Revealing the changes in water footprint at the provincial level and their drivers in the Yellow River Basin, China

Water scarcity has become the most significant limiting factor for sustainable economic and social development in the Yellow River Basin (YRB). Despite this, most current studies on water resources in the YRB from a water footprint (WF) perspective have focused on WF measurements and have explored the influencing factors of WF changes in certain industries, while the impact mechanisms driving regional WF changes remain unclear. To bridge this research gap, the WF of nine provinces in the YRB for 2012 and 2017 were quantified using an environmentally extended multi-regional input-output model (EE-MRIO), and the driving forces of regional WF changes were explored via structural decomposition analysis (SDA). The results showed that the WF of the YRB increased by 3.8% to 113.64 billion m3 between 2012 and 2017. With rapid economic development and enhanced inter-regional trade links, the external WF has played an important role in meeting local water demand. Technological advances and production structure adjustments contribute to the reduction of the WF, thus promoting the sustainable use and management of the YRB’s water resources. Both consumption patterns and final demand per capita have dominated the YRB’s WF growth, particularly in the economically developed middle and lower reaches, where urban household consumption drove the largest WF, accounting for over 40%. Therefore, in the future, continuous optimization of the consumption structure and guidance of green consumption awareness are expected to contribute more to the reduction in WF. The findings of this study reveal the primary causes of WF changes in the YRB and offer a theoretical justification for the formulation of water conservation and sustainable utilization policies.


Introduction
In recent years, the Yellow River Basin (YRB) has faced serious water shortage problems owing to climate change and human activities such as urbanization and farm irrigation, as well as a rapid increase in water demand against the background of rapid economic and social development (Dong et al 2020).Water resources constrain ecological protection efforts and inhibit the optimal development of the YRB.Promoting the development of the YRB requires rational allocation and efficient use of water resources according to the local water resource endowment (Li et al 2023), therefore, a comprehensive understanding of water resource utilization and consumption, as well as clarification of the drivers behind the impact of water resource utilization are vital for alleviating the discrepancy between water resource supply and demand and promoting the sustainable development of the YRB.
The concept of the water footprint (WF), proposed by Hoekstra, refers to the quantity of water resources consumed by a nation, region, or individual while consuming goods and services over a specific time period, thereby truly reflecting actual water consumption and helping to comprehensively describe the utilization and consumption of water resources (Hoekstra 2008, Pan et al 2012).Based on the type of water resources consumed, WF can be divided into three types: blue WF which refers to surface and underground water resources, green WF which refers to rainfall water resources consumed through the evapotranspiration of crops, and grey WF which refers to the amount of freshwater required to assimilate pollutants to meet specific water quality standards (Fu et al 2022).Computational approaches to WF consist of both bottom-up and top-down methods, the former being the most representative of the life-cycle assessment approach, which is a cradle-tograve analysis of the environmental impact of products throughout their lifetime (Gu et al 2015).In comparison, the top-down method is represented by the input-output (IO) model, which quantitatively illustrates the interconnection and interdependence of multiple economic activities.The IO model has several favorable characteristics for WF evaluation (Zhang et al 2012).For example, the IO model, including the single-region IO model (SRIO) and the multi-regional input-output model (MRIO), provides an exhaustive description of supply chains within a single region or across multiple regions, and avoids truncation errors that often occur in bottomup methods (Zhang et al 2018).The IO model can also allocate resources directly to final consumption rather than intermediate consumption, thus effectively assessing the direct/indirect resources and demand/ consumption of a country, region, lifestyle group, or household's final consumption pattern (Feng et al 2012).Therefore, the IO model is frequently employed to assess both virtual water (VW) and WF (Zhao et al 2017, Fan et al 2018, Allan et al 2021).
However, the drivers of WF change have gained extensive attention from scholars as understanding the driving forces can provide pertinent guidelines for alleviating existing water scarcity problems.This can be achieved through factor decomposition methods, including the IPAT model, the index decomposition analysis (IDA) method, and the structural decomposition analysis (SDA) method .Compared to the two previously mentioned methods, the IO theory-based SDA method performs better in the mechanism analysis of the final demand-side influencing factors (Wang et al 2017).Therefore, it is widely used to examine the drivers of pollution emissions and resource consumption in regions and sectors (Wang et al 2016, Mi et al 2017, Sun et al 2021).In terms of WF research, Yang et al (2015) and Feng et al (2015) combined the IO model and the SDA method to explore WF and its drivers in Beijing and Zhangye City, respectively.Using the SDA model, Yang et al (2016) studied the impact of socioeconomic factors on China's total change in the WF from 1997 to 2007.
The YRB is a vital area for food, energy, and industry in China and suffers from water shortages and overutilization as the demand for water resources increases.Therefore, exploring dynamic changes in WF and its driving mechanisms can help to fully understand the changing characteristics of WF and clarify the driving factors behind its impact on water resource utilization, so that corresponding measures can be taken to reduce WF, alleviate local water shortages, and achieve sustainable water resource utilization.However, existing research on water resources in the YRB mainly focuses on the measurement of WF and the study of WF drivers in specific industries, such as agriculture, coal, and chemicals (Liu et al 2022, Xu et al 2022, Zhang et al 2023), while there is insufficient research on the mechanisms driving regional WF changes.In this study, an environmentally extended multi-regional input-output model (EE-MRIO) and structural decomposition analysis (SDA) were used to measure regional WF and investigate the driving factors of regional WF changes in the YRB from 2012 to 2017.
This study aims to reveal the actual local water consumption of the YRB from the perspective of WF and introduces a quantitative analysis tool commonly used in the field of input-output technology, the SDA method, into the analysis of drivers of WF change in the YRB, providing a new perspective for the study of water resource use in this area.The research results provide a theoretical basis for alleviating the discrepancy between water supply and demand and promoting the efficient use of water resources in the YRB.The remainder of this paper is organized as follows.The study area is described in section 2. Section 3 describes the methodology and materials used in this study.The results are presented in section 4. Section 5 discusses the results of the study.Finally, section 6 presents the conclusions of this study.

Study area
With a length of 5,464 km and a basin area of 795,000 km 2 , the Yellow River is the second longest river in China (Song et al 2022a), spanning its eastern, central, and western regions and flowing from west to east through nine provinces (autonomous regions), as shown in figure 1.The YRB is strategically important for safeguarding national socio-economic development and ecological security.With only 2% of China's water resources, the YRB accounts for approximately 12% of the country's population, 15% of its arable land, and 14% of the country's gross economic output.The water resource utilization rate of the YRB exceeds 80%, and water resource shortages severely restrict sustainable development (Dong et al 2020).Therefore, it is essential to examine WF changes and the corresponding driving forces in the YRB to help policymakers formulate tailored water conservation policies to lessen the disparity between the demand and supply of water in the region.These insights will promote the optimal development of the YRB.

Data sources
In view of the fact that as green water exists only in the agricultural sector and cannot be used by other sectors, and the availability of gray water data for each industrial sector varies by province, this study only focuses on blue WF.The data used in this research mainly included the MRIO table data and data on water consumption by sector in various provinces in China.The MRIO table data originate from the 2012and 2017 China Multi-Regional IO table released by the CEADS database (https://www.ceads.net.cn/)(Zheng et al 2020).It should be noted that China updates the MRIO table every five years, and the MRIO in 2017 is the latest data currently available.In addition, as the compilation of the IO table is based on the price data of the year, this study converts the IO table in 2017 into constant price data based on 2012 to eliminate the impact of prices.The specific division of the 42 sectors can be found in Xia et al (2022) and will not be detailed in this paper.The water consumption data for each sector in each province were retrieved from the China Statistical Yearbook (http:// www.stats.gov.cn/tjsj./ndsj/),China Water Resources Bulletin, Provincial and Municipal Water Resources Bulletin (http://123.127.143.131/water_bulletin/f), and China Urban-Rural Construction Statistical Yearbook (https://www.mohurd.gov.cn/gongkai/fdzdgknr/sjfb/tjxx/jstjnj/index.html).In the absence of specific official water use records for each sector, this study decomposed overall water use by industry, construction, and tertiary sectors in physical units proportionally on the basis of the intermediate inputs of the 'water production and supply' sector to each industry sector, measured in monetary units extracted from each province's IO tables (Zhang and Anadon 2014).

Environmentally extended multi-regional IO model (EE-MRIO)
The MRIO model is an IO model containing more than two regions connected by cross-regional flows of goods and services, which provides a comprehensive description of the complete production chain between provinces and sectors.It is mainly used for inter-regional correlation research, including carbon emissions, water resources, and ecological footprints.(Zhou and Imura 2011, Feng et al 2014, Peng et al 2015, Zhang et al 2019).
Table 1 depicts the basic form of the MRIO model, which consists of the m provinces with n respective sectors.Each province's gross output includes the intermediate and final demand utilized by all provinces as well as their total exports.
The corresponding equilibrium equation of the MRIO model is shown as follows: where x i r is the gross economic output of sector i in province r. z ij rs is the intermediate input from sector i in province r to sector j in province s. f i rs denotes the final demand of province s for the goods and services of sector i produced in province r. e i r denotes the exports of sector i in province r.The symbol a ij rs denotes the direct input coefficient; it is defined as the input of the sector i in province r to produce per unit of output of sector j in province s.
The matrix representation of equation ( 1) is as below.
is the aggregate vector of the output for all regions.
the aggregate inter-regional final consumption vector.
In this study, water consumption data from various sectors of each province were introduced into the MRIO model to construct an EE-MRIO model, and the WF driven by regional final demand (consumption WF) was calculated.C c denotes the direct water use coefficient matrix, and c i r is defined as direct water The WF of regional consumption can be calculated as: where C r ˆrefers to the diagonal matrix of the province r's direct water use coefficient.F rs is the final consumption of goods and services of province s from province r.WF rs denotes the water consumption of province r to meet the final demand of province s, that is, the VW transferred from province r to province s.The aggregation of all VW transferred from other provinces to province s ( WF

Structural decomposition analysis (SDA)
SDA, proposed by Leontief, enables the decomposition of a dependent variable's change into the total various forms of change in the pertinent independent variables, thereby measuring each independent variable's contribution to the dependent variable change (Leontief and Daniel 1972).Based on IO analysis, SDA has been commonly adopted to track the influence of environmental variables over time (e.g., energy consumption, carbon emissions, VW, and footprint family) (Zhang et al 2012, Fan et al 2018, Yu et al 2019, Sun et al 2019).In this study, the following five factors are considered with reference to Cai et al (2019) in order to explore their influence on the change of WF in the YRB, and the specific decomposition model is as follows: where WF denotes the water footprint.C represents the water-use efficiency matrix for various sectors in each province; that is, water use per unit of economic production, which reflects the technological level of water consumption in various sectors in each province.L refers to the Leontief inverse, which typically denotes the production structure.F refers to the final demand matrix for different industries in each province, Y refers to the matrix of total demand in each province, and P represents the population.S denotes consumption structure and D is per capita final demand, which represents economic development or affluence.
The following equation describes the WF change (0 denotes the 2012 base period and 1 denotes the 2017 calculation period): was decomposed to provide the contribution of each of the five factors to this change: In the SDA, different decomposition forms yield different results.To handle this, it is best to use the average of the results obtained from all decompositions in the analysis, despite the fact that doing so comes at the cost of the computation being extremely large, as there are more factors to decompose (Dietzenbacher and Los 1998).For example, if there are n decomposition factors, then there are n! kinds of decomposition forms.The average of the two polar decompositions is typically used in current literature to estimate SDA outcomes (Guo 2010, Cai et al 2019).This method was also used in this study to estimate the influence of various factors on WF change.
The contributions of drivers C, L, S, D, and P to the change in WF can be expressed as:

Spatial-temporal distribution of WF in the YRB
This study calculated the WF of the provinces in the YRB in 2012 and 2017.The results are presented in appendix A and figures 2-4.The gross WF at the provincial level in the YRB was 109.43 billion m 3 in 2012.With a WF of 31.20 billion m 3 , Shandong Province had the largest WF, followed by Sichuan Province (18.80 billion m 3 ).The WF of the Qinghai Province was the smallest at 2.34 billion m 3 , which was only 7% of that of the Shandong Province.The total WF at the provincial level in the YRB in 2017 was 113.64 billion m 3 , a 3.8% increase from 2012.In 2017, Sichuan Province had the highest WF (23.05 billion m 3 ), followed by Henan Province (22.01 billion m 3 ).Qinghai Province had the lowest WF (2.21 billion m 3 ).The regional distribution of the WF in the YRB shows that provinces with a high WF are concentrated in the middle and lower reaches.These results were consistent with those reported by Zhang et al (2023).
Figure 3 shows the per capita WF of the provinces in the YRB in 2012 and 2017.As is shown in the figure, unlike the spatial distribution of provincial WF, the upper YRB provinces' per capita WF was generally larger than that of provinces in the middle and lower reaches.Among them, Ningxia had the highest per capita WF with 760.8 m 3 and 762.4 m 3 in 2012 and 2017, respectively.In addition, Qinghai Province, which is located in Northwest China, had the lowest WF but also a high per capita WF because of the vastness of the region and its sparse population (Li et al 2021, Song et al 2022b).Henan and Shandong, provinces with large populations in China, had the lowest per capita WF of 180.9 m 3 and 193.1 m 3 in 2012 and 2017, respectively, with less than onethird of Ningxia's per capita WF.
Figure 4 shows the unit GDP WF for each province in the YRB for 2012 and 2017.In a manner that is analogous to the spatial distribution of WF on a per capita basis, the unit GDP WF of the upper YRB provinces is generally higher than that of the middle and lower provinces.Ningxia, Qinghai, and Gansu had higher unit GDP WF owing to their lower GDP levels.Shaanxi, Henan, and Shandong had a smaller unit GDP WF, and from the perspective of the entire economic system, these provinces had higher water use efficiency.

Variation trend and sources of WF in the YRB
WF changes in the provinces of the YRB are shown in figure 5.As shown in the figure, the WF of Sichuan, Gansu, Ningxia, Shaanxi, Shanxi, and Henan showed an increasing trend.Furthermore, the growth trend was more significant in the provinces in the middle reaches, with growth rates of 45.7%, 32.1%, and 27.7%, respectively.The WF of Qinghai, Inner Mongolia, and Shandong showed decreasing trends, with a largest decrease of 37.9% in the Shandong Province.
Given the extensive economic trade activities that take place between various regions, the WF sources of individual provinces can be separated into local WF and external WF.The external WF can be further decomposed into VW inflows from other provinces within the YRB or outside the basin.Figure 6 shows the WF source composition for each province in the YRB in 2012 and 2017.The WF of the under-developed regions (Qinghai, Gansu, and Ningxia) in 2012 was dominated by local WF, accounting for more than 50% of the total WF source, owing to their relatively deprived economies and weak connections with the outside world.Sichuan Province had a relatively low proportion of external WF (20.5%) owing to its abundant water resources.The external WF of the middle and lower provinces were significantly larger than those of the upper provinces, with Shaanxi and Shandong having the highest external WFs, accounting for over 60%, mainly from VW inflows    As depicted in figure 6, with the intensification of economic and trade activities between regions, the proportion of external WF in the upper YRB provinces (except for Inner Mongolia) increased with varying degrees in 2017, and economic and trade ties with provinces outside the basin were strengthened.The middle YRB provinces (Shaanxi, Shanxi, and Henan) were increasingly connected with other provinces in the basin.The WFs of Shaanxi and Sichuan were primarily derived from the VW inflow of Gansu, whereas the WFs of Shanxi and Henan were mainly derived from Inner Mongolia and Shaanxi, respectively (see figure 7(b)).The proportion of local WF in the lower provinces of the YRB (e.g., Shandong) underwent a significant increase (37%→58%), with a corresponding reduction in linkages with other provinces within and outside the basin.Overall, although the proportion of external WF in the middle and lower provinces of the YRB decreased in 2017 compared to that in 2012, it remained at a higher level than that of the upper provinces.

Contribution of the drivers to WF changes in the YRB
The YRB's overall WF grew from 109.43 to 113.64 billion m 3 over the period of 2012-2017.The areas that contributed the most to this increase were the provinces in the upper (4.98 billion m 3 ) and middle (11.06 billion m 3 ) reaches.The provinces in the lower reaches decreased the YRB's total WF by 11.82 billion m 3 .Figure 8 shows the contributions of the drivers to the total WF change in the YRB during 2012-2017.As can be seen from the figure, the YRB's total WF increased primarily as a result of changes in the consumption structure and per capita final demand, both of which led to an increase in the YRB's total WF in the amount of 61.04 billion m 3 (approximately equal to 14.50 times the net increase of 4.21 billion m 3 ) and 59.81 billion m 3 (approximately equal to 14.20 times the net increase), respectively.During the study period, water use efficiency (a proxy of technology level, also known as the technology effect) was the most vital factor in offsetting the YRB's total WF growth, resulting in a reduction of 97.95 billion m 3 (approximately equal to 23.27 times the net increase) in the total WF in the YRB.In addition, adjustments in the production structure also reduced the WF by 21.00 billion m 3 (approximately equal to 4.99 times the net increase).The study also found that population had the lowest impact on the YRB's total WF, driving 2.31 billion m 3 of WF growth over the study period.

Contribution of the drivers to provincial WF changes in the YRB
Figure 9 and appendix B show the contributions of the five drivers to the provincial WF changes in the YRB from 2012 to 2017.As previously mentioned, the total WF of the YRB increased by 4.21 billion m 3 .Per capita final demand was an enormous driver of provincial WF growth, especially in Ningxia (with an increase approximately equal to 1897.22% of the net change of 0.36 billion m 3 , i.e., 6.83 billion m 3 ), Sichuan (307.31%,13.03 billion m 3 ), and Gansu (183.33%,3.30 billion m 3 ), and was a dominant factor driving the growth of the WF in these provinces.Per capita final demand reflects the level of social development.As economic growth and living standards improve, consumer demand for food, clothing, shelter, transportation, and other infrastructure has increased, directly or indirectly leading to an increase in WF.In addition to per capita final demand, changes in the consumption structure (ΔS) also actively drove provincial WF growth except Qinghai, Sichuan and Ningxia, dominating the increase of WF in Henan (496.65%,23.69 billion m 3 ), Inner Mongolia (290.00%,3.77 billion m 3 ), Shandong (251.23%,29.72 billion m 3 ), Shaanxi (165.89%,5.69 billion m 3 ), and Shanxi (148.25%,4.24 billion m 3 ).WF changes driven by population effects were positively correlated with population changes.The Shanxi, Inner Mongolia, and Gansu provinces experienced negative population growth during the study period, whereas in Shandong, the most populous province in the YRB, the population increased from 97.08 million in 2012 to 100.33 million in 2017 (NBSC, 2013(NBSC, , 2018)).Population growth has caused a huge increase in the need for materials and energy, which has led to an increase in the provincial WF.Nevertheless, the effect of population on WF changes was relatively small compared to that of other driving forces.The expansion of WF in all provinces in the YRB was offset to some extent by an increase in water use efficiency.The technology effect made the largest contribution to WF reduction in Shandong and Henan, where WF was reduced by 43.80 and 26.55 billion m 3 , respectively.In comparison, the impact of technology on reducing the WF of Gansu, Ningxia, Inner Mongolia, and Qinghai was limited due to their smaller WF bases.However, the largest decreases driven by the technology effect relative to the total changes in each province's own WF were in Qinghai (approximately equal to −2661.54% of the net change of 0.13 billion m 3 , i.e., −3.46 billion m 3 ), Ningxia (−627.78%,−2.26 billion m 3 ), Henan (−556.60%,−26.55 billion m 3 ), and Shandong (−370.25%,−43.80 billion m 3 ).In addition to water use efficiency, adjustments in the production structure drove the decline in the WF in all provinces, except Qinghai.The study found that the total GDP of Qinghai Province in 2017 was 264.280 billion yuan, with the added value of the secondary industry accounting for 118.038 billion yuan (44.7%), reflecting that the industrial system of Qinghai focused on resource development, dominated by heavy industry, and supplemented by light industry.Among them, the rolling processing industry, non-ferrous metal smelting, electricity production, heat production, the chemical industry, and oil and gas extraction industry had developed prominently, which were the advantageous industries of Qinghai and also required high water consumption.The results of the structural decomposition indicated that the current pattern of industrial structures in Qinghai is not conducive to the protection of water resources; therefore, further adjustments are imperatively required.In conclusion, there were significant regional differences in the impacts of drivers on provincial WF.

The harmonious connection that exists between economic expansion and water resource consumption
To describe the connection between economic expansion and water resource consumption, we quantified the decoupling index between these two indicators in the YRB from 2012 to 2017 using a WF perspective, as detailed in appendix C. In terms of the decoupling results, among the nine provinces in the YRB, only Qinghai, Inner Mongolia, and Shandong achieved strong decoupling from 2012 to 2017, indicating that economic expansion occurred in conjunction with a decline in water consumption and high-quality harmony between the two was achieved.Previous studies have shown that the aforementioned provinces have higher efficiency in agricultural water resource use than other provinces in the YRB (Bai et al 2022).As shown in figure 10, water use efficiency effectively contributed to the reduction in the WF in these three provinces during the study period.Sichuan, Ningxia, and Henan were in a weakly decoupled state during the study period, indicating that their water use increased at a relatively slow rate compared to their economic growth.As national water rights pilots, Ningxia and Henan, on the one hand, clarified the water volume available to water users through water rights allocation, which significantly influences users' behaviors and compels them to increase the effectiveness of their water resource use.On the other hand, water rights trading makes water rights a resource with market value.Through the market price mechanism, water rights owners with low water use efficiency are forced to save water by considering the opportunity cost of water use and actively transferring idle water rights to water users with high marginal benefits of water use.Therefore, new or potential water users have the opportunity to obtain water resources needed for development, thus improving the overall water use efficiency of society (Tian et al 2020).Gansu, Shaanxi, and Shanxi were in an expansive couple state during the study period, indicating that economic expansions in these provinces is strongly dependent on resource consumption, which reflects their extensive and unsustainable economic growth.Additionally, rapid economic growth has effectively improved people's standard of living, and changes in the consumption structure and per capita final demand have contributed to the growth of the WF in Gansu, Shaanxi, and Shanxi.

Technology effect had the greatest impact on reducing WF
The impact of each driver on provincial WF change in the YRB is shown in figure 10.The greatest impact on WF reduction was the increase in water usage efficiency due to technological advancements, in agreement with Liu et al (2022) and Chen et al (2021).The YRB is an important part of the Chinese economy in terms of both agricultural production and industrial development.Therefore, this study focused on the adoption of water conservation technologies for agricultural and industrial production.From 2012 to 2017, the government issued a series of national policy documents related to agricultural water use, including but not limited to the National Agricultural Water Conservation Program (2012-2020), the concept of contractual water conservation management, and the implementation of the inter-regional transfer of water rights in the YRB.Promoting water-saving technology transformations to change the traditional agricultural irrigation mode and improve water efficiency became the theme of agricultural production water during this period.The adoption of technologies such as efficient water-saving irrigation and irrigation channel impermeability can reduce inefficient water consumption and improve water resource usage (Shang et al 2019).During the study period, the proportion of agricultural land in the YRB irrigated using efficient water-saving methods increased from 55% to 60% (MWRPRC 2013, 2018, MAPRC 2013, 2018).In terms of industry, as mentioned earlier, the YRB is a significant industrial base in China, relying on its abundant resources and energy, and it has an obvious industrial structure with high water usage.The adoption of advanced water-saving technologies is crucial.For example, deep processing of coal in Erdos City, Inner Mongolia, has reduced water consumption by approximately 40% through the adoption of air-cooling, closed-circuit recycling, and wastewater purification and reuse technologies (Jia et al 2017).In summary, improvements in water use efficiency have significantly reduced the WF of the YRB.
Although China's water-use efficiency has improved over the past few decades, agricultural water use in the YRB has been large and inefficient.Sichuan, Qinghai, Ningxia, Shanxi, and Inner Mongolia had effective water utilization coefficients for agricultural irrigation in 2019, which were below the national average of 0.568 at 0.477, 0.5, 0.543, 0.546, and 0.547, respectively (MWRPRC 2020).Compared with the average in developed countries, there is still a significant gap (Jiang 2015, Cai et al 2019).Therefore, in the future, improving water efficiency will be an important way to reduce the WF.We suggest that in the future, financial investment and digital reform should be strengthened to promote the modernization of irrigated areas in the YRB and the largescale promotion and application of water-saving technologies.In addition, economic incentives, such as raising water prices for agricultural production and promoting agricultural water price reforms, can be used as potential tools for improving water-use efficiency; however, social factors must also be considered (Kang et al 2017, Cai et al 2019).For instance, the price of agricultural water can be reasonably determined based on the scarcity of water resources and the affordability of water to users.It is also possible to explore the implementation of classified water prices at the end-use level by distinguishing between different types of water use, such as food crops, cash crops, and aquaculture.Additionally, the government can reward users who save water based on the amount of water saved.Water savings can also be achieved through water rights on a paid basis to achieve greater efficiency and effectiveness in the use of water in agriculture.Advanced water conservation processes, technologies, and equipment, such as efficient cooling and washing, wastewater recycling, and the replacement of high-water consumption production processes, should be promoted to increase water resource effectiveness in key industries, such as steel, petrochemicals, and non-ferrous metals.
5.3.Consumption structure and per capita final demand drove a large WF increase Our findings reveal that S and D drive a significant increase in the consumption-based WF.To further discuss the differences in WF driven by different consumption categories and make recommendations accordingly, this study further refines the consumption-based WF. Figure 11 shows the proportion of WF driven by the five final demands in the nine provinces of the YRB in 2012 and 2017.Urban household consumption was the primary contributor to the total WF of the YRB, accounting for more than 40% and showing an upward trend from 2012 to 2017.The average urbanization rate in the nine provinces of the YRB increased from 48% in 2012 to 57% in 2017 (NBSC 2021).With rapid economic development and urbanization, residents' incomes have grown rapidly, and the standard of living has improved significantly.On the one hand, an increase in the proportion of residents' food expenditure will directly increase water consumption due to increased cooking, washing dishes, and other related living needs.On the other hand, the shift in the structure of residents' food consumption, such as consuming more water-intensive foods, like meat and dairy products (FAO 2020), will indirectly lead to a food-related WF increase.Given the dominant role of S and D in the increase in provincial WF, in addition to technology-based water conservation, consumer behavior-based water conservation may play a greater role in alleviating water scarcity in the future, especially for a developing country like China, with a population of 1.4 billion people.Existing research suggests that resident environmental awareness, sense of responsibility, lifestyle, and consumption habits are the most important determinants of water conservation behavior (Singha et al 2023).When people feel that they are at risk of water scarcity, there is a high probability that they will actively adopt water conservation behaviors.Therefore, we suggest that the government strengthen the role of education and guidance to help citizens fully understand the importance of water conservation and raise their awareness of water conservation.The government should promote a lifestyle for water conservation as well as helping citizens develop good water-saving habits.We also suggest that residents should be rationally guided to optimize their food consumption structure by increasing their intake of fibrous foods such as grains and reducing food waste to decrease food-associated WF (Vanham et al 2013, Walmsley et al 2015).Moreover, the government could introduce relevant policies on residential consumption, such as including the WF of products on labels and quantifying the impact of public choices on water consumption, to guide green consumption.Furthermore, gross fixed capital formation was a secondary contributor to the YRB's total WF.In 2017, fixed capital formation in Inner Mongolia and Henan became the largest contributors to the WF, accounting for 37% and 36% of the gross WF, respectively.Since the implementation of the Western Development Strategy, Inner Mongolia has gradually constructed medium-and large -sized key projects such as the Chuole, Nierji, and Wanjiazhai water conservancy projects (Li and Sheng 2010).With the implementation of major strategies for the Rise of Central China, the total fixed asset investment in Henan increased by 107.4%, from 2,145 billion yuan in 2012 to 4,449.7 billion yuan in 2017, leading nine provinces in the YRB (NBS-DCS and NBS-DRS 2013; HZSB 2018).The percentage of WF driven by gross fixed capital in Ningxia also showed a clear upward trend (12% → 27%).Along with the implementation of the two aforementioned national strategies, infrastructure such as transportation, water conservancy facilities, and power grids have been extensively developed; therefore, physical capital investment has driven construction activities and increased water consumption.Therefore, we suggest that, in the future, water-saving concepts based on green buildings should be implemented in construction projects to achieve effective water conservation.For example, the active use of fabricated building construction methods, development of reclaimed water reuse, and rainwater utilization technologies further strengthen the water conservation of green buildings (Södersten et al 2020).

Conclusions
Exploring the drivers of water footprint changes can support efforts to reduce the water footprint, which is an important step towards facilitating future water policymaking.This study adopted an EE-MRIO model to estimate the WF at the provincial level in the YRB in 2012 and 2017, and explored the driving forces of WF change via SDA.The results showed that the WF of the YRB increased by 3.8% to 113.64 billion m 3 between 2012 and 2017, with the WF of the midstream and downstream regions being significantly higher than that of the upstream regions, whereas the spatial distributions of per capita WF and unit GDP WF were the opposite.In addition, except for Qinghai, Inner Mongolia, and Shandong, all the provinces showed different degrees of WF increase, among which the WF of the middle reaches increased the most.The SDA results identified consumption patterns and per capita final demand as the dominant factors driving the increase in the water footprint of the YRB, whereas technology and production structure effects largely offset this increase.In addition, there were significant variations in the effects of the drivers on provincial WF changes in the YRB, reflecting the varying challenges faced by stakeholders in different provinces and sectors in their efforts to ensure the security and sustainability of water resources.The findings of this study reveal the changes in and drivers of the WF in the YRB, which can provide theoretical support and a reference for policy makers when formulating policies to reduce the WF and promote sustainable water resource use and management in the YRB.This study had certain limitations.In terms of data, owing to the limited publication of MRIO table data (once every five years), our study only examined WF and the driving forces of WF change in the YRB in 2017 and before, and could not describe the latest water resource consumption situation of the YRB.In terms of the model, the SDA model used in this study was based on the assumption that all of the considered factors are completely independent of each other.However, this assumption may cause some bias in the decomposition results, as the factors are likely inter-related to some extent.Further research will be conducted to address the limitations presented above and enhance the timeliness and reliability of the findings.

Appendix B
Structural decomposition of water footprint changes in nine provinces of the YRB from 2012 to 2017 (billion m 3 ).

Figure 1 .
Figure 1.Location of YRB in China and provinces through which the yellow river flows.

)
is the gross water footprint of province s, which includes both the local (WF local s ) and external water footprints (WF external s ) (Chen et al 2017, Zhang et al 2019).When r s,= WF rs denotes the local water footprint of province s, i.e., province s consumes water resources within the region to produce products consumed in the same region.When r s, represents the external water footprint of province s, i.e., province s consumes water resources from other regions to produce products consumed in that region, and also stands for the VW import of province s.denotes the VW exports of province s.
and P D represent changes in water-use efficiency, production structure, consumption structure or pattern, per capita final demand, and population, respectively.

Figure 5 .
Figure 5. Changes in the WF of provinces in the YRB.

Figure 6 .
Figure 6.The source composition of the WF of provinces in the YRB.The column on the left depicts the situation in 2012, and that on the right depicts the situation in 2017.

Figure 7 .
Figure 7. VW flow between provinces in the internal YRB.VW flow within a province, for example, the VW flow from Qinghai to Qinghai was not considered.

Figure 8 .
Figure 8. Contributions of drivers to the YRB's total WF change from 2012 to 2017.

Figure 9 .
Figure 9. Results of the SDA of drivers of provincial WF change in the YRB from 2012 to 2017.

Figure 10 .
Figure 10.Impact of drivers on provincial WF change in the YRB (unit: %).

Figure 11 .
Figure 11.The proportion of WF driven by the five final demands in the YRB in 2012 and 2017.

Table 1 .
The fundamental form of the MRIO model.