Temporal and Spatial Evolution of Grey Water Footprint in the Huai River Basin and Its Influencing Factors

Abstract

To evaluate water pollution status and sustainable development potential in the Huai River Basin, this study focused on the spatiotemporal evolution and influencing factors of the grey water footprint (GWF) across 35 cities in the basin from 2005 to 2020. This study quantifies the GWF from agricultural, industrial, and domestic perspectives and analyzes its spatial disparities by incorporating spatial autocorrelation analysis. The Tapio decoupling model was applied to explore the relationship between pollution and economic growth, and geographic detectors along with the STIRPAT model were utilized to identify driving factors. The results revealed no significant global spatial clustering of GWF in the basin, but a pattern of “high in the east and west, low in the north and south” emerged, with high-value areas concentrated in southern Henan and northern Jiangsu. By 2020, 85.7% of cities achieved strong decoupling, indicating improved coordination between the environment and economy. Key driving factors included primary industry output, crop sown area, and grey water footprint intensity, with a notable interaction between agricultural output and grey water footprint intensity. The quantitative analysis based on the STIRPAT model demonstrated that seven factors, including grey water footprint intensity and total crop sown area, exhibited significant contributions to influencing variations. Ranked by importance, these factors were grey water footprint intensity > total crop sown area > urbanization rate > population size > secondary industry output > primary industry output > industrial wastewater discharge, collectively explaining 90.2% of the variability in GWF. The study provides a robust scientific basis for water pollution control and differentiated management in the river basin and holds significant importance for promoting sustainable development of the basin.

1. Introduction

Water, as a fundamental natural resource essential for the survival and development of human society, encompasses two critical dimensions of availability: water quantity and water quality. While water quantity directly pertains to human survival and health, serving as a crucial guarantee for the normal operation of industrial and agricultural production, water quality further influences the effective utilization of water resources on the basis of quantity. The issue of water scarcity induced by water pollution-related quality degradation exacerbates the already severe situation of water resource shortages [1]. Consequently, effective governance of water pollution represents an indispensable pathway towards achieving sustainable development, with scientific and precise quantitative evaluation of water pollution serving as a pivotal prerequisite and foundational basis for addressing this challenge.

Currently, a diverse array of methodologies exists for evaluating regional water pollution, with commonly employed approaches including the single-factor index method, comprehensive pollution index method, and water footprint method, among others [2,3]. In 2002, Hoekstra et al. ingeniously integrated the concepts of ecological footprint and virtual water, thereby innovatively introducing the notion of “water footprint” [4,5]. Water footprint can be categorized into blue, green, and grey water footprints. Among them, the grey water footprint (GWF) theory focuses on the volume of freshwater required for pollutant dilution [6], enabling an objective assessment of water pollution status from the perspective of the quantitative relationship between water consumption and wastewater discharge. This provides a core indicator for quantifying agricultural and industrial pollution loads in this study. By quantifying the relationship between water consumption and wastewater discharge, the GWF offers an objective framework for assessing water pollution status, thereby providing a novel perspective and methodological approach for water pollution research.

In recent years, scholars at home and abroad have conducted extensive research on GWF, which has become an effective tool for assessing water pollution in specific regions. From the perspective of research objects, these studies can be primarily categorized into three levels: product-based, industry-based, and region-based. At the product level, certain studies have concentrated on specific products, conducting in-depth analyses of their GWF characteristics during the production process. Zhan et al. [7] compared the GWF of rice–crab co-culture vs. rice monoculture in Panjin, revealing circular agriculture’s potential in reducing the agricultural grey water footprint. Indirect processes in coal power contribute > 60% pollution load, necessitating supply chain-wide pollution control optimization [8].

At the industry level, systematic evaluations have been carried out on the impact of industrial activities on water resources across various sectors, including agriculture and industry. GWF calculations for industrial, agricultural, and domestic sectors are refined to better reflect water pollution by industries [9]. Agriculturally, research on the Huai River Basin analyzed grey water footprints of crop, livestock, and aquaculture sectors, assessing spatiotemporal efficiency and drivers like agro-economic impacts [2].

At the regional level, comprehensive assessments of GWF have been conducted for provinces, cities [10,11], and characteristic regions [12,13], providing a macroscopic perspective on regional water pollution conditions. Yin et al. [14] and Cheng et al. [15], utilizing Gansu Province and the national context as their respective samples, revealed an overall declining trend in GWF alongside pronounced regional disparities, noting that its variations are jointly driven by multifaceted factors including industrial structure, economic development, and technological advancements. Zhang et al. [16] evaluated grey water footprint intensity across Chinese provinces from 2000 to 2014, uncovering spatial heterogeneity and its underlying drivers; Wang and Yang [17] analyzed China’s blue and grey water footprints between 2002 and 2012, identifying agriculture as a primary contributor. Furthermore, Meng et al. [11] and Liu et al. [18], through case studies of Yantai City and the Yangtze River Economic Belt, respectively, elucidated that the spatiotemporal evolution of GWF is influenced by socioeconomic conditions and regional resource disparities, emphasizing the necessity for multi-scale research to underpin sustainable water management strategies. Research mostly focuses on macro-regions or specific industries, with relatively scarce studies at the watershed level.

Current research on GWF employs diverse methods, with existing approaches often emphasizing calculation and spatial distribution while insufficiently exploring the “pollution–economy–technology” interaction mechanisms during dynamic evolution. For model development, evaluation models based on GWF theory have been constructed, such as the nitrogen/phosphorus atmospheric deposition pollution model from numerical, index, and structural dimensions [19]; the improved copper GWF model incorporating biomagnification effects [20]; and the heavy metal GWF model based on hazard quotients [21]. In data processing and analysis, methodologies like life cycle assessment [22], ArcGIS spatial mapping, kernel density estimation, and standard deviational ellipse have been applied [15].

In summary, existing research still exhibits certain limitations. Existing analyses predominantly emphasize static spatial distributions, while insufficiently delving into the intricate interactive mechanisms among “pollution–economy–technology” during dynamic evolutionary processes. This shortfall renders it challenging to unveil the temporal effects of driving factors. Furthermore, research subjects are largely concentrated on macro-regions or specific industries, with a relative scarcity of studies focusing on specific river basins, particularly those that take into account varying levels of economic development within the basin. The Huai River Basin, as a vital economic zone and ecological corridor in China, exhibits pronounced disparities in economic development across its regions and is confronted with severe challenges of water pollution and water scarcity. Therefore, conducting in-depth research on the GWF in the Huai River Basin holds substantial theoretical and practical significance.

Based on this foundation, this study takes the Huai River Basin as the research object, employing the calculation method for GWF outlined in The Water Footprint Assessment Manual [23] and integrating previous research findings [24]. This study quantifies GWF across 35 Huai River Basin cities (2005–2020) across agricultural, industrial, and domestic sectors. It examines spatial differentiation via spatial autocorrelation analysis, investigates pollution–economic growth relationships using decoupling theory, and identifies key drivers of GWF variability through Geodetector and STIRPAT models. This study demonstrates dual academic–practical value by informing managerial decisions and environmental protection. For policymaking, spatiotemporal GWF patterns enable targeted watershed pollution control through regional priority identification, enhancing water management efficiency. For sustainability, quantitative pollution–economy analyses support integrated “economy–environment–society” governance frameworks, facilitating the Huai River Basin’s transition to sustainable “pollution mitigation–ecological recovery–green growth” paradigms.

2. Overview of the Study Area

Huai River is an important water system in the east of China. Its source is located in the northwest river valley of Tongbo Mountain, Tongbo County, Nanyang City, Henan Province. From west to east, the main stream flows through Henan, Anhui, and Jiangsu provinces, and enters the Yangtze River in Yangzhou, Jiangsu Province, with a total length of about 1000 km. The Huai River Basin is located between the Yangtze River basin and the Yellow River basin, which is a transitional zone between the north and south climates in China. The annual average precipitation in the Huai River Basin approximates 920 mm, exhibiting a distinct north-to-south decreasing trend. Precipitation is more abundant in mountainous regions compared to plains, and coastal areas receive higher precipitation than inland zones.

The Huai River Basin, a vital grain-producing area and industrial hub, faces severe water pollution pressures [2,25,26], making it a typical region for GWF research. In agriculture, intensive fertilizer application [27] causes significant nitrogen runoff, constituting the primary source of agricultural GWF. Industrially, its coal-reliant energy structure and energy-intensive industries [28] result in substantial industrial wastewater and pollutant discharges, directly impacting industrial GWF. Domestically, dense population and urbanization lead to massive sewage discharges, which, by 2015, were 2.54 times those of industrial wastewater [29], significantly contributing to household GWF. These characteristics collectively shape the GWF pattern in the Huai River Basin.

3. Research Methodology and Data Sources

Given the pollution profile and dominant pollutant sources in the Huai River Basin, the GWF is partitioned into agricultural, industrial, and domestic components for calculation [25,29].

3.1. Agricultural Grey Water Footprint

The agricultural grey water footprint is composed of two parts, crop farming and livestock breeding [30], and its calculation formula is as follows:

$$GW{F}_{agri}=GW{F}_{plant}+GW{F}_{live}$$

(1)

where GWF_agri represents agricultural grey water footprint (104 m³), and GWF_plant and GWF_live represent the grey water footprint of the planting industry and livestock and poultry industry (10⁴ m³) respectively.

3.1.1. Grey Water Footprint of Planting Industry

Grey water footprint of planting industry quantifies water pollution from unused fertilizers. While P³⁻ and K⁺ form insoluble soil compounds [31], N³⁻ leaches into water bodies [32], making nitrogen fertilizer the primary indicator for calculating planting-related grey water footprint. The calculation formula is as follows:

$$GW{F}_{plant}=\frac{a\times Appl}{C_{max}-C_{nat}}$$

(2)

where GWF_plant is the grey water footprint of planting industry (10⁴ m³); $\alpha$ represents the leaching rate of nitrogen fertilizer, with the national average value set at 7% [33]; Appl is nitrogen fertilizer application amount (t); C_max denotes the concentration standard for pollutants in water quality, with the standards adopted from the “Environmental Quality Standards for Surface Water” (GB 3838-2002) [34] for Class V water (COD: 40 mg/L, ammonia nitrogen: 2 mg/L, TN: 2 mg/L); and C_nat represents the natural initial concentration of nitrogen (N) in the water body (kg·m⁻³), and its value is set to 0 based on previous research studies [35].

3.1.2. Grey Water Footprint of Livestock and Poultry Farming

The main pollutants in the livestock and poultry industry are COD, TN, and TP, and feces and urine of pigs, cattle, sheep, and poultry are selected as pollution sources in the livestock and poultry industry. The number of animals with a rearing cycle of less than one year and those with a rearing cycle of one year or longer is measured by selecting the annual slaughter volume and the year-end inventory volume, respectively. The calculation formula is as follows [6]:

$$GW{F}_{live}=MAX\left[GW{F}_{live(COD)},GW{F}_{live(TN)},GW{F}_{live(TP)}\right]$$

(3)

$$GW{F}_{live(i)}=\frac{L_{live(i)}}{C_{\max }-C_{nat}}$$

(4)

$$L_{live(i)}={\displaystyle \sum _{j=1}^4}{N}_j{D}_j(1-R_j)(p_j{m}_{jp}{n}_{jp}+q_j{m}_{jq}{n}_{jq})$$

(5)

where GWF_live and GWF_live(i) are, respectively, the grey water footprint of livestock and poultry farming and the grey water footprint caused by the i pollutant in the breeding process, m³; L_live(i) is the pollution load of the i pollutant, kg; j is 4 kinds of livestock and poultry; N_j, D_j, p_j, and q_j are the feeding quantity, feeding cycle, and daily excrement and urine discharge of type j of poultry and livestock, respectively, kg/d; m_jp and m_jq are the pollutant content per unit of feces and urine of type j of poultry and livestock, respectively, kg/t; n_jp and n_jq are the pollutant loss coefficients per unit of feces and urine of type j of poultry and livestock, respectively; and R_j refers to the recycling rate of livestock excreta of type j of poultry and livestock. Detailed data are presented in

Table 1. Composition of livestock and poultry excreta.

Livestock and Poultry Category	Feces/(kg/d)	Urine (kg/d)	Raising Cycle/d	Excrement Recovery Rate
hog	2	3.3	199	0.3
cattle	20	10	365	0.2
sheep	2.6	-	365	0.2
fowl	0.125	-	210	0.5

and

Table 2. Content and loss rate of fecal pollutants.

Livestock and Poultry Category	Fecaluria	Contaminant Content (kg/t)			Pollution Flow Loss Coefficient
		COD	Ammonia Nitrogen	Total Nitrogen	COD	Ammonia Nitrogen	Total Nitrogen
hog	feces	52.00	3.08	5.88	0.0558	0.0304	0.0534
	urine	9.00	1.43	3.30	0.5000	0.5000	0.5000
cattle	feces	31.00	1.71	4.37	0.0616	0.0222	0.0568
	urine	6.00	3.47	8.00	0.5000	0.5000	0.5000
sheep	feces	4.63	0.80	7.50	0.0550	0.0410	0.0530
fowl	feces	45.65	2.79	10.42	0.0859	0.0415	0.0847

[3,36,37].

3.2. Industrial Grey Water Footprint

NH₃-N and COD are the primary pollutants in industrial wastewater. The calculation formula is as follows:

$$GW{F}_{ind}=MAX\left[GW{F}_{ind(COD)},GW{F}_{ind(N{H}_3-N)}\right]$$

(6)

$$GW{F}_{ind(i)}=\frac{L_{ind(i)}}{C_{max}-C_{nat}}$$

(7)

where GWF_ind is the industrial grey water footprint; GWF_ind(i) represents the grey water footprint of Class i pollutants; and L_ind(i) represents the emission of Class i pollutants.

3.3. Domestic Grey Water Footprint

The main pollutants of domestic sewage are COD and NH₃-N, and the calculation formula is as follows:

$$GW{F}_{dom}=MAX\left[GW{F}_{dom(COD)},GW{F}_{dom(N{H}_3-N)}\right]$$

(8)

$$GW{F}_{dom(i)}=\frac{L_{dom(i)}}{C_{max}-C_{nat}}$$

(9)

where GWF_dom is domestic grey water footprint; GWF_dom(i) represents the domestic grey water footprint of Class i pollutants; and L_dom(i) represents the emission of Class i domestic pollutants. For cities and prefectures with missing data, the pollutant generation amounts are calculated in accordance with the Methodology and Coefficient Manual for Accounting of Pollutant Generation and Discharge from Domestic Sources.

3.4. Regional Grey Water Footprint and Footprint Intensity

Regional grey water footprint quantifies the freshwater volume needed to dilute pollutants from regional production sectors [38]. Grey water footprint strength refers to the ratio of grey water footprint in a region to the gross domestic product of the region, which can be used as an indicator to measure the sewage treatment capacity [39]. The smaller the grey water footprint strength, the stronger the sewage treatment capacity of the region, and vice versa, the weaker the sewage treatment capacity [40]. Its calculation formula is as follows:

$$GWF=GW{F}_{agri}+GW{F}_{ind}+GW{F}_{dom}$$

(10)

$$T=GWF/GDP$$

(11)

where T is the intensity of grey water footprint; GWF is grey water footprint; and GDP is the total value of product.

3.5. Spatial Autocorrelation Analysis

Spatial autocorrelation assesses variable correlation across spaces and tests attribute similarity between adjacent spatial units, encompassing global and local analyses [41,42]. This study employs Moran’s I to quantify the spatial autocorrelation of GWF in the Huai River Basin, assuming homogeneous spatial patterns (i.e., constant autocorrelation intensity) across the region. A positive Moran’s I (>0) indicates spatial clustering of GWF, with higher values reflecting stronger aggregation; a negative value (<0) suggests spatial dispersion, with lower values denoting greater heterogeneity; a value of zero implies spatial randomness or no dependence [43].

$$I=\frac{\mathrm{n}}{{\displaystyle \sum _{\mathrm{i}}^{\mathrm{n}}}{\displaystyle \sum _{\mathrm{j}}^{\mathrm{n}}}{w}_{i,j}}·\frac{{\displaystyle \sum _{i=1}^n}{\displaystyle \sum _{j=1}^n}{w}_{i,j}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{{\displaystyle \sum _{\mathrm{i}}^{\mathrm{n}}}{\left(x_i-\overline{x}\right)}^2}$$

(12)

The parameter settings encompass the following: (1) the spatial weight matrix (w_i,j) is constructed using first-order queen contiguity with parallel standardization; (2) statistical significance is evaluated through Monte Carlo permutation tests. If the value exceeds 0, it can be proved that the internal space of the region is positively correlated; if the value is lower than 0, it can be proved that the internal space of the region is negatively correlated; if the value is exactly 0, it can be proved that the autocorrelation is negligible. n represents the number of cities within the Huai River Basin. w_i,j denotes the spatial weight between city i and city j; x_i and x_j represent the GWF values of city i and city j, respectively; and $\overline{x}$ signifies the mean value of the GWF.

The Global Moran’s I statistic is solely capable of demonstrating the presence or absence of clustering phenomena, yet it falls short in revealing the clustering characteristics of specific local areas. In contrast, hot-spot and cold-spot analysis can uncover the local instabilities that are obscured by the Global Moran’s I statistic. Furthermore, building upon the foundation of global spatial autocorrelation, this analysis further quantifies the regional correlation degree and local disparities in GWF, namely hot-spot and cold-spot regions. The specific formula is detailed in Equation (13) as follows:

$$G_i^{\ast }=\frac{{\displaystyle \sum _{j=1}^n}{w}_{ij}{x}_j-\overline{x}{\displaystyle \sum _{j=1}^n}{w}_{ij}}{S\sqrt{\frac{n{\displaystyle \sum _{j=1}^n}{w_{ij}}^2-{\displaystyle (\sum _{j=1}^n}w{i}_{ij}{)}^2}{n-1}}}$$

(13)

where w_i,j represents the spatial weight matrix, and S denotes the standard deviation.

3.6. Construction of Grey Water Footprint Decoupling Model

According to the Tapio decoupling elasticity method [44], a formula to measure the decoupling state between grey water footprint and economic development is constructed. The Tapio decoupling model implicitly posits that the ratio of the change rates between GWF and GDP remains constant over the study period, while overlooking nonlinear threshold effects. The Tapio decoupling model minimizes calculation bias from extreme initial/final value selection in the OECD index framework [45]. The calculation formula is as follows:

$$e=\frac{\Delta GWF/GW{F}_{t0}}{\Delta GDP/GD{P}_{t0}}$$

(14)

where e is the elasticity index of regional water pollution discharge decoupling, reflecting the changing relationship between regional grey water footprint and regional economy; ΔGWF and ΔGDP are the variation in regional grey water footprint and regional gross product, respectively; GWF_t0 is the regional grey water footprint of the base period; and GDP_t0 is the gross regional product of the base period. Based on ΔGWF, ΔGDP, and e, eight types of decoupling states are classified, as shown in

Table 3. Decoupling elasticity and decoupling state.

State		The Meaning of Status	ΔGWF	ΔGDP	Decoupling Elasticity
Connection	Expansion connection (EC)	Economic growth boosts environmental stress, with pollution rising as fast as or faster than the economy.	>0	>0	0.8 ≤ e ≤ 1.2
	Recessionary connection (RC)	Economic downturn reduces environmental stress, with pollution falling as fast as the shrinking economy.	<0	<0	0.8 ≤ e ≤ 1.2
Decoupling	Weak decoupling (WD)	Economic growth raises environmental stress, yet pollution rises much slower than the economy.	>0	>0	0 ≤ e < 0.8
	Strong decoupling (SD)	Economic growth coincides with declining environmental stress.	<0	>0	e < 0
	Recessionary decoupling (RD)	Economic decline reduces environmental stress, with pollution dropping far faster than the economy.	<0	<0	e > 1.2
Negative decoupling	Weak negative decoupling (WND)	Economic downturn reduces environmental stress, but pollution declines slightly slower than the economy.	<0	<0	0 ≤ e < 0.8
	Strong negative decoupling (SND)	Economic decline, yet environmental stress rises.	>0	<0	e < 0
	Expansionary negative decoupling (END)	Economic growth sparks a surge in environmental stress.	>0	>0	e > 1.2

. Utilizing administrative regions as the research units, this model may underestimate the impact of economic activities in neighbouring areas on the grey water footprint of the focal region.

3.7. Geographic Detector

Geographical detector analysis identifies spatial differentiation and its drivers [46]. This study applies its factor detector (quantifying explanatory power via q-statistic) and interaction detector (assessing combined factor effects) to analyze dominant influences on GWF in the Huai River Basin [47].

3.8. STIRPAT Model

There are many models for the analysis of influencing factors, and the main methods include the IPAT model, LMDI decomposition method, STIRPAT model, etc. [48]. The STIRPAT model is an extended model proposed by Dietz et al. based on the IPAT model [49]. The STIRPAT model posits a multiplicative relationship between environmental impact (I) and its driving forces, namely population (P), affluence (A), and technology (T), which can be transformed into a linear form through logarithmic conversion. This study utilizes annual data from 35 cities within the Huai River Basin spanning the period from 2005 to 2020, and the expression is as follows:

$$lnI=lna+b\left(lnP\right)+c\left(lnA\right)+d\left(lnT\right)+lne$$

(15)

where I represents environmental impact; P denotes population size; A signifies the level of economic affluence; T indicates the level of technological development; a is a constant term; b, c, d are the parameters of P, A and T; and e is the error terms.

An extended STIRPAT model for GWF is developed for the Huai River Basin based on regional conditions. Parameter statistical significance and model fit are evaluated via t/F-tests and R², with the formula presented as follows:

$$lnGW=lna+blnT+clnA1+dlnA2+elnA3+flnP+glnU+hlnG+tlnW+lne$$

(16)

where GW is the grey water footprint of the Huai River basin. T is the grey water footprint intensity, which reflects the relationship between regional economic development and water pollution. The lower the intensity, the less water pollution caused by economic development. A1, A2, and A3 are the primary industry production value, the secondary sector production value and the tertiary sector production value, respectively. The wastewater produced by different industrial sectors is different, and the change in industrial structure will affect the regional GWF. P is the population number, expressed by the permanent resident population at the end of the year of each district and county in the Huai River Basin. Population growth will increase the consumption of water resources and the discharge of domestic wastewater. U is the urban level, that is, the proportion of urban population in the total population of each district and county. The higher the urbanization level, the higher the discharge of urban domestic wastewater. G is the total sown area of crops. The higher the sown area, the greater the input of production materials and the higher the agricultural wastewater generated is. W is the discharge of industrial wastewater, representing the pollutants generated by industry in the basin. The higher the discharge of industrial wastewater, the higher the consumption of water resources and the generation of water pollution.

3.9. Data Sources

This research designates 35 prefecture-level cities in the Huai River Basin, which cover four provinces namely Henan, Anhui, Jiangsu, and Shandong, as the study area. It explores the GWF in the basin, with a particular focus on four intermediate time points between 2005 and 2020, specifically the years 2005, 2010, 2015, and 2020. The data concerning population, urbanization rate, economic indicators, livestock and poultry farming volumes, pure nitrogen fertilizer application amounts, and cultivated land areas are mainly collected from the statistical yearbooks of Henan, Anhui, Jiangsu, and Shandong provinces, as well as those of the relevant prefecture-level cities, such as the Jinan Statistical Yearbook. The emission volumes of industrial and domestic COD and ammonia nitrogen are obtained from the statistical yearbooks of respective cities or the official websites of their ecological and environmental bureaus (http://sthjj.taian.gov.cn/, accessed on 30 April 2024) [50].

The data of livestock and poultry feeding cycle, fecal excretion, pollutant content, and pollutant loss coefficient were taken from the Survey on Pollution of Large-scale Livestock and Poultry Industry in China and its Prevention and Control Countermeasures, and the recycling rate of livestock and poultry excrement was referenced from the research of Wang et al. [3,36,37].

4. Results

4.1. Spatial Autocorrelation Analysis of Grey Water Footprint

Through the global spatial autocorrelation analysis of grey water footprint in the Huai River Basin in 2005, 2010, 2015 and 2020, the global Moran index was calculated, as shown in

Table 4. Global Moran’s I statistics of GWF in the Huai River Basin at different times.

Time	Moran’s Index	Expected Index	z-Score	p-Value
2005	−0.002954	−0.029412	0.225038	0.821950
2010	0.006507	−0.029412	0.303598	0.761434
2015	0.090452	−0.029412	1.002787	0.315964
2020	0.019359	−0.029412	0.411006	0.681068

below.

It can be observed that the global Moran’s I values for the GWF in the Huai River Basin at different time points are all relatively close to 0, with p-values > 0.05. This indicates that, under a 5% significance level, the null hypothesis of “no spatial autocorrelation” cannot be rejected. Notably, in 2015, the p-value was 0.315964, the lowest among the four years, yet it still did not surpass the significance threshold. Therefore, the GWF did not exhibit a significant global spatial clustering or dispersion trend during the period from 2005 to 2020.

To further explore the local differentiation characteristics of GWF, the Getis-Ord Gi* statistic (Gi* value) was calculated using hot-spot and cold-spot analysis. Based on different confidence levels, the hot-spots and cold-spots were, respectively, classified into confidence intervals of 99%, 95%, and 90%, with their distributions shown in Figure 1.

By referring to the spatial distribution map of hot-spots and cold-spots of the GWF, it can be observed that in 2005 and 2010, low-value clusters of the GWF emerged, both located in Anhui Province, while no high-value clusters were present. In 2010, compared with 2005, the number of clustered areas decreased by two cities, indicating a gradually diminishing clustering effect. In 2015, high-value clusters appeared in three cities situated in the southeast of Henan Province and the west of Shandong Province. By 2020, these high-value clusters had shifted to Suqian City in Jiangsu Province. Meanwhile, low-value clusters persisted in both 2015 and 2020. It is evident that among the four years, 2015 exhibited the most significant spatial differences in the distribution of hot-spots and cold-spots of the GWF. However, on the whole, the proportion of cities with hot-spots and cold-spots in the Huai River Basin is relatively small, indicating only local spatial differentiation.

4.2. Spatial Differences of Grey Water Footprint

The GWF of the Huai River Basin in 2005, 2010, 2015, and 2020 was divided into five levels by using the natural break point classification method and the mean method, and the distribution map of the Huai River Basin in these four years was obtained.

As can be seen from Figure 2, agricultural grey water footprint is the predominant source of the grey water footprint in the Huai River Basin. The GWF distribution in the Huai River Basin from 2005 to 2020 presents a distribution pattern of high east–west and low north–south. In these four years, the GWFs of cities in Henan Province and Jiangsu Province is larger than those of cities in Shandong Province and Anhui Province, and the regions with high GWF values all appear in these two provinces. The distribution pattern of cities in Henan Province is high in the south and low in the north. The GWF of Nanyang City, Xinyang City, and Zhoukou City are notably higher than those of other cities in Henan Province across these four years. Specifically, the GWF of Zhoukou City exceeds 10 billion m³ in each of these four years, which is not only higher than that of other cities in Henan Province during the same periods but also generally surpasses that of cities in other provinces within the Huai River Basin. In Jiangsu Province, the GWF distribution peaks in Xuzhou City and Yancheng City, showing a bipolar distribution pattern. From 2005 to 2020, the grey water footprint of Yancheng City is 14.1 billion m³, 14.6 billion m³, 12.5 billion m³, and 12.6 billion m³, respectively, which is higher than that of other cities in Jiangsu Province during the same period. The spatial distribution of the GWF in Shandong Province and Anhui Province is relatively uniform, and the GWF is lower than that in Henan Province and Jiangsu Province and there is no particularly obvious high-value area.

To visually illustrate the changes in the weighted average position of the overall pollution load across the basin, this study has generated a dynamic shift diagram of the centroid (Figure 3). The centroid of domestic GWF in the Huai River Basin exhibits a trend of moving from west to east. It shifted from Shangqiu City, Henan Province (116.21° E, 33.85° N) in 2005 to Huaibei City, Anhui Province (116.73° E, 33.85° N), and further to Suzhou City, Anhui Province (116.82° E, 34.14° N) in 2020, representing an eastward shift of 0.61° in longitude and a northward shift of 0.29° in latitude. This shift is primarily attributed to the deceleration of population growth in eastern Henan Province, coupled with an increased weight of sewage discharge from cities at the junction of Jiangsu and Anhui provinces, which collectively pulled the overall centroid eastward.

The centroid of agricultural GWF remained relatively stable in Huaibei City, Anhui Province (116.62° E, 33.73° N) over the 15-year period, with minimal fluctuations. This stability is mainly due to the high proportion of cultivated land area and fertilizer application in the Huang-Huai-Hai Plain.

In contrast, the centroid of industrial GWF underwent the longest transfer distance. It was located in Shangqiu City, Henan Province (116.32° E, 33.95° N) in 2005, shifted to Huaibei City, Anhui Province (116.81° E, 33.80° N) in 2010, then to Zhoukou City, Henan Province (114.93° E, 33.65° N) in 2015, and finally stabilized in Zhoukou City (114.89° E, 33.44° N) in 2020, representing a westward shift of 1.43° in longitude and a southward shift of 0.51° in latitude. Among the three types of footprints, this migration is the most significant, caused by the westward shift in the industrial centroid. Therefore, it is essential to strengthen the construction of industrial wastewater treatment facilities in eastern Henan Province.

4.3. Grey Water Footprint Decoupling Analysis

In order to understand the relationship between the regional grey water footprint and regional economic growth in different periods of the Huai River Basin, based on Equation (14) and Table 1 the decoupling relationships of 35 cities and regions in the basin were calculated in three time stages from 2005 to 2010, from 2010 to 2015, and from 2015 to 2020. The calculation results are shown in Figure 4.

As can be seen from Figure 4, the decoupling relationship between the GWF and economic growth in the Huai River Basin exhibits the following phased characteristics: during 2005–2010, weak decoupling was dominant, with Huaibei City showing the best coordination (decoupling elasticity index of −0.53) and Xuchang City the worst (decoupling elasticity index of 0.20); from 2010 to 2015, many cities in Jiangsu Province shifted from weak decoupling to strong decoupling, achieving a positive coordination where economic growth was accompanied by a decline in GWF and improvement in water environment quality; between 2015 and 2020, 30 cities reached strong decoupling, with all cities in Anhui Province and Henan Province performing optimally, among which Huaibei City had the highest coordination with a decoupling elasticity index of −0.99. Specifically, Bengbu City experienced a continuous improvement path of “weak decoupling → strong decoupling → strong decoupling”, reflecting the superimposed effect of environmental protection policies; Tai’an City, however, showed a fluctuating degradation trend of “strong decoupling → weak decoupling → recessive decoupling”, exposing the risk of pollution transfer in industrial relocation. Overall, 85.7% of the cities in the basin achieved strong decoupling in 2020, with only Tai’an, Zibo, Zaozhuang, Yancheng, and Lianyungang failing to meet the standard, mainly affected by industrial structure, lagging policy implementation, and insufficient cross-regional coordination.

4.4. Spatial Geographic Detection of Influencing Factors of Grey Water Footprint

4.4.1. Factor Detector

The factor detector determined the influence degree of eight influencing factors on the regional grey water footprint (Figure 5). The q-values for the A1 from 2005 to 2020 are all > 0.58, with p-values < 0.005, indicating an extremely high level of significance. This suggests that the impact of agricultural output value on the target variable is dominant. The q-value for G increased from 0.515 in 2005 to 0.744 in 2015, and remained at 0.611 in 2020, demonstrating a gradually strengthening influence of agricultural planting scale expansion on the target variable. The q-value for T rose from 0.271 (not significant) in 2005 to 0.584 (p = 0.004) in 2020, indicating a rapidly increasing impact of industrial or domestic water pollution intensity on the target variable. The q-value for P fluctuated between 0.341 and 0.623, showing relatively high significance from 2005 to 2010 (p < 0.05), but weakened in later periods (p = 0.052 in 2020). The explanatory powers of W, A2, A3, and U are relatively weak.

4.4.2. Factor Interactive Detection

As depicted in Figure 6, the q-values for all interactive effects are significantly higher than those for single factors, with most exhibiting nonlinear enhancement, meaning that the q-values for the interactive effects between influencing factors are greater than the sum of the q-values for the individual factors. T demonstrates a generally high degree of correlation with A1 and A2 across various years. Specifically, the interaction between T and A1 remains exceptionally strong from 2005 to 2020, while the synergistic effect between T and A2 fluctuates after reaching a peak in 2010. W shows a relatively high degree of correlation with A1 in certain years, such as an interaction value of 0.84 in 2005 and 0.88 in 2010. The internal synergistic effect between A1 and G weakens after 2015, whereas the correlation between W and G is strongest from 2010 to 2015. Complex and dynamic interactive relationships exist among these factors, collectively influencing the regional water resource conditions.

4.5. Influencing Factors of Grey Water Footprint

4.5.1. Multiple Regression Analysis

P, U, W, A1, A2, A3, G, and T were set as the independent variable X, and were, respectively, used to represent lnP, lnU, lnW, lnA1, lnA2, lnA3, lnG, and lnT, and GW was set as dependent variable Y. It is represented by lnGW. SPSS 27 multiple regression analysis was used to perform regression fitting analysis on the data. The analysis results are shown in

Table 5. Statistical table of regression model.

Model	R	R2	After the Adjustment of R2	Model Error	Durbin–Watson
Numeric value	0.957	0.917	0.911	0.1826	1.554

and

Table 6. Multiple linear regression coefficients.

Variable	Unstandardized Coefficients	Standardization Coefficient	t	Significance	Collinearity Statistics
					Tolerance	VIF
Constant	−3.698		−4.521	0.000
A1	0.069	0.084	1.125	0.263	0.115	8.714
A2	0.203	0.322	3.848	0.000	0.091	10.989
A3	0.053	0.098	1.255	0.212	0.104	9.606
P	0.339	0.236	4.890	0.000	0.274	3.650
U	0.777	0.389	7.006	0.000	0.207	4.834
G	0.386	0.362	5.669	0.000	0.156	6.399
W	0.091	0.101	3.241	0.002	0.654	1.530
T	0.635	1.037	17.777	0.000	0.187	5.336

.

Through multiple regression analysis, it is found that the adjusted R² is 0.917, and the model has a good fitting effect. The Durbin–Watson coefficient is 1.554, close to 2, indicating that the sample independence is good.

The significance of A2, P, U, G, W, and T were all less than 0.05, which had significant effects on the dependent variables. The significance of other influencing factors was higher than 0.05, and the influence on the dependent variable was not significant. The VIF of A2 is greater than 10, indicating that there is a serious multicollinearity problem between this independent variable and other independent variables. In this study, the ridge regression method is employed to address the multicollinearity problem among the variables.

4.5.2. Ridge Regression Analysis

The ridge trace map (Figure 7) was obtained through ridge regression analysis, which showed the changes in standardized regression coefficients of each independent variable with K value. K value is a biassed parameter. The smaller the K value, the smaller the variance bias, and the better the simulation effect; otherwise, the greater the variance bias.

In conjunction with Figure 7, by employing a combined approach of the ridge trace plot method and the variance inflation factor method, it can be observed that each independent variable gradually stabilizes when K = 0.06. Subsequently, t-tests and F-tests were conducted on the model at this point. The test results are presented in

Table 7. K = 0.06 results of time ridge regression analysis.

K = 0.06	Non-Standardized Coefficients		Standardization Coefficient	t	P	VIF Value	R2	Adjust R2	F
	B	Standard Error	Beta
Constant	−1.461	0.668	-	−2.187	0.030 *	-	0.902	0.896	150.807 (0.000 **)
lnT	0.482	0.026	0.786	18.246	0.000 **	2.485
lnA1	0.139	0.034	0.170	4.063	0.000 **	2.349
lnA2	0.129	0.030	0.204	4.291	0.000 **	3.034
lnA3	0.008	0.026	0.014	0.304	0.761	3.037
lnP	0.357	0.053	0.248	6.773	0.000 **	1.797
lnU	0.563	0.084	0.282	6.708	0.000 **	2.356
lnG	0.352	0.045	0.330	7.837	0.000 **	2.369
lnW	0.115	0.026	0.128	4.479	0.000 **	1.089
	dependent variable: lnGW

**, * represent the significance levels of 1%, 5%, respectively.

.

From the table above, it is evident that the model has successfully passed the F-test (F = 150.807, p = 0.000 < 0.05), suggesting the presence of a regression relationship among lnT, lnA1, lnA2, lnA3, lnP, lnU, lnG, and lnW. Meanwhile, the model’s goodness-of-fit, denoted by R², is 0.902, indicating that the variables lnT, lnA1, lnA2, lnA3, lnP, lnU, lnG, and lnW can account for 90.2% of the variations in lnGW. This demonstrates that the model exhibits relatively excellent performance.

Parameters of each independent variable can be determined from Table 7. The parameters of the ridge regression model equation corresponding to independent variables lnT, lnA1, lnA2, lnA3, lnP, lnU, lnG, and lnW are as follows: 0.482, 0.139, 0.129, 0.008, 0.357, 0.563, 0.352, 0.115. As such, the ridge regression equation is as follows:

$$lnGW=-1.461+0.482lnT+0.139lnA1+0.129lnA2+0.008lnA3+0.357lnP+0.563lnU+0.352lnG+0.115lnW$$

(17)

The analysis results reveal that, except for an insignificant positive relationship with A3, the GWF in the Huai River Basin exhibits significant positive correlations with other driving factors. After eliminating the influence of dimensionality using the standardized coefficient Beta, the relative contributions of each factor to the GWF can be visually discerned. Ranked in descending order of importance, these factors are T > G > U > P > A2 > A1 > W. With the sustained growth of these key driving factors, water resource stress in the basin will concurrently escalate, posing new challenges to the construction of ecological civilization in the Huai River Basin.

5. Discussion

This study integrates a holistic analytical framework of “spatial differentiation–economic interconnection–driving mechanism” at the watershed scale, systematically uncovering the spatiotemporal evolution patterns of GWF in the Huai River Basin from 2005 to 2020. The findings reveal a spatial pattern of GWF characterized by “high values in the east and west, low values in the north and south”, with stable high-value zones in southern Henan and northern Jiangsu, and low-value areas in most cities of Anhui. This pattern closely aligns with the spatial disparities in agricultural scale and industrial structure as the high-value regions exhibit large-scale agricultural cultivation and a concentration of energy-intensive industries in northern Jiangsu, resulting in higher pollution loads; conversely, cities in Anhui demonstrate lower agricultural intensification and relatively lighter industrial structures, imposing smaller pollution pressures. Factor interaction detection indicates that the interactions between A1 and T, as well as between A2 and T, constitute the core mechanisms driving the spatial differentiation of GWF. This corroborates the theory of synergistic pollution driven by “economic structure–technological level”: the combination of agricultural intensification and low sewage treatment technologies exacerbates non-point source pollution, while the coupling of industrial expansion and technological lag amplifies point source pollution risks. The STIRPAT model further quantifies the contributions of seven key factors, among which T and G exert the most significant impacts, suggesting that technological upgrading and agricultural structural optimization are pivotal for pollution control. Decoupling analysis reveals a transition from “weak decoupling” to “strong decoupling” around 2015 in the basin, with the proportion of cities achieving strong decoupling rising from 31.4% to 85.7%, and the circular economy model in Huaibei City demonstrating notable effectiveness. This reflects that pollution control policies have begun to take effect amid economic growth in the basin. Nevertheless, the five cities that have not achieved strong decoupling, including Tai’an, serve as a warning that during industrial relocation if pollution control fails to advance synchronously then it may lead to pollution “relocation” rather than “reduction”.

This study reveals agricultural grey water footprint as the dominant contributor in the Huai River Basin, aligning with findings from the northern Tianshan Mountains [51] and Yinchuan [52], underscoring the non-negligible regional agricultural non-point source pollution. The GWF in the Huai River Basin exhibits a spatial pattern of “high in the east and west, low in the north and south”. This finding is relatively consistent with the research conclusions of Chen et al. [2] regarding the agricultural grey water footprint efficiency in the Huai River Basin for the years 2010 and 2015, and aligns with the trend of a gradual decrease in the grey water footprints of wheat, rice, corn, soybeans, and oilseeds from 2005 to 2015 as reported by Feng et al. [26]. The primary reason is attributed to the high proportion of the grey water footprint in crop production caused by the application of nitrogen fertilizers. This can be compared with the Yellow River Basin [53], where over the decade from 2012 to 2021 the GWF in the nine provinces along the Yellow River exhibited an overall fluctuating downward trend, which is relatively consistent with the trend observed in the Huai River Basin. This reflects that both water resource conservation and purification have been given significant attention across various provinces within different river basins.

However, the spatial autocorrelation of the GWF in the Huai River Basin is weaker [54]. The global Moran’s I values range from −0.002 to 0.09 with P > 0.05, suggesting weak spatial association. In contrast, the GWF in the Yellow River Basin exhibited significant spatial autocorrelation from 2000 to 2019, with global Moran’s I values ranging from 0.494 to 0.449. The disparities can be primarily attributed to two factors: first, there are significant differences in pollution control policies among the four administrative units spanning the Huai River Basin [55,56]; second, the Yellow River has a strong longitudinal gradient [57], whereas the river system of the Huai River is fan-shaped and dispersed, leading to fragmented pollution diffusion pathways and weakening spatial agglomeration [58]. In terms of economic–environmental coordination, the proportion of cities achieving strong decoupling in this study rose from 31.4% to 85.7%, aligning with the research trend observed by Sun et al. [6] in the Yangtze River Economic Belt. The watershed successfully decoupled economic growth from environmental stress, confirming the co-realizability of economic growth and pollution reduction. From 2014 to 2019, the entire Yangtze River Economic Belt consistently maintained a state of strong decoupling, demonstrating a more thorough decoupling compared to the Huai River Basin. In terms of driving mechanisms, T (standardized coefficient of 0.786) emerges as the primary driving factor, highlighting the pivotal role of technological advancement in pollution control. This finding aligns with the research conclusions of Zhang et al. in their study on water pollution in the Yangtze River Economic Belt [59]. However, the agricultural contribution in the Huai River Basin is more prominent (with a q-value of 0.743 for A1), reflecting its uniqueness as a major grain-producing region. Geographic detectors reveal strong interactions between the outputs of primary and secondary industries and T (q = 0.84–0.93), indicating that the essence of decoupling lies in the synergistic evolution of economic scale expansion and technological emission reduction effects, rather than being driven by a single factor [16]. The evolution of environmental stress relies on the dynamic coupling process between economic and technological systems, providing a theoretical anchor for cross-factor synergistic governance and deepening the cognitive framework for complex regional environment–development relationships.

The Tapio elasticity approach, rather than the OECD decoupling index, is employed in the decoupling model because the former can mitigate biases arising from extreme initial and final values [60], making it more suitable for the characteristics of significant economic fluctuations in the Huai River Basin. Methods for addressing multicollinearity encompass principal component analysis (PCA), partial least squares regression (PLSR), and ridge regression, among others. Compared with principal component analysis (PCA) and partial least squares regression (PLSR), ridge regression directly conducts regularization on the original variables, thereby avoiding issues of variable information loss and interpretational difficulties [61,62,63]. Given its frequent application in research concerning energy and pollutants, this method is selected for the present study. The integration of the Geodetector and the STIRPAT model overcomes the limitations of employing a single method as the Geodetector excels at identifying dominant factors driving spatial differentiation [64], while the STIRPAT model can quantify the marginal effects of driving factors [65]. Their combination enables a comprehensive analysis of the entire chain from “spatial pattern to driving mechanism”.

This study validates the applicability of the combined “Geodetector–STIRPAT” approach at the watershed scale, offering a methodological reference for similar research endeavours and providing quantitative evidence for zonal governance strategies. Nevertheless, the study still has certain limitations: Firstly, it fails to account for the influence of hydrological connectivity between the main stem and tributaries of the Huai River on pollution diffusion, potentially underestimating the impact of upstream areas on downstream regions. Secondly, climate change factors (e.g., annual precipitation) are not incorporated, despite precipitation’s potential to influence agricultural pollution loads through leaching processes [66,67]. Future efforts should integrate the complexity of natural–social systems to further deepen multi-factor coupling analysis, thereby supporting sustainable development decision-making in river basins.

6. Conclusions

This study employs a multidimensional analytical framework to examine the spatiotemporal evolution patterns and driving mechanisms of GWF across 35 prefecture-level cities in the Huai River Basin from 2005 to 2020. By integrating spatial autocorrelation analysis, Tapio decoupling model, geographical detector, and STIRPAT model, it quantifies GWF dynamics, explores the relationship between GWF and economic growth, and identifies key influencing factors. Major conclusions are as follows:

(1)

The GWF in the Huai River Basin shows no significant global spatial agglomeration yet presents a distinct pattern of “higher in the east and west, lower in the north and south”. High-value areas are stably distributed in southern Henan, including Nanyang, Xinyang, and Zhoukou, and northern Jiangsu, including Xuzhou and Yancheng, whereas most cities in Anhui Province remain at low GWF levels. This spatial pattern necessitates regionally differentiated governance strategies. For high-value areas in Henan and Jiangsu, stricter measures for agricultural non-point source pollution control are required, such as reducing chemical fertilizer and pesticide application and promoting ecological farming. For low-value areas in Anhui and Shandong, cross-regional ecological compensation mechanisms should be explored to encourage upstream water source protection.

(2)

The Huai River Basin achieved a critical transition from “weak decoupling” to “strong decoupling” around 2015. The proportion of cities in strong decoupling increased from 31.4% in 2005–2010 to 85.7% in 2015–2020. Huaibei City exhibited an extremely low decoupling elasticity of −0.99, verifying the effectiveness of its circular economy model. However, five cities failed to achieve strong decoupling. Continuous monitoring and early warning systems are essential for cities at risk of recessive or weak negative decoupling, such as Tai’an and Zaozhuang, to promote cleaner industrial transformation.

(3)

Geographical detector analysis demonstrates that the interaction between economic factors, A1 and A2, and technological factors, namely T, serves as the dominant driver of grey water footprint variations in the Huai River Basin. Their interaction q-values range from 0.84 to 0.93, significantly exceeding the explanatory power of single factors. The STIRPAT model further quantifies seven significant influencing factors, ranked by importance as T, G, U, P, A2, A1, and W. These factors collectively explain 90.2% of GWF variations in the basin. These results imply that mitigating grey water footprint in the Huai River Basin requires synergistic regulation of economic structure and technological progress, with priority given to factors ranked by importance, such as reducing T and optimizing G, to achieve targeted control.

This study provides a solid scientific basis for targeted water pollution governance in the Huai River Basin and offers methodological references for similar watershed-scale research on GWF dynamics. Future research may further refine driving mechanism analysis by integrating hydrological and climate factors.