Spatial disparities and water resource green efficiency improvement potential in the four-city area in middle China

Urban agglomerations are the inevitable outcome of China's emerging industrialization and urbanization. Such urban environments have become not only dynamic and promising sites for current and future economic development but also highly sensitive, considering the associated concentrated and intensified environmental challenges. Urban agglomerations in China account for 75% of the national economic output, as well as 67% of the industrial wastewater, gas emissions, and solid waste generated in the country (Fang 2015). In addition, under future climatic and hydrological uncertainty, water-related challenges caused by continuous expansion of urban centers and economic growth pose emerging challenges to sustainable management of water resources in urban agglomerations, and are key sources of concern for researchers and governments.

Today, green development has become a key trend globally, and its aim is to minimize resource consumption and environmental pollution, while facilitating sustainable development across economic, social, and ecological spheres (Gilbert 2016). The water resource green efficiency (WRGE) concept was put forward based on a combination of China's and global water resource and sustainable development requirements. It refers to the ratio of water resource input to economic, social, and ecological output, and is used to evaluate the social benefits of water resource services (Sun et al. 2017; Guo et al. 2020), with the aim of facilitating the improvement of human well-being and increasing happiness in populations, which are also fundamental aspects of the green development concept (Zhang et al. 2014). Consequently, based on the WRGE concept, social benefit output must be considered as a desired output and integrated in evaluation indexes for WRGE.

Recently, the WRGE, which considers diverse social and development indicators, has attracted the attention of scholars globally. For example, Sun et al. (2018a, 2018b) evaluated WRGE in China using the slack-based measure-data envelopment analysis (SBM-DEA) model, with social development index applied as the desirable output. Guo et al. (2020) employed the social benefit index as the desired output and assessed WRGE in 18 cities in Henan Province, China, using the SBM-DEA model. In addition, Zhang et al. (2020) evaluated WRGE in China using the HDDF-GML model, with social development index as the desired output, whereas Huang et al. (2021) evaluated WRGE in the Yangtze River Economic Belt, with SBM-DEA, Malmquist index, and social network analysis models as undesirable outputs, and China human development index as desirable output. Furthermore, Wu et al. (2021) evaluated WRGE in Northwest China using the super-SBM model, with built-up greenery area, green coverage rate, and park greenbelt area as desirable outputs. With the human social development index (HDI) proposed by Tian (2008) as a basis, Yang & Xie (2019) added the environmental index to develop a human sustainable development index (HSDI) and evaluated WRGE in the Yangtze River Economic Belt using the SE-SBM model. Some recent studies have employed the SBM-DEA model to calculate the green efficiency of industrial water resource use in China (Xiao et al. 2020; Zhang et al. 2021), Zhejiang Province (Zhang et al. 2019), the Poyang Lake region (Lv et al. 2021), and the Huaihe river Basin (Tian et al. 2021).

Overall, existing studies have largely focused on specific provinces in China and the Economic Belt and Basin, with diverse social development indicators adopted in the evaluations. Notably, few studies on water resource use in China under urban agglomeration have considered the regional heterogeneity, considering the great disparities in water resource endowment, economic foundation, urbanization level, and industrial structure across different urban agglomerations. Furthermore, different regions across urban agglomerations may have distinct water-use trends (Yue et al. 2016), and WRGE may vary significantly across various administrative regions within urban agglomerations. Consequently, different water conservation and efficiency improvement strategies should be adopted in different urban agglomeration regions.

The present study selects the ‘four-city area in middle China’ urban agglomeration as the research subject and makes the following improvements based on current research. First, the groups were characterized and the group members were determined based on water-use matching indexes, which were employed for categorization. Secondly, water resource ecological footprint (WREF), water pollution ecological footprint (WPEF), and HSDI were incorporated into the WRGE evaluation index system. Third, WRGE was evaluated using meta-frontier undesirable super-efficiency slack-based measure (Meta-US-SBM) incorporating regional heterogeneity. Finally, WREF saving potential and gross domestic product (GDP) growth potential were used in combination to analyze the improvement potential of WRGE. The specific objectives of the present study were as follows: (i) to uncover differences in WRGE across regions in urban agglomerations; (ii) to identify the improvement room of WRGE across regions and groups; and (iii) to explore novel approaches for improving WRGE potential.

Considering data availability, the scope of the study included the administrative regions of each city. Indicator datasets were derived from the Statistical Yearbooks (2011–2020) of Hubei Province, Hunan Province, Jiangxi Province, and Anhui Province, as well as the Water Resources Bulletin (2011–2020) of each province, Statistical Bulletin on National Economic and Social Development, and the Statistical Bulletin of the Environmental Situation of 38 regions in the four-city area in middle China from 2011 to 2020. A small amount of missing data was supplemented according to the environmental protection planning of provinces and prefecture-level cities. In the present study, the required indicators are divided into two categories: input and output indicators. Capital investment, labor force, and WREF are input indicators; GDP and HSDI are desirable outputs, whereas WPEF is an undesirable output. The specific indicators are defined as follows:

• (1)

  Capital investment input: The fixed asset investments of the entire society over the years (Chen & Jia 2017; Zhou et al. 2018; Yu et al. 2019b). The total fixed asset investments over the years were converted into 2000 prices (constant prices) based on the fixed asset investment price index.

• (2)

  Labor force input: The number of employed persons at each year's end.

• (3)

  WREF input: WREF, i.e., water volume ecological footprint, refers to water resource consumption by human beings for production and life (Wang & Liu 2019). Total WREF in the region was selected as the water resource input, excluding water quality ecological footprint. WREF in a region was calculated using the method of Li et al. (2020).

• (4)

  Desirable output: The desirable outputs include real GDP and HSDI for each province and city.

GDP: The GDP of each region over the years is converted to 2,000 prices (constant prices) using the producer's price index.

HSDI: HSDI integrates the environmental dimension (such as per capita industrial wastewater discharge, per capita industrial waste gas discharge, per capita industrial solid waste emissions, and per capita electricity consumption) in the HDI, which already covers two out of three dimensions of sustainability, namely, social and economic dimensions (Goodland 1995), to develop a comprehensive HSDI (Togtokh 2011). HSDI is then calculated and transformed to obtain the environmental performance index, which is integrated into the income index, and modified to obtain an HSDI (Tian et al. 2007) that contains environmental impact factors, which is taken as expected output. The maximum and minimum ranges of each of the indicators are defined. According to data collection and trends over the years, in the present study, 200 ton/person, 400,000 standard cubic meters/person, 0.3 ton/person, and 15,000 kWH/person are selected as the maximum values of per capita wastewater discharge, per capita waste gas emissions, per capita industrial solid waste emissions, and per capita electricity consumption, respectively, and their minimum values are 0. More details on the formula for the calculation of HSDI can be found in Bravo (2014, 2015), Wang (2016), and Hickel (2020).

• (5)

  Undesirable outputs: WPEF, i.e., water quality ecological footprint, refers to the area of water resources required to absorb pollutants produced by a certain population that exceeds the carrying capacity of the existing water body and to dilute the pollutants to a certain water quality standard (Wang & Liu 2019). The present study focused on the four-city area in middle China and selected total emission reduction indicators for major pollutants – chemical requirements (e.g., Chemical Oxygen Demand (COD)) and ammonia nitrogen (e.g., NH₃-N) emissions – as research content and undesirable outputs. Considering the two pollutants have obvious overlap in environmental impacts, the greatest pollution footprint between them is considered the final WPEF. WPEF in the region is calculated using the method of Feng & Zhao (2020).

In Equations (1) and (2), represents the weight vector in the production technology frontier of the nth DMU reference in category h. ⁠, and denote slack variables of input, desirable output, and undesirable output, respectively. ε represents the non-Archimedes infinitesimal. The objective function aims to enhance desirable outputs and minimize undesirable outputs and inputs, simultaneously. Moreover, they can reflect the improvement potential of various regional efficiencies (Wang et al. 2015a, 2015b). Conversely, the goal of the constraint is to ensure the denominator of the objective function is >0.

The closer the TGR is to 1, the lower the heterogeneity in technology, which also implies that the GWRGE is closer to the MWRGE. In contrast, the smaller the TGR, the higher the heterogeneity in technology; therefore, the MWRGE would be considerably lower than the GWRGE.

Table 1 presents the descriptive statistics of the input and output indicator data for each region in 2010–2019. Due to space limitations, only average values are listed. (1) From 2010 to 2019, the average WREF was 365.22 thousand m³/hm². Among them, YC was the region with the highest WREF, at 699.87 thousand m³/hm², while the region with the lowest WREF was QJ, at 113.02 thousand m³/hm². (2) Capital investment increased from 2010 to 2019, with an average increase of 112.11 billion yuan per year. Wuhan had the highest capital investment, at 437.61 billion yuan, with TM having the lowest, at 20.96 billion yuan. (3) Labor force increased in 2010–2019 by 244.04 thousand persons per region on overage. The region with the highest labor force was Wuhan, at 543.33 thousand persons, while QJ had the lowest labor force, at 59.53 thousand persons. (4) Average GDP increased in 2010–2019, with an average increase of 124.14 billion yuan per year. Wuhan had the highest GDP, at 635.75 billion yuan, and TM had the lowest GDP, at 25.39 billion yuan. (5) Average HSDI increased in 2010–2019, with an average increase of 0.7691 per year. Changsha had the highest HSDI at 0.8160, with SR having the lowest HSDI, at 0.6850. (6) The average WPEF was 223.92 thousand m³/hm² in 2010–2019. YY had the highest WPEF, at 800.56 thousand m³/hm², whereas CiZ had the lowest, at 52.16 thousand m³/hm².

Table 2 presents the summary statistics of the input and output indicators for different categories. According to the coefficient of variation (CV), the dispersion degree of each variable was less than 0.65, indicating moderate heterogeneity among variables in different categories. Conversely, the GDP in EOU and capital investment in DWEU showed fluctuation, at 2.401 and 1.165, respectively, indicating strong heterogeneity in GDP in EOU and capital investment in DWEU. With regard to inputs, WREF and labor force inputs were higher than capital inputs across the four categories. In particular, capital inputs were the highest in DSEY, followed by UOE. With regard to outputs, HSDI output growth rates across the four categories were roughly similar. In addition, the highest GDP growth rate was observed in DWEU, and WPEF from undesirable outputs in EOU far exceeded those in the other categories, with the DWEU having the lowest undesirable outputs. Overall, the four categories exhibited heterogeneous growth patterns, implying that certain regional differences in WRGE exist in the four-city area in middle China.

The average GWRGE varied from roughly 0.5201 in HN to 1.3317 in HS with an overall average value of 0.9112 (Table 3). A representative GWRGE in HN had 47.99% room for improvement. In contrast, the GWRGE in HS decreased by 33.17%, saving its production inputs by improving its technical management ability. The average MWRGE in the 38 regions ranged between 0.3104 and 1.1292, with an overall mean value of 0.6331. MWRGE in QJ (1.1292), XY (1.0771), and TM (1.0066) were the highest, whereas MWRGE in SR (0.3104), HG (0.3148), and JA (0.3233) were the lowest. The results indicate that the MWRGEs of the 38 regions in the four-city area in middle China require further improvement. Our results are consistent with those of WRGE studies conducted by Zhang et al. (2020) in different regions in China. Moreover, the average GWRGE was generally greater than the average MWRGE (Table 3). For example, in NC, the GWRGE was relatively high (1.1529); however, the MWRGE was only 0.554. Similarly, the GWRGE of HS was >1.2, whereas that of MWRGE was 0.802. This result is similar to the findings of Wang et al. (2018) on carbon reduction efficiency and Yu et al. (2019a, 2019b) on energy efficiency. Among the 38 regions, GWRGE and MWRGE were only equal in TM, and both were greater than 1, indicating that the production frontier had been reached.

When Technology Gap Ration (TGR) approaches 1, the group-frontier is closer to the optimal production technology in the meta-frontier. Conversely, when TGR approaches 0, the group-frontier is far below the optimal production technology in the meta-frontier. From Table 3, it is obvious that the technology gaps of WRGE among different regions in the four-city area in Middle China were large. The average TGR of the 38 regions ranged from 0.171 to 1.000, with an overall average value of 0.694. Among them, the average TGR value in TM was 1, which means that the group-frontier achieved the optimal production technology in the meta-frontier. The average TGR for QJ was 0.993, ranging from 0.959 to 1.000, indicating a low technology gap between group-frontier and meta-frontier. The average TGRs in XT, WH, YT, XY, HN, and CS were 0.964, 0.925, 0.919, 0.905, 0.903, and 0.898, respectively, indicating that the six regions had achieved 96.6, 92.48, 91.86, 90.48, 90.26 and 89.79% of the national potential optimal production technology, respectively, in green water resource utilization. However, the average TGRs for AQ, Whu, NC, LA, MAS, FZ, and YiY, were 0.4884, 0.4871, 0.4753, 0.4696, 0.4695, 0.4377, and 0.4132, respectively, which indicated that there was 51.16, 51.30, 52.47, 53.04, 53.05, 56.23, and 58.68% room for improvement in the production technology levels in the seven regions above, respectively, when compared with the national potential optimal technology level.

The average TGR in DWEU was significantly higher than those in the other categories (Table 4), implying that the DWEU has the optimal green water resource utilization technologies, followed by DSEU and EOU. The average TGR in UOE was 0.5325, implying there was 46.75% room for improvement compared to the national potential optimal production technology level. The average TGRs in DSEU, EOU, and DWEU were 0.7591, 0.7376, and 0.8396, respectively, which implied that compared to the national potential optimal production technology, DSEU, EOU, and DWEU had 24.09, 26.24, and 16.04% room for improvement, respectively. However, TGR exhibited an overall downward trend in DSEU, DWEU, and UOE, indicating that WRGE gaps among DSEU, DWEU and UOE were widening. Conversely, TGR exhibited an overall upward trend in EOU, implying that the WRGE gaps between EOU and DSEU, and between DWEU and UOE, were narrowing.

To verify the GWRUE technology gap across the four categories, the Kruskal–Wallis test was used to examine whether TGR varies among categories, and the descriptive statistics and test results are listed in Table 5. TGR ranged from 0.1714 to 1.000. In addition, the results presented in Table 5 strongly reject the null hypothesis that TGR in different categories comes from the same group. DWEU generally had the highest TGR. Specifically, the average TGR in DWEU was 0.846, implying the group-frontier in DWEU was quite close to the meta-frontier, while the TGR value of UOE was 0.5409. In other words, given the same capital and labor force inputs, UOE required 1.5 times more water resources than DWEU to yield the same output. This also indicates that the UOE in the ‘four-city area in middle China’ has greater potential for WRGE improvement.

WE^IPR was highly negatively correlated with WRGE value in group-frontier and meta-frontier (Figure 5). For example, YC ranks 33rd based on the mean GWRGE; however, it ranks first based on WE^IPR; HS ranks first based on mean GWRGE; however, it ranks 19th based on WE^IPR (Figure 5(a)). In addition, JA ranks 36th based on mean MWRGE, while it ranks first based on WE^IPR; QJ ranks first based on MWRGE, and ranks 35th based on WE^IPR (Figure 5(b)). This result is similar to the findings of Zhang et al. (2015b) who found that the potential improvement rate of carbon productivity is highly negatively correlated with the super-efficiency value. However, the negative correlation is not strictly established. For example, the mean WE^IPR in the group-frontier in YC was the highest; however, the mean GWRGE in YC was not the lowest, and further examination revealed that the WE^IPR in YC was obviously larger (Figure 5(a)). Such a phenomenon was also observed in JA in the meta-frontier (Figure 5(b)). The potential reason is the disruption of capital investment and labor force slacks. Compared to other regions where capital investment and labor force slacks can be reduced, YC and JA had less capital investment and labor force input, so that in order to achieve the optimal ratio among factors, the pressures of their WREF reduction would be greater.

In JA, the WE^IPR was 0.6208 under group-frontier and 1.2196 under meta-frontier, and the W^IPR accounted for more than 80% of WE^IPR, which implies that the higher WE^IPR in JA is not due to the low WREF input, but the high output, while the high output may be attributed to other input factors and the influence of technology and knowledge. In addition, 16 regions, including PX, XG, HG, JJ, YC, FZ, SR, AQ, CZ, LA, XN, XT, LD, HN, CiZ, and XC, under group-frontier, and 31 regions under meta-frontier, also had large proportions of W^IPR. The results indicate that to reduce the WE^IPR and improve the actual WRGE level, the regions should not only pay attention to the exploitation of the production potential of existing factors but also the optimization and adjustment of the ratio relationships among factors. In contrast, 18 regions under group-frontier, including WH, CS, HF, etc., and WH, HF, XY, XT, QJ, TL, and CiZ under meta-frontier, showed relatively low economic output and economic development remains a top priority of the regions. This may be because the effect of improved water efficiency did not offset the effect of increased water use resulting from the increase in total output (Liu & Zhang 2015). However, NC, EZ, and YiY had no room for improvement for WRGE (Figure 5(a)) under group-frontier. The results above also reflect inconsistency in factor allocation and production capacity in different regions. Therefore, when formulating economic development strategies and water conservation policies, each region should aim to achieve sustainable water resource utilization by the dual water supply based on their development characteristics and needs.

In a trend distinct from that of WRGE, the annual average WE^IPR in UOE was 0.242, which was significantly higher than those in DSEU, EOU, and DWEU (0.108, 0.176, and 0.160, respectively). Among them, the overall WE^IPR in DSEU exhibited an upward-downward trend, with the highest value in 2016 and the lowest value in 2019. WE^IPR in EOU exhibited an upward-downward-rising trend, with the highest and lowest values in 2019 and 2014, respectively. The WE^IPR in UOE generally exhibited a downward trend, while that of DWEU exhibited a downward-upward trend, with the highest and lowest values in 2019 and 2015, respectively. The results indicated that the improvement potentials of WRGE in DSEU and UOE tended to be optimized, while those in EOU and DWEU tended to be deteriorating. The conclusion was also verified by the CVs of different input variables in the four categories (Table 2). The CVs of different input variables were 0.117–0.324 and 0.200–0.433 in DSEU and UOE, respectively, and 0.426–0.540 and 0.348–1.165 in EOU and DWEU, respectively.

The annual average WE^IPR values in DSEU, EOU, UOE, and DWEU were 0.2228, 0.2510, 0.7316, and 0.4522, respectively. Among them, the WE^IPR in DSEU generally showed a ‘W’-type upward trend, and the highest values were observed in 2013 and 2019; WE^IPR in EOU generally decreased first and then increased, with the lowest and highest values observed in 2011 and 2019, respectively; the WE^IPR values in UOE and DWEU showed upward trends on the whole. The results indicate that the improvement potential of WRGE in four categories was deteriorating.

In conclusion, there were considerable differences in allocative efficiency and overall utilization efficiency of production factors in different categories during the study period under group-frontier and meta-frontier (Figures 6 and 7).

In the present study, the Meta-US-SBM model was used to analyze the technical heterogeneity of WRGE in different regions. XY, QJ, and TM were the most efficient regions in both group-frontier and meta-frontier over the study period. In addition, DSEU did not necessarily perform optimally based on WRGE. Moreover, TGR in most regions, DSEU, DWEU, and UOE, exhibited a fluctuating and generally decreasing trend and only remained stable at 1 in TM, and exhibited an overall upward trend in EOU. The results are an important basis for addressing several water resource challenges associated with sustainable economic and social development within urban agglomerations.

To facilitate WRGE improvement, the present study considers slack improvement in addition to input (output) improvement to ensure efficiency evaluation results are as accurate as possible. Under group-frontier and meta-frontier, WE^IPR is highly negatively correlated with WRGE, and the potential improvement rate of WREF and GDP varies greatly among regions and categories. However, under group-frontier, in DSEU and EOU, the improvement potential of GDP is significantly higher than that of WREF, with major policy implications for optimizing the allocation of input factors, considering the heterogeneous characteristics of the water-use regions.

This work was supported by the Social Science Foundation of Jiangxi Province (20GL14), the Humanities and Social Science Projects of Colleges and Universities in Jiangxi Province (JC20101), the Science and Technology Projects of the Jiangxi Provincial Education Department (GJJ2200518), and the Ministry of Education in China Layout Project of Humanities and Social Sciences (20YJAZH037). The authors also thank anonymous reviewers for their comments that improved this manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.