Multi-Scenario Analysis of Brackish Water Irrigation Efficiency Based on the SBM Model

Abstract

The North China Plain faces severe water scarcity, and the efficient use of brackish water has become a crucial pathway for sustaining agricultural development. In this study, we combine scenario analysis with Data Envelopment Analysis to establish a multi-scenario efficiency evaluation framework. Focusing on six counties in Handan, Hebei Province, we employ an input-oriented Slack-Based Measure Data Envelopment Analysis (SBM-DEA) model to systematically evaluate brackish water irrigation efficiency (BWIE) across a baseline year (2020) and eight projected scenarios for 2030. The results show that the mean efficiency values across scenarios range from 0.646 to 0.909. Scenarios combining universal adoption of water-saving irrigation with normal hydrological conditions achieve the highest mean efficiency (>0.9), with minimal regional disparities and optimal system stability. The promotion of water-saving irrigation technologies is the primary driver of improved BWIE, whereas simply increasing brackish water application yields only limited marginal benefits. Redundancy analysis further indicates that water resource inputs are the main source of efficiency loss, with brackish water redundancy (42.3%) far exceeding that of land inputs (10.5%). These findings provide quantitative evidence and methodological support for optimizing regional water allocation and advancing sustainable agricultural development.

1. Introduction

With the continuous increase in water demand from agriculture, industry, and domestic use, pressure on water resources has intensified, and water scarcity has become a major challenge faced worldwide [1,2]. While China possesses one of the world’s largest total freshwater endowments, limited per capita availability and pronounced spatiotemporal disparities have sharpened the conflict between water supply and socio-economic growth. The North China Plain, a critical hub of China’s grain production, has long been constrained by the twin challenges of acute water scarcity and chronic groundwater overexploitation [3,4,5]. Within this context, harnessing unconventional water resources has emerged as an imperative to safeguard the sustainability of regional agriculture [6,7]. Brackish water is an unconventional water resource with abundant reserves and wide distribution, which can serve as an important substitute and supplement to freshwater in agricultural irrigation [8,9,10,11,12,13,14]. In the North China Plain, shallow brackish groundwater is extensively distributed and constitutes a substantial share of regional aquifers, with estimates in Hebei Province alone indicating more than 3.5 × 10⁹ m³ of exploitable resources each year [15,16]. Moreover, this resource is characterized by ease of extraction and considerable volume, holding significant potential to alleviate agricultural water scarcity in the region. Nevertheless, the use of brackish water for irrigation poses the risk of salt buildup in soils. When salinity remains below the crop tolerance threshold, its increase has little effect on yield; once this threshold is exceeded, crop yields decline approximately linearly with rising soil salinity [17,18,19]. A variety of crops exhibit different levels of salt tolerance. Common salt-tolerant crops include cotton, barley, sorghum, sugar beet, and certain maize and wheat varieties [20,21]. The major crops in the North China Plain are winter wheat, summer maize, and cotton, which generally exhibit certain salt tolerance and can be irrigated with brackish water [22,23]. Among them, cotton is the most tolerant, surviving in soils with electrical conductivity (EC) up to ~7–10 dS/m, while winter wheat and summer maize can tolerate moderate salinity, with EC thresholds of ~4–6 dS/m and ~3–4 dS/m, respectively [24]. Beyond crop tolerance, brackish water irrigation introduces practical challenges. Salts can rapidly corrode drip and sprinkler systems, necessitating resistant and often costly materials [25]. Effective management, including monitoring salinity, periodic leaching, and using robust irrigation infrastructure, is essential. Prior studies indicate that moderate substitution of freshwater with brackish water can relieve water demand while maintaining yield stability [26,27,28,29,30,31,32], but successful implementation depends on both crop tolerance and real-world technical and economic constraints.

The Data Envelopment Analysis (DEA) method, owing to its advantages in evaluating the efficiency of multi-input and multi-output systems, has been widely applied in water resources management and agricultural production [33,34,35,36]. For example, Wang et al. employed a DEA model to evaluate the improvement of regional agricultural production efficiency brought by China’s South-to-North Water Diversion Project, highlighting the importance of cross-regional water transfer in alleviating agricultural water constraints [37]. Ait Sidhoum et al. combined DEA with an eco-efficiency framework to examine how water technology adoption under drought conditions shapes the sustainability of agricultural systems [38]. Lu et al. used a super-efficiency DEA model to estimate the agglomeration level of grain production and agricultural environmental efficiency across different functional zones in China [39]. Wei et al. applied a super-efficiency SBM-DEA model to measure agricultural water use efficiency in the Yellow River Basin [40]. Compared with traditional radial DEA models, the Slack-Based Measure (SBM) model effectively characterizes input redundancies and output shortfalls, making it more suitable for systems such as agricultural irrigation, which feature complex input structures and are strongly affected by external conditions [41]. It is noteworthy that, in recent years, some scholars have integrated SBM-DEA with scenario-based methods. For instance, Guo et al. employed an SBM-DEA model integrated with life cycle analysis to build a scenario-based evaluation framework, systematically assessing the efficiency of 16 energy-saving technology combinations in the aluminum industry across various policy scenarios [42]. These methodological advances suggest that combining scenario analysis with the SBM-DEA model provides a new pathway for evaluating brackish water irrigation efficiency (BWIE) under varying hydrological and management conditions.

In the fields of agriculture and water resource utilization, considerable research has explored irrigation efficiency and brackish water use, but most studies focus on crop yield effects, with systematic evaluations of BWIE still lacking. In particular, quantitative assessments under future scenario settings remain scarce. To fill this gap, this study takes typical counties in Handan City as research cases, constructs future scenarios, and systematically evaluates BWIE under different combinations of water-saving irrigation ratios, hydrological year types, and brackish water utilization levels. The objectives of this study are: (1) Employ an input-oriented SBM-DEA model to evaluate BWIE under baseline and multiple future scenarios, revealing overall efficiency levels and their variation patterns; (2) Compare efficiency differences among counties and identify their efficiency types under the integration of multiple scenarios; (3) Use redundancy analysis to quantitatively detect inefficient components of resource inputs and propose pathways for optimizing brackish water utilization. The findings will help guide the scientific use of brackish water in agricultural production and provide a scientific basis for optimizing regional water resource allocation and promoting sustainable agricultural development.

2. Materials and Methods

This study develops an integrated research framework that combines scenario analysis with Data Envelopment Analysis (DEA) to evaluate BWIE under multiple conditions (Figure 1). The framework establishes key input and output indicators, incorporates baseline and future scenarios reflecting variations in water-saving irrigation adoption, hydrological conditions, and brackish water utilization, and applies an input-oriented SBM-DEA model to assess efficiency and redundancy. By integrating efficiency evaluation with redundancy analysis, the framework provides a systematic basis for revealing spatial disparities, comparing scenario outcomes, and deriving targeted policy recommendations.

2.1. Study Area

Handan City is located in the southern part of Hebei Province, between 113°27′–115°38′ E and 36°04′–37°02′ N. It borders Liaocheng City in Shandong Province to the east, Anyang City in Henan Province to the south, the Taihang Mountains to the west, and Xingtai City to the north. The total area is approximately 12,000 km², with terrain sloping from west to east, consisting of the foothills of the Taihang Mountains in the west and an alluvial plain in the east. The region has a warm temperate, semi-humid continental monsoon climate with four distinct seasons. The average annual precipitation is approximately 500–600 mm, with rainfall concentrated between June and August, characterized by uneven spatial and temporal distribution and significant interannual variability. The dominant soil types are fluvo-aquic and cinnamon soils. This study focuses on six counties in the eastern plain of Handan City—Qiuxian, Guantao, Quzhou, Guangping, Cheng’an, and Feixiang (Figure 2).

2.2. SBM-DEA Model

Data Envelopment Analysis (DEA) is a non-parametric method based on linear programming, widely used to assess the relative efficiency of decision-making units (DMUs) with similar functions under multiple input-output scenarios. It was first introduced by Charnes, Cooper, and Rhodes in 1978, known as the CCR model [33]. The method does not require pre-set weights or rely on the functional relationship between inputs and outputs, thereby maintaining a high level of objectivity in efficiency measurement. To quantitatively assess the efficiency of brackish water use in agricultural irrigation, this study adopts the input-oriented SBM (Slack-Based Measure) model [41]. The model can identify input redundancies while maintaining existing output levels, overcoming the limitations of traditional radial DEA models that may overlook non-proportional changes. It is suitable for agricultural irrigation systems, which involve complex input structures and strong technological heterogeneity.

The input-oriented SBM model is represented as follows:

$$\min \mathsf{\rho }=1-{\displaystyle \frac{1}{m}}\sum _{i=1}^m{\displaystyle \frac{s_i^{-}}{x_{io}}}$$

(1)

The constraints are as follows:

$$\mathrm{X}\mathsf{\lambda }+s^{-}=x_o$$

(2)

$$\mathsf{\lambda }\ge 0,s^{-}\ge 0$$

(3)

where $\mathsf{\rho }$ represents the efficiency value calculated by the model, where $\mathsf{\rho }\in \left[0, 1\right]$; $n$ denotes the number of Decision-Making Units (DMUs), while $\mathrm{m}$ and $\mathrm{q}$ represent the number of input and output indicators, respectively; $\mathrm{X}\in R^{m\times n}$ and $\mathrm{Y}\in R^{q\times n}$ are the input and output matrices, respectively; $x_o$ and $y_o$ are the input and output vectors of the $\mathrm{o}$-th DMU to be evaluated; $\mathsf{\lambda }$ represents the weight vector; $s^{-}$ represents input redundancy. When $\mathsf{\rho }=1$, the DMU is considered efficient. When $\mathsf{\rho }=1$ and $s^{-}=0$, the DMU is classified as a DEA strongly efficient unit.

2.3. Determination of Input and Output Indicators

To evaluate the irrigation efficiency of the research subjects using the constructed SBM-DEA model, it is first necessary to determine the selection range and number of Decision-Making Units (DMUs). According to the basic principles of the DEA method, to avoid a large number of DMUs being evaluated as efficient units, the number of DMUs should be no less than five times the total number of input and output indicators [33]. This study constructed an efficiency evaluation system with three input indicators and one output indicator, so the minimum required number of DMUs is 20. Based on the actual situation of brackish water use in Handan City’s districts, six representative counties were selected as evaluation objects. A total of 54 DMUs (S0–S8) were constructed through combinations of year, water-saving irrigation ratio, hydrological year type, and brackish water utilization level. This not only satisfies the DEA method’s sample size requirement but also reflects regional and scenario variations. Specific scenario settings will be detailed in the following sections.

The indicator system in this study follows the “resources-process-output” logical framework and references related studies on agricultural irrigation efficiency [43,44,45,46]. On the input side, the selected indicators include salt-tolerant crop planting area, freshwater irrigation volume, and brackish water irrigation volume, reflecting land use scale, conventional water input, and unconventional water input, respectively. Salt-tolerant crops include three major crops: cotton, winter wheat, and summer maize. These crops have wide planting areas, strong adaptability, and high salt tolerance in the study area, making them typical candidates for brackish water irrigation. To avoid introducing redundant variables in the DEA model and maintain the simplicity of the indicator system, the area and yield data for these three crops are not listed separately in the calculations, but are integrated into a composite indicator for salt-tolerant crops. On the output side, the total yield of salt-tolerant crops is selected as the agricultural output indicator, corresponding to the three input crops, and is used to measure the overall effect of different water resource utilization methods on crop production (

Table 1. Main Input and Output Indicators for BWIE Evaluation.

Indicator Type	Indicator	Unit
Input Indicators	Salt-tolerant crop planting area	mu
	Freshwater irrigation volume	104 m3
	Brackish water irrigation volume	104 m3
Output Indicators	Total yield of salt-tolerant crops	ton

). Detailed data sources are listed in

Table 2. Data sources.

Indicator	Source	Note
Brackish water irrigation volume	Handan City Water Resources Bulletin (2020)	County-level irrigation water data
Freshwater irrigation volume	Derived from irrigation quotas and crop planting areas	Estimated for study area
Irrigation water quotas (wheat, maize, cotton)	Irrigation Water Quotas for Wheat, Maize, and Cotton	Official standards
Salt-tolerant crop yields (wheat, maize, cotton)	Handan Statistical Yearbook (2015–2020)	County-level agricultural yield data
Yield improvement coefficients under water-saving irrigation (${\mathsf{\lambda }}_i$)	Published literature	15% (maize), 8% (wheat), 13% (cotton) [47,48,49]
Total yield of salt-tolerant crops	Handan Statistical Yearbook (2015–2020)	Agricultural land use statistics

.

2.4. Scenario Design and Data Construction Methods

2.4.1. Scenario Design Scheme

To systematically assess the impact and optimization potential of different resource configurations and technological conditions on BWIE, this study uses the 2020 statistical data as a baseline, constructing one baseline scenario (S0) and eight future simulated scenarios (S1–S8), resulting in a total of nine scenario combinations (

Table 3. Different Irrigation Scenario Numbers and Assigned Variables.

Scenario	Year	Proportion of Water-Saving Irrigation	Hydrological Year Type	Brackish Water Utilization
S0	2020	Current level	Actual hydrological conditions	Statistical values of the year
S1	2030	90%	Dry year	Maintain baseline level
S2	2030	90%	Normal year	Maintain baseline level
S3	2030	90%	Dry year	Increase by 15%
S4	2030	90%	Normal year	Increase by 15%
S5	2030	100%	Dry year	Maintain baseline level
S6	2030	100%	Normal year	Maintain baseline level
S7	2030	100%	Dry year	Increase by 15%
S8	2030	100%	Normal year	Increase by 15%

Maintaining the baseline level and increasing by 15% both refer to the change in brackish water utilization relative to the baseline year.

). The scenario design covers the following three key variables:

(1)

Water-saving irrigation ratio: Based on regional agricultural water resource planning and policy goals, the scenarios set the water-saving irrigation ratio for 2030 at 90% and 100%, while S0 retains the actual ratio for 2020.

(2)

Hydrological year type: To reflect the impact of meteorological and hydrological changes on irrigation demand, two types of years are defined: normal year (average multi-year precipitation conditions) and dry year (significant reduction in precipitation and increased irrigation water demand).

(3)

Brackish water utilization level: Set at the same level as the baseline year, with an additional 15% increase for one of the scenarios, which simulates the potential substitution capacity under enhanced salt-tolerant crop planting and water quality management.

The baseline scenario (S0) reflects the actual production conditions of 2020 and serves as a reference for comparison with the future scenarios. The future scenarios (S1–S8) correspond to different combinations of brackish water utilization levels, water-saving irrigation ratios, and hydrological year types, reflecting various possible resource and technological configurations in the study area. This scenario framework provides the data and methodological foundation for the subsequent efficiency calculations, identification of efficient scenarios, and optimization of regional irrigation strategies based on the DEA model.

2.4.2. Indicator Construction Under Scenarios

To ensure transparency and reproducibility, the specific sources of the raw data for each input and output indicator are summarized in Table 2. Based on these data, this study constructs an input dataset that includes the baseline year (2020) and future scenarios (2030). The input indicators for the model are salt-tolerant crop planting area, brackish water irrigation volume, and freshwater irrigation volume, while the output indicator is the total yield of salt-tolerant crops. Each indicator is calculated at the county scale based on different crops. The specific calculation steps are as follows.

(1)

Salt-tolerant crop area prediction

Based on the 2020 measured area and referencing crop planting area variation data (Table 2), the annual growth rate $r_i$ for the three salt-tolerant crops is calculated. The trend extrapolation method is applied to forecast the total salt-tolerant crop area for 2030:

$$A_{i,2030}=A_{i,2020}\times {\left(1+r_i\right)}^t$$

(4)

$$A_{\mathrm{total}}=\sum _{i=1}^3{A}_{i,2030}$$

(5)

where $A_{i,2030}$ represents the predicted area of crop $\mathrm{i}$ in 2030; $A_{i,2020}$ represents the area of crop $\mathrm{i}$ in the baseline year; $r_i$ represents the annual growth rate of crop $\mathrm{i}$ from 2015 to 2020; $\mathrm{t}$ represents the forecast period in years, with $\mathrm{t}=10$; $A_{\mathrm{total}}$ represents the total planting area of salt-tolerant crops. Based on this, the total area is further divided into water-saving irrigation areas at 90% or 100%, according to policy goals, to construct different scenario combinations.

(2)

Brackish water irrigation volume

The brackish water irrigation volume was set according to baseline values and adjusted ratios (see Table 2). In this study, different levels of brackish water utilization were set in the scenario construction: maintaining the 2020 measured value and increasing by 15% over the baseline level, to simulate future utilization patterns under two conditions: maintaining the current situation and moderate expansion.

$$W_{\mathrm{saline},2030}=\left\{\begin{array}{c}{W}_{\mathrm{saline},2020},Baseline scenario\\ W_{\mathrm{saline},2020}\times \left(1+0.5\right),Increased scenario\end{array}\right.$$

(6)

where $W_{\mathit{saline},2020}$ denotes the brackish water irrigation volume in 2020, and $W_{\mathit{saline},2030}$ represents the brackish water irrigation volume under different scenarios in 2030.

(3)

Freshwater irrigation volume

Freshwater irrigation volume was calculated based on irrigation water quotas and projected planting areas of salt-tolerant crops (Table 2). The total water demand was first estimated and the conventional freshwater input was then determined by deducting the assigned brackish water irrigation volume, as expressed below:

$$W_{\mathrm{total}}=\sum _i{A}_i\times q_{i,m}$$

(7)

$$W_{\mathrm{fresh}}=W_{\mathrm{total}}-W_{\mathrm{saline}}$$

(8)

where $W_{\mathrm{total}}$ is the total water demand, $W_{\mathrm{fresh}}$ the freshwater irrigation volume, $A_i$ the planting area of crop $i$, $q_{i,m}$ the unit water quota of crop $i$ under irrigation method $m$, and $W_{\mathrm{saline}}$ the brackish water irrigation volume.

(4)

Total yield of salt-tolerant crops

The yield of salt-tolerant crops is calculated based on the projected planting areas in 2030 and the unit yields under different irrigation methods. The total area is divided into conventional and water-saving irrigation categories, with yields computed separately and then aggregated. The calculation is expressed as follows:

$$y_{i,ws}=y_{i,trad}\times \left(1+{\mathsf{\lambda }}_i\right)$$

(9)

$$Y_{i,2030}=\left(1-p_i\right)\times A_{i,2030}\times y_{i,trad}+p_i\times A_{i,2030}\times y_{i,ws}$$

(10)

$$Y_{\mathrm{total},2030}=\sum _{i=1}^3{Y}_{i,2030}$$

(11)

where $A_{i,2030}$ is the projected planting area of crop $i$ in 2030, $y_{i,ws}$ and $y_{i,trad}$ are the unit yields of crop $i$ under water-saving and conventional irrigation, respectively, with yield parameters obtained from statistical yearbooks and relevant literature (Table 2); ${\mathsf{\lambda }}_i$ represents the relative yield gain of water-saving irrigation over conventional irrigation: 15% for maize, 8% for wheat, and 13% for cotton [47,48,49]; $p_i$ is the proportion of crop $i$ cultivated under water-saving irrigation; $Y_{i,2030}$ is the projected total yield of crop $i$ in 2030 (t); $Y_{\mathrm{total},2030}$ represents the aggregate yield of the three salt-tolerant crops in the study area.

All indicators were calculated at the county level, and an input dataset covering all counties and scenario combinations was ultimately constructed to support the DEA efficiency analysis. The specific data sources of input and output indicators are summarized in Table 2, while descriptive statistics are presented in

Table 4. Descriptive Statistics of Indicators.

Indicator	Sample Size	Mean	Standard Deviation	Minimum	Maximum
Salt-tolerant crop planting area	54	354,050	54,087	257,080	475,500
Freshwater irrigation volume	54	3409	756	2160	5203
Brackish water irrigation volume	54	414	201	90	805
Total yield of salt-tolerant crops	54	305,929	62,584	190,062	416,802

.

2.5. Evaluation of Brackish Water Input Redundancy

In DEA theory, input slack variables quantify the potential reduction in inputs without compromising outputs, serving as indicators of inefficiencies in resource allocation. Based on the results of the SBM model, this study introduces the ratio of input slack to actual input as the redundancy rate of brackish water input, which is used to identify the water-saving potential of each county under different scenarios. This metric quantifies regional redundancy under given conditions, offering insights to support efficient brackish water use and optimized resource allocation. The redundancy rate is calculated as follows:

$$R_{jk}={\displaystyle \frac{s_{jk}^{-}}{x_{jk}}}\times 100\%$$

(12)

where $R_{jk}$ represents the redundancy rate of brackish water input for county $j$ under scenario $k$, $s_{ik}^{-}$ is the slack value derived from the SBM model, while $x_{ik}$ indicates the corresponding actual input. A redundancy rate of zero signifies that the input has no further improvement potential, indicating an optimal allocation level.

3. Results and Discussion

3.1. Overall Efficiency Analysis Results

3.1.1. Efficiency Levels and Stability Across Scenarios

While previous studies have predominantly examined the agronomic impacts of brackish water irrigation [8,9,10,11,12,13,14], this study shifts the focus to efficiency in resource allocation. By employing a multi-scenario DEA framework, we provide a systematic evaluation of BWIE under various future conditions, moving beyond static assessments to offer a dynamic perspective on optimizing water inputs.

The average efficiency and standard deviation for each scenario are summarized to evaluate the overall efficiency level and system stability (

Table 5. BWIE for Each Scenario.

Scenario	Rank	Mean Efficiency	Standard Deviation
S0	7	0.696	0.209
S1	8	0.663	0.142
S2	5	0.726	0.136
S3	9	0.646	0.133
S4	6	0.711	0.13
S5	3	0.782	0.144
S6	2	0.903	0.117
S7	4	0.756	0.121
S8	1	0.909	0.122

). Results show clear differences in BWIE across scenarios, ranging from 0.646 to 0.909. The highest value occurs in S8 (0.909), followed by S6 (0.903), both of which are close to the efficiency frontier, reflecting more efficient resource allocation in these scenarios. The lowest BWIE is observed in S3 (0.646), which is lower than the baseline scenario S0 (0.696), indicating that efficiency levels may decrease rather than improve in certain future settings. S8 and S6 perform the best, followed by S5 and S7, while S1 and S3 are at the lowest levels, demonstrating that the differences in BWIE between future scenarios remain significant.

The standard deviation further reveals differences in system stability. The highest standard deviation is observed in S0 (0.209), reflecting substantial differences in BWIE between counties under current conditions, indicating weaker system robustness. In contrast, the lowest standard deviations are observed in S6 and S8 (0.117 and 0.122), indicating that under efficient scenarios, not only is the overall efficiency higher, but the differences between counties are smaller, showing better coordination and stability.

Considering both efficiency and stability, S8 (comprehensive water-saving + normal year + increased brackish water utilization) performs the best in terms of both efficiency and stability and can be considered the ideal scenario. This result suggests that a reasonable combination of policy and hydrological conditions can significantly improve the overall level of regional BWIE, reduce county disparities while enhancing system stability, and provide a potential optimal path for future brackish water utilization.

3.1.2. Distribution Characteristics and Proportion of High-Efficiency Units

To further analyze the impact of different scenarios on the distribution of BWIE, this study calculates the number of counties in each scenario that achieve DEA efficiency (efficiency value = 1) and classifies them into three levels: high efficiency (=1), medium efficiency (0.6–1), and low efficiency (<0.6). The study finds that different policy combinations have a significant impact on the distribution of BWIE levels. The specific distribution is shown in Figure 3.

3.1.3. Regularity of Efficiency Response Among Different Scenarios

Based on the aforementioned efficiency averages and distribution patterns, this subsection employs the control variable method to compare specific scenario pairs (e.g., S2↔S6, S5↔S6, S6↔S8), to identify the independent effects and interactive influences of three variables: water-saving irrigation ratio, brackish water irrigation, and hydrological year type, and further reveal the underlying mechanisms of BWIE (Figure 4).

The water-saving irrigation ratio plays a pivotal role in enhancing BWIE. Under the same hydrological conditions and brackish water irrigation levels, transitioning from “high water-saving (90%)” to “comprehensive water-saving (100%)” can result in a significant increase in the average efficiency from 0.726 in S2 to 0.903 in S6, with an increase of 0.177, showing a clear threshold effect and nonlinear structural response characteristics. This indicates that the widespread adoption of water-saving irrigation technology can significantly optimize the resource input structure and improve BWIE.

Compared to directly changing the input-output structure, the hydrological year type, as an exogenous natural variable, manifests in the DEA model as an environmental condition that affects resource availability and irrigation intensity. In the “high water-saving” scenario, switching from a dry year (S1) to a normal year (S2) results in a 0.063 efficiency improvement, whereas under “comprehensive water-saving” conditions (S5→S6), this increase expands to 0.121, indicating an asymmetric interaction effect between hydrological conditions and water-saving policies. When the water-saving foundation is weak, improvements in hydrological conditions are insufficient to drive structural improvements in BWIE; whereas under the premise of an optimized irrigation system, favorable hydrological conditions significantly enhance the matching of resource use, bringing the DMU closer to the efficiency frontier.

Increasing brackish water irrigation volume shows a weak marginal effect and is constrained by the hydrological background. Under the ideal conditions of a normal year and comprehensive water-saving, a 15% increase in brackish water usage (S6→S8) only leads to a 0.006 increase in efficiency; whereas under the same water-saving conditions in a dry year (S5→S7), increasing brackish water usage results in a decrease of 0.026 in irrigation efficiency. This suggests that increasing brackish water irrigation volume should not be advanced as the main measure independently; its effectiveness is highly dependent on the water-saving foundation and hydrological environment, and under strong resource constraints, it may even have a negative impact. Brackish water is better suited as a supplementary resource based on systematic optimization.

Although DEA has been used in agricultural water studies [38,39,40,41,45,46,47], existing applications are often static. A key innovation of our study is the integration of hydrological variability and management policy into a scenario-based SBM-DEA framework. This approach allows us to uncover nonlinear threshold effects and asymmetric interactions between water-saving policies and hydrological conditions. These dynamic insights, unique to our multi-scenario analysis, could not be captured by previous static models.

3.2. Regional Differences Analysis

Building upon the overall analysis, this study further examines the differences in BWIE across the six counties under different scenarios. To this end, the efficiency mean (μ), standard deviation (σ), and range (R) were chosen as the discriminative indicators for each county across the nine scenarios. Based on these indicators, the six counties were categorized into three typical types: stable and efficient (mean remains high with low volatility), improvement potential (baseline level is low but can reach the efficiency frontier under ideal scenarios), and low-efficiency and fragile (overall efficiency is low or significantly declines across scenarios) (

Table 6. Characteristics and Classification Results of BWIE in Each County.

County	Mean	Standard Deviation	Range	S8–S0 Difference	DEA Efficient (Count)	Classification
GP	0.934	0.048	0.14	0.065	2	Stable and efficient
QX	0.655	0.186	0.55	0.55	1	Improvement potential
GT	0.798	0.153	0.374	0.374	2	Improvement potential
FX	0.824	0.134	0.39	0.39	2	Improvement potential
CA	0.662	0.082	0.227	0.209	0	Low-efficiency and vulnerable
QZ	0.689	0.15	0.311	−0.311	0	Low-efficiency and vulnerable

).

Guangping County represents a stable and efficient region, with its BWIE consistently maintained at a high level. Under the baseline scenario, its efficiency is 0.935, and under the ideal scenario, it reaches the efficiency frontier, with an improvement of only 0.065, indicating limited room for improvement and high marginal difficulty. With an average efficiency of 0.934, a standard deviation of 0.048, and a range of 0.140, the results demonstrate a highly concentrated efficiency distribution and minimal variability. As illustrated by the efficiency heatmap (Figure 5) and trend curve (Figure 5), Guangping County reaches the efficiency frontier in some scenarios and maintains high levels in others. Its overall efficiency is stable, offering a valuable benchmark for other counties to improve.

Qiuxian, Guantao, and Feixiang counties fall into the improvement potential category, sharing the feature of low baseline efficiency but achieving the efficiency frontier under ideal scenarios, demonstrating considerable potential for improvement. In Qiuxian, the baseline BWIE is only 0.450, which rises to 1.000 under the ideal scenario, representing an absolute increase of 0.550 and a relative improvement of more than 120%, highlighting the greatest improvement potential. In Guantao and Feixiang, baseline efficiencies are 0.626 and 0.610, respectively, both of which increase to 1.000 under the ideal scenario, with gains of 0.374 and 0.390, respectively. The average efficiency in these three counties ranges from 0.655 to 0.824, with standard deviations between 0.134 and 0.186 and ranges between 0.374 and 0.550, indicating high variability in BWIE but also considerable plasticity. As further illustrated in Figure 5, these counties exhibit large fluctuations in BWIE across different scenarios, indicating greater sensitivity to external condition adjustments.

Quzhou and Cheng’an counties are classified as Low-efficiency and vulnerable types, with their BWIE being highly constrained and difficult to improve. Quzhou County is on the efficiency frontier under the baseline scenario; however, under the ideal scenario, its BWIE drops to only 0.689, a decrease of 0.311 compared to the baseline, indicating a significant decline in efficiency (Table 6). As shown in the heatmap (Figure 5), Quzhou County is at a high level in S0 but declines significantly under multiple future scenarios. Cheng’an County shows a different pattern: its BWIE increases from 0.557 in the baseline scenario to 0.766 in the ideal scenario, an increase in only 0.209, and its maximum level never reaches the efficiency frontier. Its average efficiency is only 0.662, with a standard deviation of 0.082 and a range of 0.227, indicating a persistent low-efficiency state. The trend chart (Figure 6) shows that this county consistently remains in the low-efficiency range across different scenarios, with limited improvement.

3.3. Identification of Brackish Water Utilization Potential

To further identify inefficiencies in resource allocation, this study introduces the redundancy rate indicator to analyze the input structure across different counties and scenarios. Results reveal that brackish water irrigation volume has the highest average redundancy rate (42.3%), followed by freshwater irrigation (20.7%), with salt-tolerant crop area showing the lowest (10.5%). This suggests that the key to improving BWIE lies in optimizing water resource allocation and use, rather than simply adjusting farmland scale. In terms of input quantities, some counties show freshwater redundancy exceeding ten million cubic meters, brackish water redundancy at the million-cubic-meter scale, and redundant salt-tolerant crop areas reaching hundreds of thousands of mu, highlighting pronounced misallocations in water resource scheduling and crop structure management.

Figure 7 illustrates the redundancy structure of the three input indicators across counties and scenarios. Guangping County consistently shows low redundancy, re-maining close to the efficiency frontier. In contrast, Qiuxian, Cheng’an, and Quzhou display a “high redundancy–low efficiency” pattern, particularly under dry year conditions, while Feixiang and Guantao fall within a moderate range.

Among all counties, Guangping ranks highest in average efficiency (0.934), with the lowest redundancy rates across all three inputs, markedly outperforming the others. In the baseline scenario, redundancy rates are merely 1.6% for crop area (43,000 mu), 13.8% for freshwater (1.27 million m³), and 4.5% for brackish water (0.14 million m³), suggesting an almost optimal allocation with minimal scope for further gains. Qiuxian recorded the lowest average efficiency (0.656), with significant redundancy problems: 21.9% for crop area (209,000 mu), 16.0% for freshwater (15.44 million m³), and 65.4% for brackish water (3.89 million m³), underscoring major shortcomings in land use and water allocation. Cheng’an often records efficiency below 0.6. Redundancy is 13.9% for crop area (149,000 mu), 41.2% for freshwater (18.86 million m³), and 58.0% for brackish water (2.09 million m³), reflecting multi-dimensional redundancy, with structural inefficiencies particularly evident in water use. Quzhou shows pronounced scenario dependence: in the baseline scenario (S0), all inputs record zero redundancy, suggesting a superficially optimal state. However, in other scenarios, redundancies rise sharply, particularly under dry year scenarios (S3, S4): 21.1% for crop area (about 1.005 million mu), 28.1% for freshwater (about 10.05 million m³), and as high as 59.6% for brackish water (about 2.10 million m³). Redundancy in this county is mainly concentrated in water-related inputs, with freshwater and brackish water redundancies generally high, while crop area redundancy remains relatively low. This suggests that although Quzhou shows no redundancy under baseline conditions, once external environments change, its water-use patterns reveal clear redundancy, reflecting vulnerability in resource utilization. Guantao and Feixiang counties maintain average efficiency in the range of 0.79–0.82. Their crop area redundancies are relatively low, but freshwater and brackish water irrigation still have significant optimization potential. For example, in the baseline scenario, Feixiang shows a freshwater redundancy rate as high as 45.3% (25.37 million m³) and a brackish water redundancy rate of 20.1% (1.90 million m³), suggesting that water-saving management and source optimization remain priorities for future improvements.

Scenario comparisons further indicate that external conditions and policy combinations have significant impacts on redundancy rates. In dry year scenarios without full adoption of water-saving irrigation (S1, S3), redundancy rates of freshwater and brackish water irrigation exceed 36% and 53%, respectively, which are significantly higher than the overall averages, indicating that external water resource constraints exacerbate internal system inefficiencies. By contrast, in ideal scenarios with full water-saving irrigation and increased brackish water use (S6, S8), redundancy is significantly reduced. For example, in Qiuxian, input redundancies are completely eliminated, and BWIE improves to the efficiency frontier; in Cheng’an, freshwater redundancy decreases from 41.2% to less than 5% (3.13 million m³), while brackish water redundancy drops to about 22.6% (2.26 million m³); and in Feixiang, freshwater redundancy decreases from 45.3% (25.37 million m³) to zero. These results suggest that water-saving irrigation technology reduces total system water demand, while brackish water substitution optimizes the water source structure. Their combined effect provides an effective pathway to reduce redundancy and push the system closer to the efficiency frontier.

Prior research has focused on mitigating soil salinity and maintaining yield [26,27,28,29,30,31,32]. However, efficiency loss mechanisms from a resource allocation perspective remain unexplored. Applying redundancy analysis within our scenario framework, we quantify input overuse, revealing brackish water redundancy (42.3%) as the primary inefficiency source. This scenario-specific diagnosis provides actionable guidance for optimizing resources under different future conditions.

4. Conclusions and Implications

4.1. Conclusions

This study evaluated BWIE across six counties of Handan City using an input-oriented SBM-DEA model, based on a 2020 baseline and eight projected 2030 scenarios. Results demonstrated that efficiency values, ranging from 0.646 to 0.909, were highly influenced by the combination of water-saving irrigation adoption and hydrological conditions. Scenarios with full water-saving irrigation under normal years (S6 and S8) consistently achieved the highest efficiency, lowest regional disparity, and greatest system stability, highlighting the “water-saving irrigation + normal year” strategy as the optimal pathway. In contrast, efficiency declined during dry years even with increased brackish water use, underscoring the limitations of relying predominantly on brackish water expansion.

Spatial analysis categorized the regions into three distinct types: Guangping County was classified as “Stable and efficient”; Qiuxian, Guantao, and Feixiang Counties were identified as having “Improvement potential”; and Cheng’an and Quzhou Counties were labeled “Low-efficiency and vulnerable”. Redundancy analysis further revealed that water inputs—particularly brackish water—constituted the primary source of inefficiency, emphasizing that optimizing water allocation holds greater importance than expanding agricultural land.

These findings underscore the need to promote comprehensive water-saving technologies, implement tailored regional management strategies, and incorporate brackish water as a supplementary resource within an integrated and optimized framework. Future studies should incorporate crop salinity response models and dynamic efficiency analysis to refine sustainable irrigation strategies under variable hydrological conditions.

4.2. Policy Implications

This study demonstrates that improving hinges on system optimization centered on water-saving irrigation. To this end, the widespread adoption of advanced water-saving technologies such as drip and sprinkler irrigation should be accelerated, particularly in major salt-tolerant crop-producing areas, to lay the foundation for efficiency gains. At the same time, differentiated regional strategies should be implemented according to local conditions: Stable and efficient areas should maintain their advantages and strengthen water quality monitoring. Improvement potential areas should promote the integrated use of water-saving and brackish water irrigation, and low-efficiency and vulnerable areas should optimize crop structure and water source composition while carefully managing the share of brackish water to reduce risks. From a management perspective, a hydrological response mechanism should be established to dynamically adjust freshwater and brackish water allocation in wet and dry years, thereby enhancing system resilience. In addition, a county-level dynamic management system based on efficiency evaluation should be established, incorporating DEA efficiency and redundancy rates into assessment frameworks. Brackish water use should also be integrated into regional water resource strategic planning, thereby promoting coordinated innovation between policy and technology and ensuring long-term sustainability.