Comparison and Prediction of the Ecological Footprint of Water Resources—Taking Guizhou Province as an Example

Abstract

Water resources are considered to be of paramount importance to the natural world on a global scale, being critical for the sustenance of ecosystems, the support of life, and the achievement of sustainable development. However, these resources are under threat from climate change, population growth, urbanization and pollution. This necessitates the development of robust and effective assessment methods to ensure their sustainable use. Although assessing the ecological footprint (EF) of urban water systems plays a critical role in advancing sustainable cities and managing water assets, existing research has largely overlooked the application of geospatial visualization techniques in evaluating resource allocation strategies within karst mountain watersheds, an oversight this study aims to correct through innovative methodological integration. This research establishes an evaluation framework for predicting water resource availability in Guizhou through the synergistic application of three methodologies: (1) the water-based ecological accounting framework (WEF), (2) ecosystem service thresholds defined by the water ecological carrying capacity of water resources (WECC) thresholds, and (3) composite sustainability metrics, all correlated with contemporary hydrological utilization profiles. Spatiotemporal patterns were quantified across the province’s nine administrative divisions during the 2013–2022 period through time-series analysis, with subsequent WEF projections for 2023–2027 generated via Long Short-Term Memory (LSTM) temporal forecasting techniques.

1. Introduction

Functioning as both a critical ecological cornerstone and a key driver of macroeconomic stability, water resource stewardship serves as a critical enabler for maintaining equilibrium within dynamic socioecological frameworks. Its multi-scalar governance directly determines the viability of biosphere-coupled development paradigms [1,2]. Owing to the essential role of water in sustaining ecosystems and human societies, the global governance of freshwater systems has become a core policy objective for advancing sustainable development agendas [3]. The escalating challenge of water shortages jeopardizes both ecological stability and societal progress, thereby establishing hydrological security as an enduring academic concern across multidisciplinary fields. Guizhou, a region characterized by its karst topography and significant ecological importance as an upper-reach barrier for the Yangtze and Pearl Rivers, exhibits a comparatively abundant water supply, though with considerable spatial and temporal variability. The province is confronted with persistent water shortages, both in terms of spatial distribution and engineering demands, which are identified as primary impediments to achieving optimal development. The economy and society of Guizhou are currently underdeveloped and imbalanced, with a limited economic base and inadequate utilization of water resources, particularly in comparison to other regions of China. This research applies the WEF to assess current dynamics and projected trajectories of water consumption patterns and ecological carrying thresholds in nine prefecture-level administrative units within Guizhou Province. The analytical outcomes establish empirical evidence for developing region-specific water governance policies and alleviating anthropogenic stress on aquatic ecosystems in this ecologically fragile karst terrain.

The framework of the EF was originally formulated by Rees et al. during the early 1992 [4], with subsequent refinements introduced by Wackernagel and colleagues in 1998. This analytical approach quantifies regional sustainability through a comparative assessment between productive land/water areas demanded by human populations and the availability of such resources within specific geographic boundaries, thereby establishing an evaluative system for ecological carrying capacity measurement [5].

Employing a spatial Durbin modeling framework, Sun et al. [6] quantified interdependencies between emergency ecological footprints (EEFs) and eco-efficient industrial output across the Yangtze River Economic Belt (YREB). The results indicate that water resource limitations critically modulate EEF–industrial green GDP synergies, with the YREB exhibiting the following: (1) statistically robust EEF-driven productivity enhancements, (2) transboundary spillover coefficients averaging 0.32 via spatial autoregressive linkages, and (3) stratified responsiveness differentiated by upper/middle/lower basin developmental gradients. Developed by Zhang et al. [7], the ecological footprint account for the Tarim River Basin (TRB) was utilized to assess the watershed’s sustainable development status by integrating the landscape ecological risk index. Furthermore, their research investigated approaches for ecological security regulation within the watershed through multi-objective planning and scenario-based simulations. With the gradual maturation of ecological footprint theory, its application has expanded to numerous study regions. The conceptual framework of ecological footprint measurement has undergone systematic refinement through iterative methodological advancements. Within this paradigm, hydrocentric metrics encompassing virtual water tracking (water footprint) and water–energy–food nexus WEF analysis have emerged as dominant methodological approaches for quantifying sustainable water governance performance across spatial scales. A “water footprint” is defined in ISO 14046:2014, Principles, Requirements and Guidelines for Water Footprinting in Environmental Management, as an indicator that quantifies the potential environmental impacts associated with water [ISO 14064:2014] [8]. The ecological footprint quantification framework initially formalized by Wackernagel’s [9] research consortium, grounded in biosystem resilience thresholds, adopts a six-dimensional spatial accounting architecture. This modeling paradigm systematically integrates regenerative land classes (arboreal systems, pastoral ecotones, and agricultural domains) with anthropogenic energy-transformation buffer zones and hydrological footprint indicators. Huang et al. [10] established a WEF model, in which a water account was added to characterize regional water consumption and water production by land area. Consequently, in contrast to conventional water footprint metrics, the WEF provides dual consideration of resource depletion patterns and replenishment capacities, rendering it superior for evaluating hydro-socioeconomic sustainability at subnational scales. Empirical studies demonstrate that WEF quantification methodologies offer robust mechanisms for diagnosing systemic water governance challenges through incorporating mass-balance principles. Internationally, research efforts predominantly employ water footprint analyses at macroscale administrative boundaries (e.g., nation-states) to perform holistic evaluations of gross hydrological budgets. While extant WEF investigations concentrate on developing localized accounting protocols—particularly for discrete geographies or commodity production chains—and subsequently deriving water security indices, critical knowledge gaps persist. Systematic inter-regional comparisons of WEF spatial patterns and drivers remain underexplored, significantly limiting cross-jurisdictional policy interoperability.

A substantial corpus of research has previously been dedicated to the study of water resource sustainability in China’s urban regions. A longitudinal assessment spanning 2012 to 2021 was performed by Li et al. to quantify hydrological allocation dynamics within China’s principal metropolitan clusters: the Beijing–Tianjin–Hebei metropolitan cluster, Yangtze River Delta urban agglomeration, Pearl River Delta megalopolitan corridor, and Chengdu–Chongqing regional economic nexus. Utilizing the water ecological footprint theory, they conducted a comparative assessment of regional water resource sustainability [11]. Building on this foundation, Bin et al. developed a more sophisticated analytical framework for the Beijing–Tianjin–Hebei region (2010–2021). Their methodology integrated multiple models: (1) a hydrological footprint accounting framework, (2) ecosystem load capacity assessment system, and (3) sustainable water utilization metrics. This multi-model approach enabled detailed spatiotemporal analysis of water resource dynamics. To resolve hydrologic forecasting uncertainties, a gated recurrent neural architecture with LSTM was implemented for multi-temporal simulation of aquatic ecosystem service indicators across 2022–2026 cycles. Quantitative verification revealed that adaptive machine learning algorithms achieved 15.3% higher Nash–Sutcliffe efficiency (NSE) than conventional regression models, establishing their superiority in addressing data sparsity limitations inherent to water resources informatics [12]. Employing the WEF, prior investigations [13] have quantified the spatiotemporal variability in WEF indices and environmental carrying thresholds across key urban centers within the MinTuo River Basin (2010–2020). This analysis incorporates a multidimensional sustainability assessment via four hydro-ecological metrics: (1) aquatic ecosystem service surplus/deficit ratios; (2) WEF intensity per CNY 10,000 of economic output; (3) pressure coefficients reflecting water resource utilization limits; and (4) hydro-economic harmonization coefficients. Qin et al. [14] used the WEF model, system dynamics method and scenario analysis method, and designed four development scenarios, namely, maintaining the status quo (JS1), economic growth (JS2), water conservation (JS3), and comprehensive development (JS4), by constructing the system dynamics model of water resource utilization in Jiangxi Province, and conducted a WEF study for the current (1999–2021) and future (2022–2050) WEF in Jiangxi Province. The WEF in Jiangxi Province was evaluated and predicted for the present (1999–2021) and future (2022–2050) scenarios. Extending the water–energy–food nexus analytic framework, Wang et al. [15]. (2023) conducted a spatiotemporal evaluation of resource interdependencies in the Yellow River’s lower catchment through logarithmic mean divisia index (LMDI) decomposition (2007–2020 temporal scope). They subsequently implemented the GM (1,1) gray prediction model to forecast water–energy–food system trajectories over the 2021–2030 period. This integrated methodology establishes analytical foundations for dual objectives: quantifying hydrologic resource carrying capacity thresholds and developing adaptive management protocols under coupled climatic–anthropogenic stressors. By implementing the ecological footprint analytical framework, Wei and colleagues [16] formulated a tri-metric evaluation system to quantify spatiotemporal variations in ecological carrying capacity and WEF nexus dynamics across provincial jurisdictions within the Yellow River Basin during the period of 2005–2019. Complementarily, the researchers employed an integrated system dynamics model to project evolutionary trajectories of these coupled systems through 2030. Wang et al. [17] utilized the system dynamics (SD) method and the ecological footprint model to design four different development scenarios (including the status quo continuation scenario (ES1), the economic development scenario (ES2), the water conservation scenario (ES3), and the coordinated development scenario (ES4)) by constructing an SD model of water resource utilization in Hebei Province, and analyzed the development trends in the status quo (2006–2021) and future (2020) using a combination model. Four different development scenarios (including ES1, ES2, ES3 and ES4) were designed, and the ecological footprints of water resources in the present (2006–2021) and the future (2022–2050) were evaluated and predicted.

A safety assessment of hydro-environmental sustainability across five Ningxia municipalities was performed by Chang’s [18] research team using the WEF nexus metrics. Empirical findings indicate pronounced hydrological stress within the study area, with all evaluated urban centers exhibiting measurable ecological pressure and persistent resource deficits. Spatial analysis revealed distinct heterogeneity: (1) Yinchuan demonstrated conflicting metrics—despite presenting the most severe supply–demand imbalance, its adaptive governance mechanisms maintained hydraulic resilience; (2) Zhongwei’s composite indicators suggested systemic vulnerability, combining acute deficit magnitudes with diminished assimilative capacity; (3) Guyuan exhibited moderated deficits concurrent with constrained ecological buffering potential. Leveraging the Penman–Monteith equations with crop-specific evapotranspiration coefficients, a recent investigation [19] systematically quantified the hydrologic fluxes characterizing agricultural production in the Yangtze Delta region. The study engineered three novel metrics: Composite Water Footprint Stress (AWS), Irrigation Water Scarcity Index (AWS_(blue)), and Rainfall-Driven Water Deficit (AWS_(green)), applied to six staple crops. Through scenario-based modeling calibrated with historical hydrometeorological datasets, the research team mapped geospatial variability in water resource security thresholds and projected 2030 depletion patterns for both anthropogenic and precipitation-sourced agricultural water systems. Through the development of a multidimensional metric framework, Qiao and co-researchers [20] systematically characterized the spatiotemporal dynamics associated with the Beijing–Tianjin–Hebei urban agglomeration’s modernization trajectory. This investigation employed a water–energy–food nexus analytical model to quantitatively delineate the resource utilization patterns across the studied metropolitan network. To address hydrological sustainability challenges and stressor mitigation in Jilin Province, Zhu and colleagues [21] executed a geotemporal evaluation of the water–energy–food nexus (WEFN) through ecological footprint analytic frameworks. Their methodology systematically evaluated resource allocation patterns across the region’s heterogeneous ecosystems. Early-stage analytical frameworks developed through Zhang’s [22] academic consortium implemented the Tapio elastic decoupling formalism, systematically examining interdimensional detachment mechanisms between water–energy–food (WEF) metrics and subnational GDP progression patterns. Current karst hydrology studies, however, inadequately address water–energy–food (WEF) nexus modeling, especially in deploying LSTM networks for multiscale WEF forecasting across geologically fragmented regions exemplified by Guizhou’s karst terrain. This computational approach establishes a novel methodological framework for evaluating urban water resource sustainability in fragile karst ecosystems.

2. Study Zone Characteristics and Dataset Provenance Analysis

2.1. Composite Analysis of the Research Sector

Geopolitically positioned between latitude 24°37′ N–29°13′ N and longitude 103°36′ E–109°35′ E, Guizhou occupies a transitional zone where three tectonic plates converge. Administratively, it shares boundaries with Hunan and Guangxi Zhuang Autonomous Region along its northeastern and southeastern margins, while exhibiting karst-dominated geological continuity with Yunnan and Sichuan through its southwest corridor. It is bordered by Chongqing to the north, and by the city of Guiyang, the prefecture-level city of Zunyi, the prefecture-level city of Boluo, the prefecture-level city of Anshun, and the prefecture-level city of Liupanshui. The province is further subdivided into four autonomous prefectures: the prefecture-level city of Khamyian, the prefecture-level city of Qiannan, and the prefecture-level cities of Biaojia and Tianzhu. The topography of Guizhou is characterized by a high degree of complexity, with a predominance of rock formations. Geospatial analyses quantify Guizhou’s lithospheric configuration as exhibiting extreme terrain diversification, where satellite-derived slope classifications identify 146,000 km² of orogenic belts and karstic plateaus, accounting for 92.5% spatial dominance across this Southwestern China province. The province’s geographical conditions are diverse, with a preponderance of karst topography. The distribution of these geological features is extensive and varied, contributing to the formation of a distinct karst hydrological system. Fluvial networks within the province exhibit a dendritic drainage patterning, categorized into eight principal systems spanning five distinct hydrodynamic domains: (1) upstream Yangtze River tributary segments, (2) downstream Yangtze corridors, (3) Honghe River basin catchment areas, (4) Xiangjiang River source regions, and (5) Jinsha River headwater zones. The topography of the region is predominantly mountainous and hilly, with a high percentage of the terrain classified as rock-based. The region is characterized by a diverse array of geological features, including a wide distribution of rock fissures. These geological features are distributed throughout the landscape, forming a distinct geological system. The geomorphological evolution of this area shows pronounced hydrological modulation from major aquatic systems, notably encompassing: (1) Yangtze River’s upper/lower basin segments, (2) Red River’s fluvial networks spanning upstream/downstream zones, and (3) dendritic tributary configurations within the Jianjiang River’s headwaters. Hydrological zoning follows operational definitions specified in 20250781-T-332 Technical Specifications for Water Resources Zoning (derived from the National Water Resources Comprehensive Plan), employing a standardized hierarchical framework that prioritizes watershed lithology thresholds and annual runoff variability indexes [23]. This classification was compared with the results of previous assessments and utilization plans for water resources, which were conducted from 1980 onwards. Under a hierarchical classification framework, Guizhou’s hydrological zoning system establishes the following: (1) first-tier units, two principal divisions corresponding to the headwaters of the Yangtze River basin and Pearl River basin; (2) second-tier subdivisions, six subsidiary basins encompassing downstream reaches of major river systems (Yangtze, Pearl, Wu, Xiangjiang, Dian, Hong-Liangjiang), with explicit geospatial boundaries validated through DEM analysis; (3) third-tier sub-basins, eleven distinct sub-basins demarcated in compliance with the Technical Guidelines on Ecological Zoning Control General Outline (Draft for Comment) (EIAO Marking Letter [2024] No. 30), supporting finer hydrosystem management interventions [24]. Guizhou’s provincial landscape exhibits a hierarchical hydrological division framework, structured across three tiers: (1) two broad-scale zones corresponding to the upper basins of Yangtze and Jialing river systems; (2) six subregional segments delineated sequentially as lower basin sections of Yangtze, Jialing, Xiangjiang, Diancang, Jianhe, and Jiuquan watercourses; (3) an 11-unit tertiary schema encompassing the upper basins of Xiangjiang, Jialing, Jianhe, Jiuquan, and Diancang valley systems, supplemented by the lower reaches of Diancang, Xiangjiang, Jialing, and Jiuquan [25]. Figure 1 shows the location of Guizhou Province. Figure 2 shows the jurisdictional subdivisions and water resource allocation patterns within Guizhou Province. The color scheme serves solely to differentiate between distinct geographical regions and does not represent quantitative data.

Hydrologically, Guizhou’s defining characteristics stem from its geographic position in southwestern China’s interior and its unique karst-dominated lithostratigraphic framework. The region exhibits significant spatiotemporal heterogeneity in precipitation-to-storage conversion efficiency, compounded by low hydraulic retention coefficients inherent to karstic aquifers. Such hydrogeological constraints induce structural deficiencies in water supply infrastructure development, posing a critical constraint on regional socioecological resilience and industrial productivity optimization.

The fluvial networks of Guizhou serve as critical environmental buffers strategically positioned in the headwater regions of the Yangtze–Pearl River basins to stabilize ecosystem integrity. The region is characterized by a substantial network of rivers, with a total of 4696 rivers recorded within the province. Notably, 1059 of these rivers possess a watershed area of 50 km² or more, underscoring the region’s significant hydrological diversity. The Miao Ridge functions as the primary hydrological divide, demarking the northern Yangtze basin that contains four major fluvial networks: Wu River, Yuan River, Niulan River, along with the composite systems of Hengjiang–Chishui–Qijiang. This hydrologic domain occupies 115,800 km² (65.7% areal coverage), representing the predominant territorial component. In contrast, the Pearl River Basin in southern Guizhou comprises four principal drainage systems—Nanpanjiang, Beipanjiang, Hongshui, and Liujiang rivers—with cumulative catchment areas totaling 60,400 km² (34.3% provincial coverage).

In regard to water resources, Guizhou is characterized by an abundance of water resources, as evidenced by its average annual rainfall of 1179 mm. Long-term hydrologic records demonstrate a mean annual surface water availability of 106.2 × 10⁹ m³ in Guizhou Province, ranking ninth nationally in water resource endowment. The Yangtze River system dominates this allocation, contributing 68 × 10⁹ m³ (64% of provincial total), while the Pearl River system retains 38.2 × 10⁹ m³ (36% share). According to the Seventh National Population Census, the province sustains 38.56 million inhabitants, yielding a per capita water availability of 2754 m³/yr—equivalent to 140% of the nationwide mean value.

Water resource characteristics manifest through three dimensions: (1) Spatiotemporal het erogeneity: Annual water fluxes demonstrate extreme interannual variability (peak: 137.6 × 10⁹ m³; trough: 62.6 × 10⁹ m³). Hydrological seasonality dictates that four consecutive months (June–September) contribute >60% of total annual discharge. Spatial yields span 300,000–1.1 × 10⁶ m³/km², exhibiting distinct geographical gradients—greater water productivity in southeastern highlands versus northwestern plains. (2) Demand–supply mismatch: Hydroeconomic analysis reveals incongruity between resource distribution patterns and socioeconomic development zones. The predominantly populated and economically advanced regions cluster along the Yangtze–Pearl watershed corridors, with the Qianzhong Economic Zone constituting a prominent case—generating 46% and 57% of provincial economic output GDP while containing merely 27% of regional water reserves. The exploitative management of such aquatic assets is further impeded by dissolution-dominated geomorphology, manifesting as alternating elevated ridges and depressed valleys with limited surface runoff capacities—key signatures of karst hydrogeological systems. Operational complexities are exacerbated by the strategic placement of major water impoundments (>100 million m³ capacity) in topographically constrained areas [26].

2.2. Dimensional Data Origins and Principal Metric Attributes

The hydrological dataset integrated in this research comprises three key variables—mean annual precipitation, aggregate water resources, and cross-sectoral water withdrawal metrics—extracted from the Guizhou Water Resources Bulletin (2013–2022). Corresponding social–economic indicators, specifically population size and gross domestic product, were acquired through systematic retrieval of the Guizhou Statistical Yearbook records spanning the identical decadal period (2013–2022). Hydro-productive potential across metropolitan zones was quantified through integrative analysis of decadal freshwater inventories (2012–2021 annual averages) concurrent with areal topographic parameters. This dual-parameter approach enabled the quantification of water-producing faculties per urban agglomeration.

3. Research Methodology

3.1. WEF Model

The WEF model, which has been derived from the ecological footprint theory, has been developed as a means to quantify regional water consumption. This is achieved by converting actual water use into standardized footprint accounts. For Guizhou Province’s nine prefecture-level divisions, the WEF has been categorized into three functional accounts based on local water utilization patterns: water consumption in production, household, and ecological sectors. The WEF index derives from the tripartite computational methodology outlined below:

$$F_W=N{f}_W={\gamma }_{W}W/P_W$$

(1)

$$F_P={\gamma }_W{W}_P/P_W$$

(2)

$$F_D={\gamma }_W{W}_d/P_W$$

(3)

$$F_E={\gamma }_W{W}_e/P_W$$

(4)

where $F_W$ [hm²] denotes the WEF; N [person] quantifies the annual resident population at demographic closure; $f_W$ [hm²/person] represents per inhabitant WEF intensity; ${\gamma }_W$ corresponds to the global equivalence coefficient for water resources; W [m³] aggregates gross water withdrawals, inclusive of productive, domestic, and restorative consumption phases; $P_W$ [m³/hm²] defines normalized biospheric water yield potential, derived from multi-year hydrological flux averages at the planetary scale; $F_P$ [hm²/person] denotes production-oriented hydro-ecological footprint intensity; $F_D$ [hm²/person] corresponds to household water consumption footprint density; $F_E$ [hm²/person] quantifies eco-service water allocation indices; $W_P$ [m³] characterizes industrial-process water requirements; $W_D$ [m³] registers per capita domestic water utilization thresholds; and $W_e$ [m³] tracks environmental flow sustainability metrics. ${\gamma }_W$ = 5.19 and $P_W$ = 3140 m³/hm [2,12,13].

3.2. Water Carrying Capacity Model

The concept of water resources carrying capacity (WRCC) is operationally defined as the maximum sustainable freshwater exploitation magnitude attainable through optimized allocation strategies, constrained by both techno-industrial innovation baselines and institutionally embedded socioeconomic frameworks across defined temporal intervals. Methodologically, the WRCC calculation algorithm mandates the reservation of no less than 60% of total water resources for maintaining baseline ecosystem functions, a critical requirement for environmental preservation. The resultant value undergoes scaling through the application of a dimensionless 0.4 [27] multiplier. The governing WRCC formulation is expressed below:

$$C_W=N{c}_w=0.4\psi {\gamma }_{W}Q/P_W$$

(5)

where $C_W$ [hm²] denotes the water ecological carrying capacity (WECC); $c_w$ [hm²/person] represents per capita WECC, calculated as $C_W$ divided by resident population; The water yield coefficient (denoted as $\psi$) quantifies hydrological efficiency through comparative analysis, calculated as the quotient between region-specific mean water productivity and corresponding global benchmark values; and $Q$ [m³] aggregates all exploitable surface and groundwater within jurisdictional boundaries.

3.3. Metric Framework for Hydro-Resource Stewardship Assessment

The WECC quantifies the upper threshold capacity at which aquatic resources sustain socio economic advancement within designated administrative boundaries. Conversely, the WEF measures the aggregate volumetric expenditure of water allocations attributable to anthropogenic activities. The water resource equilibrium index (surplus/deficit) mathematically represents the arithmetic discrepancy between these two hydrological sustainability metrics. The water resource sustainability index $D_W$ represents the arithmetic discrepancy between regional hydrological ecological carrying capacity (ECC) and water footprint (WF) in equivalent measurement units. Non-negative h values ($D_W$ ≥ 0) indicate either ECC-WF equilibrium (at $D_W$ = 0) or resource abundance scenarios ($D_W$ > 0), whereas $D_W$ < 0 diagnostically identifies systemic overdraft conditions where demand surpasses regeneration capacity.

The Water Stress Index (WSI) serves as a critical indicator for assessing ecological pressure and water resource sustainability within the study region. The WSI is calculated as follows:

$$E{P}_W=F_W/C_W$$

(6)

where $E{P}_W$ is the water stress index. According to Ren et al. [28], when $E{P}_W$ < 0.5, regional water resources are safe; when 0.5 ≤ $E{P}_W$ < 0.8, regional water resources are safer; when 0.8 ≤ $E{P}_W$ ≤ 1, regional water resources are critical; when $E{P}_W$ > 1, regional water resources are unsafe.

The WEF per CNY 10,000 of GDP acts as a diagnostic metric to assess hydrologic productivity in economic output systems. Reduced WEF magnitudes correlate with lower water inputs required to yield equivalent monetary outputs, signifying enhanced hydrological efficiency in economic processes. Evaluating WEF/CNY 10,000 ratios enables quantification of resource utilization sustainability levels across spatial scales, where elevated values (>3.5 m³/yuan) imply overexploitation thresholds jeopardizing long-term water security. The computational framework is formalized through the subsequent equation:

$$F_{GDP}=F_W/GDP$$

(7)

where $F_{GDP}$ [hm²/CNY 10,000] is the WEF for CNY 10,000 of GDP; $GDP$ [yuan] is the gross domestic product.

3.4. LSTM Time-Series Neural Network Model

The LSTM networks constitute a modified recurrent neural network (RNN) architecture that resolves gradient dissipation/amplification issues characteristic of standard recurrent networks. This configuration incorporates gated memory units with self-optimized data retention mechanisms:

(1)

Memory Mechanisms: cell state preservation through time, self-regulating memory updates via gate operations.

(2)

Gating Architecture: input gate—controls new information flow; forget gate—manages memory retention; output gate—governs prediction outputs.

(3)

Temporal Processing Advantages: maintains stable gradient flow during backpropagation, captures multi-scale dependencies (short/long-term), and demonstrates particular efficacy for hydrological time series forecasting [29].

Figure 3 illustrates the computational topology of Long Short-Term Memory neural architectures, organized into three core structural modules: (a) input layer for feature ingestion, (b) hidden layer for temporal state propagation, and (c) output layer for prediction generation, in which $H_{t-2}$, $C_{t-2}$ are the output information, unit state value at the moment $t-2$, respectively; $X_{t-1}$, $H_{t-1}$, $C_{t-1}$ are the input information, output information, unit state value at the moment $t-1$, respectively; $X_t$, $H_t$, $C_t$, $f_t$, $i_t$, $O_t$ are the input information, output information, unit state value, forgetting gate, memory gate, and output gate at moment $t$, respectively; $H_{t+1}$, $C_{t+1}$ are the output information, unit state value at the moment $t+1$, respectively; σ denotes the Sigmoid function; and $\tilde{C}t$ is the new candidate value obtained by the tanh function. The LSTM’s hidden layer is architectured around cellular memory modules featuring tripartite control mechanisms: forgetting regulation, memorization modulation, and output filtering.

The LSTM model was implemented with the following optimized configuration:

(1)

Architecture specifications—network topology: single hidden layer LSTM; hidden units: 20 neurons (optimized for water resource time-series complexity); input/output dimensions: matched to hydrological feature dimensions.

(2)

Numerical parametrization framework—gradient descent operations employed the adaptive moment estimation algorithm (Adam optimizer, β₁ = 0.9, β₂ = 0.999 with coefficient-preserved initialization) with step size magnitude α = 0.01.

(3)

Learning rate scheduling protocol: the optimizer adopted a progressive scaling strategy wherein the initial learning rate received a multiplicative factor of 1.2 (Δ = +0.2) post-training phase transition at epoch 20, balancing convergence acceleration with regularization requirements; training duration was bounded at 100 epochs, incorporating real-time validation loss plateau detection (patience = 5 epochs) to prevent model over-specialization.

(4)

Regularization strategy—implicit learning rate reduction for loss plateau avoidance, batch normalization between layers, and gradient clipping (threshold = 1.0) for stability.

(5)

Hydrological data considerations—input window size: [X] temporal steps (aligns with water cycle periodicity); output horizon: [Y] steps (matching forecast requirements); feature scaling: standardized to N (0,1) using training set statistics [30].

LSTM networks are capable of handling long-term dependencies, capturing nonlinear relationships, and are robust to outliers. This method exhibits robust performance in processing temporal sequential datasets, attributed to its capability for modeling interdependencies across extended temporal intervals. Consequently, it has gained notable adoption across disciplines ranging from multilingual neural machine translation to sentiment analysis [31].

3.5. Evaluation of Projected Results

Selected experimental datasets were partitioned into distinct cohorts: 60% allocated to algorithmic parameter optimization through iterative training processes, with the residual 40% designated for post-calibration validation phases. Predictive performance quantification was executed via a multi-criteria assessment framework comprising relative error term (e), percentage-based deviation measure (MAPE), quadratic dispersion estimator (RMSE), along with stability indicators α and Ω, computed via established formulas incorporating three statistical metrics and two reliability indices.

$$e={\displaystyle \frac{\left|Ri-Oi\right|}{Oi}}$$

(8)

$$MAPE={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n\frac{\left|Ri-Oi\right|}{Oi}}$$

(9)

$$RMSE=\sqrt{{\displaystyle \frac{{\displaystyle \sum _{i=1}^n{(Ri-Oi)}^2}}{n}}}$$

(10)

$$\alpha =1-{\displaystyle \frac{{\displaystyle \sum _{i=1}^n\left|Ri-Oi\right|}}{{\displaystyle \sum _{i=1}^n(\left|Ri-O^{\prime }\right|+\left|Oi-O^{\prime }\right|)}}}$$

(11)

$$\Omega =1-{\displaystyle \frac{{\displaystyle \sum _{i=1}^n{(Ri-Oi)}^2}}{{\displaystyle \sum _{i=1}^n{(Oi-O^{\prime })}^2}}}$$

(12)

where $Oi$ denotes measured ground-truth values; $Ri$ represents experimentally obtained quantities; $n$ indicates cohort size; and $O^{\prime }$ expresses the arithmetic mean of observational data. Model precision exhibits an inverse relationship with e and RMSE values; superior accuracy correlates strongly with the adjacency of evaluation indices α and Ω to unit scaling thresholds [32].

3.6. Technical Route of the Study

The research technical route is divided into 4 steps as follows:

Step 1: Data sources. Complete data collection and organization and the value of the main parameters.

Step 2: Research methodology. Complete the water resources sustainable use evaluation method and WEF modeling method.

Step 3: Change analysis. Complete the water resources sustainable utilization calculation analysis and WEF calculation analysis.

Step 4: Predictive analysis. This phase encompasses conducting the sustainability assessment of water resource utilization and developing the LSTM-based predictive modeling. Figure 4 outlines the methodological workflow.

4. Variation Dynamics Characterization

4.1. Spatiotemporal Variation in WEF Nexus Coupling Mechanisms

This investigation examined the 2013–2022 per-capita water usage metrics (m³/capita·yr) across nine prefecture-level divisions in Guizhou, Southwest China. Sectoral water consumption dynamics were quantified through constructed WEF temporal profiles, specifically distinguishing productive (industrial/agricultural), residential, and ecological demand components. The analysis revealed that the decadal mean provincial WEF reached 4.61 hm²/capita, exceeding the 2022 national benchmark by 565% (0.6927 hm²/capita). Figure 5a reveals that the majority of cities and states in Guizhou Province exhibit no discernible trend in the change curve of per-capita WEF. The WEF dynamics exhibited a critical transitional phase in 2015. Zunyi demonstrated a notable basin-and-peak pattern during this period, evidenced by per-capita values of 0.56 → 0.28 → 0.58 hm² (2014–2016 baseline) alongside synchronized fluctuation records of 0.51 → 0.75 → 0.58 hm². This interannual volatility culminated in a province-wide low in 2020. The decade-long trajectory (2013–2022) ultimately reveals a descending WEF profile accompanied by persistent annual instability. This outcome demonstrates the efficacy of Guizhou’s recent regulatory frameworks through (a) mandatory enforcement of tiered water governance protocols, (b) systemic institutionalization of quantifiable boundary conditions for hydrological utilization, and (c) optimized stewardship of efficiency-driven allocation mechanisms. Figure 5b indicates that temporal variations in WEF per capita across Guizhou Province closely mirror the patterns observed in Figure 5a, reflecting a dominantly oscillatory decline. A distinctive temporal characteristic manifests as a subdued peak in 2020, while the decadal average WEF (2013–2022) remains at 0.38 hm²/capita. Elevated WEF metrics were documented in 2015 across Anshun, Qiandongnan, and Tongren municipalities, reaching a maximum intensity of 0.69 hm²/capita that denotes the most substantial per capita water resource appropriation throughout the study period. However, Qiandongnan Prefecture and Zunyi city exhibited lower values, with an ecological footprint of 0.21 hm²/capita, indicating the lowest per capita production of water resources. It is noteworthy that the disparity between the upper and lower figures is substantial. This result is inextricably linked to the increased efforts of Guizhou in recent years to strengthen the supervision of the whole process of water abstraction, strict statistical surveys of water use, and strict implementation of water use quotas and other measures. As quantified in Figure 5c, domestic water appropriation per capita throughout Guizhou Province spans 0.03–0.14 hm²/capita. Analysis of the ecological footprint trajectory reveals three distinct phases: (1) 2013–2019 stabilization: Minimal interannual variation characterizes this period (standard deviation = 0.004 hm²), with marginal annual increments averaging 0.007 hm². (2) Post-2019 escalation: Guiyang emerges as the primary growth driver, achieving 127% expansion within a single hydrological cycle (2019: 0.059 hm² → 2020: 0.134 hm²). (3) Regional variability: Zunyi followed with a more gradual rise (0.06 → 0.10 hm²) that remained below Guiyang’s escalation magnitude. Peripheral municipalities maintained growth trajectories aligned with provincial averages yet underperformed core urban centers. The sequential implementation of integrated urban–rural water distribution networks, expanded centralized supply infrastructure, and upgraded agricultural water security mechanisms has stimulated progressive 0.85% annual growth (2015–2020 baseline) in both supply reliability indices and household consumption metrics. As delineated in Figure 5d, the per capita ecological water consumption across Guizhou Province exhibits a characteristic fluctuation band of 0–0.03 hm²/individual. Notably, the temporal variation pattern demonstrates phase-specific behaviors: (1) a plateau phase (2013–2019) where values remain stationary, aligning with domestic water use trends; (2) a growth phase post-2019 marked by pronounced escalation in Qianxinan Prefecture (0.003 → 0.026 hm²/person during 2019–2021). Concurrently, Liupanshui City manifests an intermittent peak in 2021, followed by marginal reduction across provincial municipalities, suggesting post-surge stabilization mechanisms. Notwithstanding this short-term reduction, the post-decline rebound magnitude surpasses historical maxima. This recovery pattern systematically correlates with Guizhou’s established eco-management protocols, wherein ecological water allocation has been prioritized in regional governance frameworks since 2018, as evidenced by the following: (1) full implementation measures of the Mega-Ecology Strategic Initiative; (2) ecohydrological regulation enhancements targeting fluvial systems; (3) operationalization of ecological flow assessment protocols for lacustrine environments.

4.2. Changes in the Distribution of the WECC

The WECC of nine Guizhou prefectural administrative units was quantified from 2013 to 2022 through Equation (5) modeling, as visualized in Figure 6. A tiered classification framework stratifies these regions into three clusters. Guiyang, Liupanshui, and Anshun comprise Tier I, exhibiting clustered WECC values within 0.04–0.1 hm²/person throughout the study period. This cohort shows limited fluctuations—variation coefficients remain below 15%—in contrast to lower-tier units, as evidenced by the interannual stability patterns in Figure 6. It is evident that these three cities exhibit high levels of urbanization and comparatively limited water resources carrying capacity. The second layer includes Bijie city, Tongren city and Qianxinan city. Dynamic variations in municipal and regional water ecological footprints (2013–2022) are graphically depicted, demonstrating interannual oscillations within the 0.07–0.2 hm² per-capita bandwidth, which is indicative of variability in water resource utilization. The third layer includes Zunyi City, Qiandongnan Prefecture and Qiannan Prefecture. The WECC in these three cities and prefectures from 2013 to 2022 demonstrates a greater degree of fluctuation, ranging from 0.14 to 0.34 hm²/person. It is evident that these three municipalities and prefectures exhibit higher precipitation levels in comparison to other municipalities and demonstrate relatively elevated instability factors. The evolutionary trajectory of water sustainability metrics (2013–2022) for nine municipal jurisdictions is plotted in Figure 6a, with Figure 6b delineating the spatiotemporal heterogeneity in domestic water accessibility per capita across these regions.

Derived through the methodological framework outlined in Section 3.3,

Table 1. Zonal ecological balance status (surpluses/deficits) of water resources across the nine prefecture-level administrative units comprising Guizhou Province.

City (State)	Zonal Ecological Balance Status (Surpluses/Deficits) of Water Resources (hm2/Person)
	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Guiyang	0.024	0.054	0.042	0.022	0.047	0.041	0.040	0.050	0.037	0.024
Liupanshui	0.023	0.060	0.047	0.039	0.057	0.040	0.050	0.064	0.055	0.058
Zunyi	0.118	0.229	0.166	0.176	0.119	0.152	0.200	0.244	0.181	0.123
Anshun	0.018	0.071	0.069	0.045	0.050	0.064	0.065	0.087	0.053	0.057
Bijie	0.108	0.158	0.141	0.148	0.146	0.139	0.150	0.163	0.151	0.103
Tongren	0.104	0.171	0.122	0.173	0.157	0.102	0.149	0.183	0.170	0.103
Qianxinan	0.063	0.132	0.114	0.088	0.122	0.102	0.115	0.144	0.091	0.118
Qiandongnan	0.194	0.236	0.316	0.247	0.212	0.173	0.238	0.301	0.239	0.204
Qiannan	0.134	0.227	0.252	0.213	0.218	0.220	0.195	0.258	0.200	0.178

presents water resource ecological surplus/deficit metrics for Guizhou’s nine prefecture-level units. Synchronization analysis demonstrates an observable equilibrium state across the administrative regions during the 2013–2022 timeframe, where recurrent hydrological surplus intervals exhibit temporally concordant oscillation patterns.

4.3. Changes in the WSI

Figure 7 quantifies hydrological stress levels across nine administrative divisions (sub-provincial tier urban administrative areas and ethnic autonomous administrative regions) in Guizhou, with indices below the 0.5 threshold indicating sustainable conditions. From 2013 to 2022, the WSI of all nine cities and states in Guizhou is less than 0.5, with the exception of Guiyang city, which has an overall high WSI.

4.4. Changes in the WEF per CNY 10,000 of GDP

Figure 8 constructs the temporal variation pattern of water footprint intensity per CNY 10,000 of GDP in Guizhou using successive decadal datasets (2013–2022). Analysis reveals the water footprint intensity metric manifests consolidated attenuation characteristics, registering a 45.8% cumulative reduction over the study interval. While episodic increases occurred in discrete temporal intervals, these statistically insignificant fluctuations fail to reverse the prevailing diminishing tendency. These metrics collectively underscore Guizhou’s water use efficiency optimization trajectory during the evaluated period.

5. Spatiotemporal Coupling Assessment of Water–Energy–Food Nexus Footprints in Guizhou

5.1. Diagnostic Evaluation on Model Authenticity Metrics

WEF quantification for Guizhou’s nine administrative divisions was performed through Equations (1)–(6). These division-level datasets spanning 2013–2019 served as the LSTM temporal modeling inputs, while 2020–2022 entries were designated as prediction accuracy benchmarks. The temporal validation findings are compiled in

Table 2. LSTM model validation results.

City (State)		WEF (hm2/Person)			Evaluation of Results
		2020	2021	2022	MAPE	RMSE	$\mathit{\alpha }$	$\mathit{\Omega }$
Guiyang	Actual Value	0.0189	0.0204	0.0179	0.0287	0.0007	0.5785	0.5042
	Predicted Value	0.0187	0.0192	0.0182
Liupanshui	Actual Value	0.0116	0.0141	0.0124	0.0507	0.0007	0.6780	0.5675
	Predicted Value	0.0121	0.0145	0.0134
Zunyi	Actual Value	0.035	0.0394	0.0361	0.0250	0.0011	0.6441	0.6662
	Predicted Value	0.0361	0.0379	0.0359
Anshun	Actual Value	0.0115	0.0137	0.0114	0.0505	0.0006	0.6538	0.6391
	Predicted Value	0.012	0.0141	0.0123
Bijie	Actual Value	0.017	0.02	0.0182	0.0300	0.0006	0.7901	0.7259
	Predicted Value	0.0166	0.021	0.0179
Tongren	Actual Value	0.0116	0.0143	0.0149	0.0492	0.0007	0.7727	0.7832
	Predicted Value	0.0122	0.015	0.0156
Qianxinnan	Actual Value	0.0102	0.0119	0.012	0.0652	0.0009	0.5106	−0.2948
	Predicted Value	0.01	0.0104	0.0114
Qiandongnan	Actual Value	0.0171	0.0204	0.021	0.0233	0.0005	0.8727	0.9116
	Predicted Value	0.0169	0.0211	0.0215
Qiannan	Actual Value	0.0161	0.018	0.0165	0.0154	0.0003	0.7857	0.8505
	Predicted Value	0.0163	0.0175	0.0166

. Table 2 validates that the LSTM architecture achieves close congruence between simulated and measured values throughout the validation phase. Computational robustness metrics across all nine administrative units—specifically MAPE < 0.0652, RMSE < 0.0011, $\alpha$ > 0.5106 (training phase), and $\Omega$ > 0.5042 (validation phase)—exhibit compliance with precision thresholds for hydrological forecasting. Notably, the architecture attained 100% prediction validity when applying the reference criterion |ε| < 10% [33]. The validation outcomes therefore support the application of this LSTM-based temporal modeling framework for prognostic analyses of water–energy–food nexus dynamics in Guizhou’s regional systems.

5.2. Analysis of the Projected Results

The WEF nexus trajectories of nine prefecture-level units in Guizhou Province between 2023 and 2027 were simulated using LSTM architecture. As demonstrated in Figure 9a, building upon empirical evidence from the 2013–2022 observation period, projected WEF indices exhibit stabilized variation patterns during the initial triennium of prediction. Zunyi maintains dominance in WEF magnitude accompanied by accelerated expansion pace, whereas Guiyang and Bijie display more moderate progression rates. This spatial heterogeneity correlates with the context of Guizhou’s resource-efficient economic transition policy framework. Notably, Zunyi exhibits the most significant surge in water utilization—a pattern aligning with the production expansion trajectory observed in Zunyi’s dominant economic sector (liquor manufacturing sector, experiencing sustained growth). This correlation necessitates stringent enforcement of water extraction licensing protocols coupled with dynamic resource allocation mechanisms targeting the liquor production value chain. Figure 9b demonstrates that the production water use ecological footprints (EFP-WU) across all nine administrative units in Guizhou exhibit comparable magnitudes with minimal temporal fluctuations (±5%). This equilibrium suggests coordinated regional development during Guizhou’s socioeconomic transition phase, where industrial expansion (GDP growth rate > 7% from 2015 to 2020) is achieved while maintaining a stable water resource demand baseline. The observed trend further implies enhanced water productivity arising from technological and managerial optimizations. Figure 9c,d reveal sustained increments in household and environmental water appropriation across all nine Guizhou prefectures. This spatial–temporal pattern aligns with (1) enhanced living standards (evidenced by 68% urbanization rate progression since 2015), and (2) intensified ecological governance prioritization given Guizhou’s role as China’s first provincial-level pilot zone for eco-civilization institutional innovation.

Guizhou’s water resource systems currently exhibit hydrological equilibrium condition, necessitating demand-responsive strategic interventions to enhance long-term stewardship practices: (1) For municipalities exhibiting significant expansion in per capita domestic water-related ecological footprints (e.g., Guiyang), prioritized interventions should integrate the following water stewardship strategies: (a) optimize urban spatial configurations through land use zoning to enhance infrastructure hydrological performance; (b) develop hybrid green–gray infrastructure networks combining constructed wetlands, permeable pavements, and bioswales for stormwater retention capacity maximization; (c) institutionalize rainwater harvesting mandates in municipal building codes to achieve sustainable urban water cycle closure, thereby establishing water conservation municipalities. (2) Enhance water resource management through hydraulic infrastructure optimization. This involves the following: (a) prioritizing spatial rationalization of rural water systems to establish interconnected distribution grids (ring-shaped topology preferred for reliability); (b) scaling regionalized supply schemes in alignment with integrated watershed development protocols; (c) deploying automated source-to-tap monitoring frameworks covering abstraction, treatment, and household delivery phases. Such coordinated measures facilitate techno-economic parity between decentralized and centralized water provision systems. (3) Regions exhibiting marked elevation in per-capita hydro-ecological footprints (e.g., Qianxinan Prefecture) necessitate prioritized surveillance protocols and adaptive governance frameworks for aquatic resource allocation. This strategic emphasis will catalyze capacity building aligned with ecological civilization imperatives during Guizhou’s transitional modernization phase. Critical interventions include the following: (a) enforcing tiered water withdrawal permits through IoT-enabled metering infrastructure; (b) implementing predictive analytics platforms for real-time consumption pattern tracking; (c) developing blockchain-based water rights trading mechanisms to optimize allocation efficiency; (d) deploying remotely sensed watershed health assessment systems to guide restoration prioritization. Synthesizing these measures establishes a closed-loop management architecture that operationalizes circular hydro-economy principles.

6. Conclusions

This WEF analysis and forecasting study provides critical insights for the assessment of regional economic development and the sustainable utilization of water resources, thus offering a scientific basis for the formulation of rational water resource management policies. As economic growth progresses and water use structures become more optimized—supported by the implementation of comprehensive water conservation measures—the pressure on sustainable water resource utilization in Guizhou Province’s nine cities and prefectures shows a gradual decline over time. This trend is further corroborated by the decreasing WEF per CNY 10,000 of GDP, indicating improved water use efficiency and resource sustainability.

Following 2019, a considerable divergence in trends pertaining to the WEF nexus emerged across all prefectures in Guizhou Province, encompassing production, domestic, and ecological water usage. It is worthy of particular note that there has been a steady linear increase in domestic water demand in Guiyang, while Qianxinan Prefecture has exhibited a similar upward trajectory in ecological water requirements. Despite these localized variations, the aggregate impact on the province’s total WEF remained statistically insignificant, suggesting effective counterbalancing through regional water resource management strategies.

During the 2013–2022 decade, Guizhou’s provincial economy demonstrated sustained growth patterns wherein per-capita water–energy–food (WEF) indicators maintained homogeneity among its administrative regions, punctuated solely by negligible oscillatory deviations. However, a notable shift occurred after 2019, marked by significant changes in both domestic and ecological water consumption patterns. The observed systemic shifts primarily derive from macroeconomic framework realignments, sectoral productivity rationalization, and codified hydrological governance enforcement. The observed trends suggest that policy interventions and advancements in science and technology have played a pivotal role in reshaping the WEF nexus in Guizhou Province. Specifically, the post-2019 divergence in water usage highlights the province’s transition toward more sustainable resource management practices, likely driven by national ecological initiatives and regional efforts to balance economic development with environmental preservation.

Hydrogeological projections confirm Guizhou’s capacity to preserve balanced resilience states in aquatic resources through 2026, exhibiting progressive growth in production-oriented water footprint indicators alongside synchronous expansion patterns in domestic water consumption and environmental flow allocations. To operationalize precision water governance frameworks, strategic allocation protocols must prioritize municipalities demonstrating disproportionate per-capita domestic water impacts—exemplified by Guiyang’s demographically dense urban core—through tiered demand-side management interventions. Concurrently, the allocation of water resources should be continuously strengthened, the level of water supply security should be improved, and the management of water resources should be reinforced. It is imperative to direct attention towards the WEF in Guizhou Province, thereby ensuring the sustainable development of this region.

This research implements a custom-developed LSTM architecture for joint analytical forecasting of hydro-ecological footprint metrics across nine municipal divisions comprising Guizhou’s urban water governance network. Subsequent steps may include comparison and validation with other single models (e.g., ARIMA or process-based hydrological models), combination of two or more models, or optimization of the model. This will facilitate further in-depth research.