2030 and 2050, identifying high-risk windows and quantifying the effectiveness of policy interventions [47].

(3)

Causal mechanisms between indicators and outcomes remain unclear. While the DPSIR–CRITIC framework reveals the relative contributions of indicators to WLRES, it cannot disentangle non-linear feedback or threshold effects within the “driving force–pressure–statues–impact–response” chain. Future work can apply SEM or Bayesian networks to quantify how key drivers, such as land reclamation rate and population density, affect WLRES, informing targeted policy.

4.3. Recommendations

The results of the study show that although the WLRES condition of MGPAs in China is improving, low level pressure and state subsystems hinder further enhancement. Moreover, factors like climate, resource endowment, and economic development vary across regions, causing uneven urban WLRES, weakening regional synergistic development effect, and leaving some cities with notably low WLRES. To address the above problems, this paper puts forward the following recommendations:

(1)

Reduce the negative impact of agricultural production on WLRES. Apply smart agriculture technology, with the help of Internet of Things and artificial intelligence technology and real-time monitoring of soil moisture, meteorological conditions, and crop growth. This method enables precision irrigation and targeted fertilization, enhancing water and nutrient-use efficiency. Consequently, soil salinization and nutrient loss are minimized. Promote green planting techniques, adopt biological control techniques to reduce the use of pesticides, apply organic fertilizers to improve soil structure and fertility, and reduce the pollution of soil and water by chemical fertilizers.

(2)

Tailor water-land protection and ecological-restoration policies to each region’s natural endowment and level of economic development. In terms of resource allocation, priority should be given to areas with scarce resources and ecological fragility. In the Huang-Huai-Hai Basin, with severe water scarcity, increase policy support and investment in boosting industrial and agricultural water use efficiency. Promote water-saving irrigation and high-standard farmland construction, optimize water resource allocation, and ensure agricultural sustainability. For the northeast region, over–utilization of arable land has led to soil fertility decline and soil erosion. To tackle this issue, conduct conservation agriculture technology innovation to promote technologies such as conservation tillage, straw mulching, and deep tillage. To address the low WLRES in some Yangtze River Basin cities due to underdeveloped economies, promote city cluster development, strengthen regional economic cooperation, share resources, and leverage complementary advantages to boost the regional economy.

(3)

Promote regional synergistic water and land resources protection. On the one hand, the government should strengthen the integrated management of natural resources, optimize the land use pattern, and strengthen the environmental regulation of high-pollution-emission cities to reduce the spillover of negative effects. On the other hand, a regional cooperation mechanism must be established to take advantage of the positive spatial spillover effect of high-level WLRES cities to elevate neighboring levels and coordinate sustainable water and land resource use.

5. Conclusions

Scientific evaluation of WLRES is crucial for ensuring food security. This study selected 180 cities in China’s MGPAs as research objects. Based on the DPSIR model, a WLRES assessment system with 25 indicators was constructed. The CRITIC method and Natural Breaks classification were applied to calculate and grade the WLRES index values. Kernel density estimation and Moran’s I were combined to analyze the dynamic trends and spatial distribution characteristics of WLRES across different scales. The main conclusions are as follows:

(1)

From 2005 to 2020, the aggregate WLRES index for China’s MGPAs rose from 0.470 to 0.518, reflecting an overall gain of 10.15%. This trajectory comprised three sequential phases—initial ascent, subsequent decline, and renewed rise—culminating in a shift from relatively low to moderate ecological security. The response subsystem contributed most, rising 38.74%, while the driver, pressure, and influence subsystems increased by 2.31, 2.20, and 8.13%, respectively. However, the state subsystem fell by 6.54%, indicating that while progress has been made in SDG-aligned actions (e.g., SDG 2.4 sustainable agriculture, SDG 6.4 water-use efficiency), land and water degradation (SDG 15.3) remains a challenge. This underscores the need to integrate WLRES monitoring into national SDG frameworks.

(2)

During the study period, the WLRES levels of the 180 cities were mainly concentrated in the three levels of relatively low ecological security, moderate ecological security, and relatively high ecological security. Cities with rising WLRES levels were mostly in the Huang-Huai-Hai Basin, and those with declines were predominantly in the northern region. Spatially, WLRES displayed marked heterogeneity, with higher ecological security in the northeast and southwest and weaker security in the central region.

(3)

The overall WLRES level in MGPAs increases, but the inter-city WLRES gap gradually widens. For the three subregions, there are obvious differences in the dynamic evolution characteristics of different regions, with the WLRES differences within the Yangtze River Basin narrowing and the WLRES differences within the northern region and the Huang-Huai-Hai Basin widening.

(4)

Spatial autocorrelation analysis showed that WLRES of MGPAs had significant global spatial positive correlation from 2005 to 2016, which gradually weakened and eventually disappeared. Local spatial autocorrelation was mainly “high–high” (HH) clustering, prominent in the northeastern part of the northern region, and “low–low” (LL) clustering, significant in the central and southern Huang-Huai-Hai Basin.