Research on the Ecological and Environmental Risk Assessment of Inter-Basin Water Transfer Projects Based on an Improved Sparrow Search Algorithm–Projection Pursuit Model

Abstract

The imbalance between water supply and demand is intensified by population growth and economic development. While water diversion projects are capable of mitigating water shortages, multiple ecological and environmental risks, such as accidental pollution and impairment of ecosystem structure, are introduced by their long-distance water transport and complex corridor environments. The reduction in potential losses hinges on the accurate assessment of these risks. This study integrates the Driving Force–Pressure–State–Impact–Response (DPSIR) model with a projection pursuit model optimized by an improved Sparrow Search Algorithm (SSA) based on seagull optimization and whale optimization operators. A comprehensive risk assessment model was constructed and validated using data from the Chuhe Main Canal for the period 2015 to 2024 as a case study. The results indicate that “water resource utilization rate”, “biodiversity index”, and “public satisfaction” are key factors; project risks have gradually escalated from “relatively low risk” to “relatively high risk”. By this model, the key risk factors and evolutionary patterns of ecological and environmental risks in water diversion projects are able to be scientifically identified, thereby providing a quantitative basis for risk early warning and differentiated management strategies, as well as serving as a reference for the ecological risk assessment of similar inter-basin water diversion projects.

1. Introduction

Water is the foundation of life, production, and ecology. Water demand is increased by rapid urbanization and population growth, while the problems of water scarcity and uneven spatiotemporal distribution are being exacerbated, leading to an intensified supply–demand contradiction. Inter-basin water transfer projects are one of the key engineering measures for optimizing water resource allocation, enabling water to be transferred from areas of relative abundance to those of scarcity. However, water diversion projects are confronted with multiple risks due to their long distances and the complex geographical and climatic zones they traverse, including sudden pollution incidents, flood overflows, and channel instability [1]. Taking the South-to-North Water Diversion Project as an example, although water quality is maintained at a high standard, it is still exposed to risks such as diverse pollution sources and sensitive ecosystems, which may threaten drinking water safety and regional ecological stability. Therefore, it is crucial to conduct ecological and environmental risk assessments for water diversion projects and to incorporate comparative analyses of different water diversion projects [2].

Existing research largely concentrates on individual factors like water quality, hydrological change, or operational efficiency. Richter [3] developed hydrological alteration indicators to quantify flow regime disturbances caused by water diversion; however, his work addresses only the hydrological dimension, neglecting the dynamic responses of socioeconomic pressures and ecosystems. Shaokun [4] optimized the timing of cascade reservoir water storage through reservoir storage analysis models and flood risk assessment methods, thereby mitigating regional disparities in water resource allocation, but the study focused on the benefits of engineering operations and did not address ecological risks such as sudden pollution or biodiversity loss. Shi [5] quantitatively analyzed the mitigation effects of inter-basin water diversion on water scarcity risks in the Baiyangdian region, revealing the critical role of water diversion projects in ecological replenishment during droughts; however, this study focused primarily on the risk of water scarcity and did not consider the evolution of risks during operation. Jing [6] established a risk assessment system for the water conveyance channels of the South-to-North Water Diversion Project’s Henan section, providing quantitative evaluation methods for the safe operation and ecological risk management of long-distance water conveyance projects, but the framework did not explicitly address the nonlinear relationships between high-dimensional indicators. Long [7] attempted to construct a risk indicator system for water resource transfer projects, employing the Analytic Hierarchy Process to determine indicator weights. However, different experts exerted significant influence on weight calculations and the final risk analysis. In summary, existing research lacks systematic modelling of the interaction mechanisms between multiple risks, such as sudden pollution and ecosystem disturbances, and pays insufficient attention to the dynamic evolution of risks during engineering operations; consequently, it struggles to adapt to fluctuations in risk caused by changes in the external environment, such as accelerated urbanization. Furthermore, while some studies [8,9] have attempted to integrate multi-source risks, they have failed to establish a comprehensive logical framework linking risk causes to impact pathways and corresponding mitigation strategies. Therefore, there is an urgent need for integrated research to develop a scientific evaluation system for comprehensively assessing the ecological and environmental risks associated with water diversion projects.

Establishing a comprehensive and clearly structured system of indicators forms the foundation of risk assessment. Many existing indicator systems suffer from fragmentation; whilst methods such as SWOT and PEST are capable of conducting macro-environmental analysis, they struggle to establish quantitative links and causal pathways between indicators. By applying the DPSIR model, originally scattered and isolated ecological and environmental problems can be integrated, and a clear explanation is given of what the risk is, why it is generated, and how it should be addressed. Consequently, an effective framework is offered for investigating the causal linkages among social, economic, and environmental systems. Based on the DPSIR model, Gao [10] constructed an evaluation method for urban water resource resilience, providing decision support for the adaptive management of urban water resources. Li [11] applied the DPSIR framework to assess the ecological security of Anping Port in Taiwan from 2015 to 2021. Key indicators such as water transparency, nitrogen oxides, and ship fuel consumption were identified, revealing a U-shaped trend in the port’s ecological security. However, the DPSIR model struggles to characterize the quantitative relationships among indicators, fails to capture the dynamic interactions between dimensions, and cannot reflect the time lag of response measures. The projection pursuit (PP) model addresses the quantitative deficiency of the DPSIR model by projecting high-dimensional indicator data onto a low-dimensional space, preserving the original data structure automatically, and deriving indicator weights through the identification of optimal projection directions. Nevertheless, the determination of the optimal projection direction is a high-dimensional nonlinear optimization problem, where conventional methods are easily trapped in local optima and suffer from slow convergence. Therefore, this study is built upon the indicator system by integrating optimization algorithms and the PP Model to enhance its capabilities in quantitative assessment and the identification of risk trends.

In terms of evaluation methodologies, traditional approaches such as the Analytic Hierarchy Process (AHP) and the Fuzzy Comprehensive Evaluation Method were widely applied in the early stages. However, these methods are generally characterized by issues including high subjectivity and insufficient retention of data structural characteristics. The PP model constructs projection values by analyzing key indicators that influence the ecological environment of water diversion projects and determines risk levels by combining clustering or threshold rules. It has attracted attention for its ability to effectively handle high-dimensional nonlinear data, preserve structural features of the data, and demonstrate advantages in the assessment of aquatic environments and disaster risks. Zhao [12] employed the Fuzzy Analytic Hierarchy Process to determine indicator weights, integrated the PP model and cluster analysis to classify flood risk levels, and proposed prevention and control strategies considering local conditions to reduce losses caused by flood disasters. However, determining the optimal projection direction for the PP model is inherently a high-dimensional nonlinear optimization problem. Limitations are encountered by traditional methods during data preprocessing, and the model itself lacks self-learning and parallel processing capabilities. For this purpose, the Chicken Swarm Algorithm [13] and the Whale Algorithm [14] have been introduced by scholars to solve high-dimensional optimization problems. For instance, Sun [15] constructed a water security system resilience evaluation index system based on resistance–resilience–adaptability, with a projection pursuit model optimized by the Sparrow Search Algorithm (SSA) being employed to assess the water security resilience level of the Yellow River Basin from 2009 to 2022. Similar to other algorithms, SSA also has drawbacks such as slow convergence, susceptibility to local optima, and reduced population diversity in later stages. In response to these issues, researchers have proposed improvements in areas such as population iteration mechanisms, position updating strategies, and optimization search methods. Leng [16] addressed SSA’s slow convergence and imbalance between global exploration and local exploitation by dynamically adapting the producer ratio and step size factor. This approach was applied to renewable energy scheme selection in a Chinese province, validating the effectiveness of the enhancements. Ji [17] enhanced SSA through multi-algorithm fusion and search strategy optimization, integrating the neighborhood search mechanism of the mayfly algorithm with the mutation-crossover strategy of the differential evolution algorithm. Comparisons with SSA, GA, PSO, and other algorithms demonstrated that ISSA achieved significant improvements in both optimization accuracy and robustness for sparse array optimization problems. The above studies have improved the SSA from various perspectives, demonstrating that the enhanced SSA optimized for PP can effectively enhance global optimization capabilities whilst preserving the characteristics of high-dimensional data structures, and may serve as a useful reference for ecological and environmental risk assessments in water diversion projects.

In summary, existing risk assessment theories are characterized by limitations such as unsystematic modeling, fragmented indicator systems, and inadequate optimization of evaluation methods. The objectives of this study are as follows: (1) to systematically integrate multidimensional factors, including socioeconomic conditions and ecological stress based on the DPSIR model, and construct an indicator system; (2) to propose an improved Sparrow Search Algorithm (SSA) using the seagull algorithm and the Whale Algorithm for optimizing the PP model and the K-means clustering method, so that indicator weights and risk levels can be determined objectively; and (3) to identify key risk factors through case validation and reveal the evolutionary patterns of risks, thereby providing a basis for early warning and differentiated management of ecological and environmental risks in water diversion projects. The innovations are as follows. Indicator fragmentation is overcome by the DPSIR model. The SSA is improved and used to optimize the PP model, by which the convergence speed and global optimization capability for processing high-dimensional data are enhanced, and the reliability of the evaluation results is strengthened. Theoretically, this study enriches the application of the DPSIR framework and the improved projection pursuit model in the ecological and environmental risk management of inter-basin water transfer projects. Practically, scientific basis and technical support are provided for risk prediction, early warning, and the formulation of differentiated control strategies for similar projects.

2. Materials and Methods

2.1. Research Area Overview

This study was conducted at a macro-regional scale to assess the eco-environmental risks associated with an inter-basin water diversion project, taking the Chuhe Main Canal and its affected area as the research subject, as shown in Figure 1.

The Chuhe Main Canal is located in the northern part of Hefei City, Anhui Province, extending westward from the southern side of the Jianghuai watershed and eastward to the Chuhe River Basin. This project, which integrates agricultural irrigation, urban flood control, and water supply, has a total length of approximately 103 km and was fully opened to water conveyance in 1971. The canal alignment passes through several administrative units, including Shushan District, Luyang District, and Changfeng County, along which farmlands, forests, wetlands, and drinking water source protection zones are distributed. The study area is defined as covering the canal body, riparian buffer zones, adjacent land areas, and major catchment zones, thereby effectively incorporating key risk elements such as land use, population distribution, water quality monitoring sections, and ecologically sensitive areas. The canal is characterized by a trapezoidal cross-section and operates under a combination of seasonal regulation and continuous operation, with the water diversion volume being dynamically adjusted according to real demand and water source availability, and the water conveyance being achieved through both open channels and pipelines. The region as a whole receives abundant precipitation, but it is mostly concentrated in summer. The multi-year average annual rainfall is approximately 900–1000 mm. In 2016 and 2018, the precipitation reached 1500 mm and 1310 mm, respectively, representing typical wet years, while in 2022, the precipitation was only 550 mm. In the water-receiving area, a relatively high population density and a high level of urbanization are observed, with industrial zones posing certain demands on water resources. In terms of agricultural planting, characteristic agriculture—such as strawberry and grape cultivation for tourist picking—has been developed, which exerts a significant influence on the regional aquatic ecology and socioeconomic pattern.

2.2. An Ecological and Environmental Risk Assessment Framework for Inter-Basin Water Transfer Projects Based on the DPSIR Model

Risk assessment serves as the foundation of risk management. In contrast to comprehensive scoring methods that rely on subjective weights and fixed ranking rules, a comprehensive risk assessment approach is proposed in this study. In this approach, the optimal projection direction is adaptively determined using an optimization algorithm, a comprehensive projection value is objectively generated, and the risk levels are subsequently classified. The detailed procedure of this method is illustrated in Figure 2.

The selection of risk evaluation indicators is a crucial step in the assessment process. In this study, a hierarchical analysis of the risk assessment indicator system was performed based on the DPSIR framework. For example, the enhancement of response measures can effectively alleviate pressures and improve the state, thereby mitigating negative impacts. Moreover, the deterioration of the state can further promote stronger response measures. The DPSIR framework was adopted as a logical structure for indicator system construction and classification, through which the interconnections among the different dimensions are clearly demonstrated. The logical relationships among the various factors are illustrated in Figure 3.

To ensure the scientific validity and rationality of the evaluation, adherence to the principles of scientific rigor, sustainability, comprehensiveness, and operational feasibility was maintained. An integrated approach combining the DPSIR model, literature review methodology, and expert interviews was employed to establish the indicator system. The specific development process is illustrated in Figure 4.

In this study, driving forces, pressures, state, impacts, and responses were adopted as the first-level indicators for the eco-environmental risk assessment of the inter-basin water diversion project. The conceptual basis of the DPSIR framework is derived from the environmental system causal chain analysis method proposed by institutions such as the European Environment Agency [18]. This framework has been widely used for the construction of indicator systems in fields such as water resources management [19] and ecological security [20]. To determine the specific second-level indicators, a literature search was conducted in databases, including Web of Science, China National Knowledge Infrastructure (CNKI), and Wanfang Data, using keywords such as “inter-basin water diversion”, “eco-environmental assessment of water diversion projects”, and “eco-environmental risk indicator system for water diversion projects”. Only empirical research articles published from 1 January 2015 to 1 March 2026 were selected. The collected literature was read and analyzed, and candidate indicators were preliminarily summarized. After removing duplicates and refining the selection, the indicators were grouped according to the first-level categories, and a preliminary indicator library was established. Subsequently, interviews were conducted with a total of 20 experts from relevant professional fields, including environmental science and engineering, hydraulic engineering and hydrology and water resources, landscape ecology, and geographic information science, thereby ensuring the diversity of the expert panel. Indicators were retained if over 60% of experts deemed them important; those with significant disagreement underwent further analysis to determine their inclusion. Concurrently, indicators not covered by the current system but with substantial practical impact were identified and supplemented. For indicators that were difficult to quantify or for which data was hard to obtain, expert opinions were solicited regarding their handling. Finally, all expert opinions were integrated to further refine and optimize the preliminary indicator database. The final ecological and environmental risk assessment indicator system for inter-basin water diversion projects is presented in

Table 1. Ecological and environmental risk assessment indicator system for water diversion projects.

Primary Indicator	Secondary Indicator	Indicator Type	References
Driving force: $D$	Regional population density (persons per km2): $D_1$	Reverse	[21,22,23]
	Annual growth rate of regional GDP (%): $D_2$	Positive	[24,25,26]
	Natural population growth rate (%): $D_3$	Reverse	[27,28]
	Per capita water resources (m3/person): $D_4$	Positive	[27,29]
	Urbanization rate (%):$D_5$	Reverse	[30,31]
Pressure: $P$	${SO}_2$ emission intensity (tons/100 million yuan): $P_1$	Reverse	[31,32]
	Probability of accidental oil spills from vessels (times per year): $P_2$	Reverse	[31,33]
	Water resource utilization rate (%): $P_3$	Reverse	[34,35]
	Proportion of farmland and construction land (%): $P_4$	Reverse	[36,37]
	Soil erosion index: $P_5$	Reverse	[38,39]
State: $S$	Water quality compliance rate at monitoring sections (%): $S_1$	Positive	[21,22,23]
	Biodiversity index: $S_2$	Positive	[21,25]
	Noise compliance rate along the canal (%): $S_3$	Positive	[24,27]
	Air quality compliance rate (%): $S_4$	Positive	[31,33]
Impact: $I$	Aquatic organism loss rate (%): $I_1$	Reverse	[35,36,37]
	Engel’s coefficient (%): $I_2$	Reverse	[34,35]
	Water supply reliability rate (%): $I_3$	Positive	[27,29]
	Public satisfaction: $I_4$	Positive	[31,33]
Response: $R$	Environmental investment as a percentage of GDP (%): $R_1$	Positive	[30,31]
	Tertiary industry as a percentage of GDP (%): $R_2$	Positive	[24,25,26]
	Centralized sewage treatment rate (%): $R_3$	Positive	[21,22,23]
	Industrial wastewater compliance rate (%): $R_4$	Positive	[34,35]
	Comprehensiveness of aquatic organism protection measures: $R_5$	Positive	[27,29]
	Green coverage rate (%): $R_6$	Positive	[24,27]

.

Driving force is defined as the underlying causes and momentum of ecological and environmental change, primarily driven by socioeconomic and demographic factors. Therefore, the “Indicators for Building Ecological Counties, Cities, and Provinces” was referenced in this study, and the indicators $D_1$, $D_2$, $D_3$, $D_4$, and $D_5$ were selected. Among these indicators, $D_1$ is primarily focused on the spatial agglomeration intensity of the population and is regarded as a key driving force indicator for measuring the intensity of human activities and ecological disturbance in the region. $D_3$ reveals the characteristics of natural population replacement and age structure. In Hefei, which is a rapidly urbanizing area, although the natural growth rate is lower than the migration rate, an indirect but non-negligible effect on the balance between water supply and demand is still exerted through its influence on household size and long-term water use habits. $D_5$ characterizes the stage of urban–rural structural transformation, reflecting the overall changes in land use and industrial patterns. Taken together, these three indicators describe the socioeconomic driving forces from three perspectives: “spatial intensity”, “natural replacement”, and “structural transformation”. Pressure is defined as the impact exerted on the environment under the influence of driving forces, encompassing natural disasters and risks from human activities. Guided by the Technical Specifications for Ecological Environment Status Evaluation (HJ 192-2015) [40], large amounts of ${SO}_2$ are emitted by urban industrial activities. After being deposited into water bodies via dry and wet deposition, water acidification is exacerbated, thereby exerting indirect but long-term negative impacts on the water quality of the water diversion project and on the aquatic ecosystems along the canal. Although the Chuhe Main Canal is not classified as a high-grade waterway, certain risks are posed by local navigation, maintenance operations, and the land-based transport of dangerous goods via bridges crossing the canal. In the event of an oil spill or a hazardous material leakage accident, the open water conveyance channel would be directly contaminated, leading to devastating consequences for the safety of downstream water supply and for the aquatic organisms inhabiting the canal. Therefore, it is of practical significance to regard these events as low-probability but high-consequence risk sources. Additionally, increased water resource consumption has been driven by rapid urban development, with vast ecological lands being converted into urban construction sites, hydrological cycles being disrupted, and soil erosion being exacerbated, ultimately resulting in damage to ecological landscapes. In summary, $P_1$, $P_2$, $P_3$, $P_4$, and $P_5$ are identified as significant risk sources. The state S directly reflects the current environmental condition under pressure, primarily characterized by $S_1$, $S_3$, and $S_4$, representing regional water, air, and noise quality states, while $S_2$ is used to characterize the functional state of landscape ecosystems. Impacts I encompass socioeconomic and ecological dimensions, with $I_1$, $I_2$, $I_3$, and $I_4$ being selected. Responses R are defined as countermeasures implemented in response to driving forces, pressures, and states, primarily through economic and ecological–environmental management, and policy responses, with the indicators $R_1$, $R_2$, $R_3$, $R_4$, $R_5$, and $R_6$ being selected. The calculation methods and data acquisition approaches for each indicator are detailed as follows:

(1) $D_1$. This indicator is defined as the ratio of the permanent population in the study area to its total land area. A higher value of this indicator indicates greater demand for water resources, which impacts the supply–demand balance and water quality safety of water diversion projects.

(2) $D_2$. The growth rate of GDP during the evaluation period relative to the previous period. This indicator reflects that economic expansion has driven increased demand for industrial and domestic water use, potentially intensifying water resource exploitation and pollution discharge.

$$R_{GDP}={\displaystyle \frac{{GDP}_t-{GDP}_{t-1}}{{GDP}_{t-1}}}.$$

(1)

(3) $D_3$. The ratio of natural population growth to the average total population over a given period. Population growth is considered to directly increase water demand, placing long-term pressure on the water supply capacity of water diversion projects. In rapidly urbanizing areas, population increase is mainly driven by migration, whereas the natural growth rate is considered more as a reflection of population structure characteristics and serves as an indicator of the stage of population development. Consequently, the structure of water demand and the transmission pathways of ecological risks are indirectly influenced.

$$R={\displaystyle \frac{B-D}{\overline{P}}}\times 100\%.$$

(2)

(4) $D_4$. This is defined as the ratio of a region’s annual total water resources to its annual resident population, which is used to precisely reflect how a region’s water supply and demand capacity influences the ecological risks of water diversion projects. The lower the value of this indicator, the greater the dependence of the region on the water diversion project is observed, and the higher the operational pressure and ecological disturbance risk of the project are consequently induced.

(5) $D_5$. This is defined as the ratio of the urban resident population to the total resident population in the study area. Data can be obtained directly from the National Bureau of Statistics. The accelerated urbanization process drives the expansion of built-up land, resulting in the conversion of cropland and ecological land into impervious surfaces. Consequently, the surface runoff coefficient is increased, and the runoff processes along channel corridors are altered. Concurrently, regional water demand pressures are intensified by the agglomeration of urban populations and industrial activities, while the loads of domestic and industrial pollutants are elevated. As a result, the water quality security of water diversion projects and the stability of ecosystem functions are significantly impacted.

(6) $P_1$. This is defined as the ratio of the annual total ${SO}_2$ emissions in the region to its annual GDP. Data is primarily obtained from the Pollutant Discharge Statistics Bulletin issued by the ecological environment authorities. Urban industrial emissions impact air quality and acid rain formation, indirectly affecting water acidification and ecosystem health.

(7) $P_2$. This indicator is typically calculated as the ratio of the total number of oil spills in the region over the past decade to the number of years covered in the statistics. Although the Chuhe Main Canal is not classified as a high-grade navigable waterway, regular passage of water conservancy patrol vessels, dredging boats, and a limited number of agricultural boats is carried out within the canal. Multiple highway bridges crossing the canal are situated along its route. In the event of a traffic accident involving a road-based hazardous material transport vehicle on these bridges, oil leakage into the water body could be triggered. Although the probability of such an event is considered extremely low, severe consequences would be posed, by which aquatic ecological integrity and water supply security are directly threatened. For specific details, see [33].

(8) $P_3$. This is defined as the proportion of water consumption in a study area relative to its total water resources. The higher the indicator, the more pronounced the water supply–demand imbalance is considered to be, and the greater the ecological stress risk.

(9) $P_4$. This is defined as the proportion of cultivated land and construction land combined relative to the total land area within the study region. Data should be obtained primarily from land use status records maintained by national land and spatial planning authorities. Land use changes are recognized to impact surface runoff, soil erosion, and ecological connectivity, thereby increasing the risk of ecological disturbance from engineering projects:

$$L={\displaystyle \frac{A_c+A_b}{A_t}},$$

(3)

where $A_{\mathrm{c}}$ represents the area of cultivated land, $A_{\mathrm{b}}$ represents the area of construction land, and $A_{\mathrm{t}}$ represents the total land area of the region.

(10) $P_5$. This is defined as the regional average soil erosion intensity level, which can be determined by referring to the “Soil Erosion Classification and Grading Standard” (SL 190-2007) [41] issued by the Ministry of Water Resources. Increased soil erosion is considered to accelerate sediment input, affecting water quality, causing channel siltation, and damaging riparian ecosystems:

$$S={\displaystyle \frac{\sum (G_i\times A_i)}{A_t}},$$

(4)

where $G_i$ represents the soil erosion intensity grade value for the i-th type, $A_i$ denotes the area corresponding to that grade, and $A_{\mathrm{t}}$ signifies the total area of the region.

(11) $S_1$. The ratio of compliant monitoring instances to total monitoring instances during the evaluation period, which is obtained through routine monitoring by ecological and environmental authorities, directly reflects the water quality status of the water diversion project.

(12) $S_2$. This indicator is used to measure biodiversity levels within water diversion areas. A decline in biodiversity is considered to indicate an increased risk of ecological degradation:

$$H^{\ast }=-{\displaystyle {\sum }_{i=1}^S}\left(P_i\times {ln}_{P_i}\right),$$

(5)

where $S$ denotes the number of species, $n_i$ represents the number of individuals in the $i$-th species, and $N$ is the total number of individuals across all species, $P_i={\displaystyle \frac{n_i}{N}}$.

(13) $S_3$. This indicator is determined based on the “Environmental Noise Quality Standards” (GB 3096-2008) [42], with data being obtained through fixed-point monitoring and specialized reports. It reflects the current quality of the acoustic environment along the route. This noise pollution is considered to disrupt wildlife behavior and habitats, particularly affecting sensitive species such as birds and mammals, thereby compromising ecosystem stability.

(14) $S_4$. This is defined as the ratio of the number of days meeting air quality standards to the total number of valid monitoring days in the study area, which is determined based on the Environmental Air Quality Standards (GB 3095-2012) [43]. These pollutants are considered to enter water bodies through wet and dry deposition, impacting water quality and the health of aquatic organisms.

(15) $I_1$. This indicator measures the relative reduction in biomass within aquatic ecosystems under environmental stress, which is used to quantify the direct negative consequences experienced by these ecosystems:

$$L={\displaystyle \frac{B_0-B}{B_0}},$$

(6)

where $B_0$ represents the initial biomass, and $B$ denotes the current biomass. In this study, $B_0$ was defined as the average biomass over three consecutive years before the operation of the Chuhe Main Canal project, with data sourced from reports published by the Hefei Municipal Bureau of Ecology and Environment and from field sampling surveys. $B$ is obtained from standardized sampling data taken at the same season and the same monitoring section each year, and consistency in sampling method, section location, and analytical procedure was ensured.

(16) $I_2$. This is defined as the ratio of annual household food expenditures to total annual consumption expenditures in a given region, which is considered to indirectly influence public awareness of ecological environments and participation in environmental protection efforts.

(17) $I_3$. This is defined as the ratio of actual water supply to total water demand in the study area. Data can be obtained from the China Water Resources Bulletin. A low security rate is considered to trigger conflicts over socioeconomic and ecological water use.

(18) $I_4$. This indicator is generally obtained through questionnaires, which are used to reflect the public’s perception and acceptance of the environmental impacts of water diversion projects.

(19) $R_1$. This is defined as the proportion of ecological and environmental-related investments in the study area relative to the total GDP during the same period. This indicator reflects the national emphasis on safeguarding water security and restoring the ecology of water sources, as well as the level of fiscal investment. It is considered to influence risk prevention and control capabilities and ecological restoration capacity.

(20) $R_2$. This is defined as the proportion of the tertiary industry’s value-added output in the total GDP of the study area during the same period, which can be obtained from the National Bureau of Statistics. This indicator is considered to help reduce industrial pollution emissions and alleviate environmental pressure.

(21) $R_3$. The ratio of centralized treatment capacity at wastewater treatment plants to total wastewater discharge volume is considered a core control measure, with its treatment rate directly reflecting the effectiveness of the response:

$$\rho ={\displaystyle \frac{T_t}{T_d}},$$

(7)

where $T_{\mathrm{t}}$ represents the centralized treatment volume of the wastewater treatment plant, and $T_{\mathrm{d}}$ represents the total volume of wastewater discharged.

(22) $R_4$. The ratio of annual compliant industrial wastewater discharge volume to total annual industrial wastewater discharge volume, as referenced in the China Ecological Environment Statistical Yearbook, is considered a key measure for reducing point source pollution and safeguarding water quality.

(23) $R_5$. Data for this indicator is typically evaluated and scored by experts through engineering document reviews and field surveys, directly impacting the effectiveness of biodiversity conservation efforts.

(24) $R_6$. This is defined as the proportion of green coverage area within a study region relative to its total land area, which is considered to effectively mitigate engineering disturbances and enhance ecosystem service functions.

To validate the scientific rigor and reliability of the indicator system, this study conducted reliability and validity tests upon completion of the data collection process. For specific verification, refer to Section 2.5.1 on data collection.

(1)

Reliability test

Cronbach’s alpha coefficient is a commonly used measure of internal consistency for multidimensional scales or indicator systems. A higher value indicates stronger correlations among indicators and greater reliability of results [44]. The formula for assessing internal consistency using Cronbach’s alpha coefficient is as follows:

$$\alpha ={\displaystyle \frac{k}{k-1}}\left(1-{\displaystyle \frac{{\displaystyle {\sum }_{i=1}^k} {\sigma }_i^2}{{\sigma }_T^2}}\right),$$

(8)

where $k$ represents the number of indicators, ${\sigma }_i^2$ denotes the variance of the $i^{th}$ indicator, and ${\sigma }_T^2$ signifies the variance of the total score. Generally, $\alpha >0.7$ indicates a good level, while $\alpha >0.8$ signifies an excellent level.

(2)

Validity test

The validity of the indicator system was assessed through KMO and Bartlett’s sphericity tests. The KMO test compares simple and partial correlation coefficients among variables; a KMO value closer to 1 indicates stronger inter-indicator correlations, making the system suitable for risk analysis [45]. Bartlett’s sphericity test examines whether the correlation matrix is an identity matrix. A significant $p<0.05$ indicates substantial correlations among variables, warranting further analysis [46].

2.3. Indicator Classification and Grading

Indicator classification and grading were guided by the multi-level evaluation principle outlined in the Technical Specification for Ecological Environment Status Evaluation (HJ 192-2015) [40], which explicitly adopts a five-tier grading system for assessing ecological environment status. In this study, by adhering to the given specification, the risk levels were preliminarily classified into five categories: lower risk (Level I), low risk (Level II), moderate risk (Level III), higher risk (Level IV), and high risk (Level V). The ecological risk of the Chuhe Main Canal is driven by the coupled effects of socioeconomic and natural ecological factors, showing a continuous gradient pattern. A three-tier classification is too coarse to capture moderate risks, while a seven-tier or higher resolution is excessively detailed and poses implementation challenges in practice. In contrast, a five-tier risk classification is widely adopted as the mainstream approach, through which clear distinctions from low to high risk are able to be precisely made. However, this classification is employed solely for indicator standardization and subsequent standard sample generation, and is not regarded as the criterion for final risk level determination. The final risk level is derived from a data-driven objective clustering based on the projection values output by the projection pursuit model using K-means clustering, with each cluster then mapped to a risk level according to engineering practice.

Based on indicator characteristics, they are categorized into qualitative and quantitative indicators. Qualitative indicators are primarily evaluated through expert scoring, while risk level classifications for quantitative indicators are generally determined using actual data from engineering case studies. Each qualitative indicator shares identical quantitative intervals for its risk levels: $\left[0,\left.2\right)\right.$, $\left[2,\left.4\right)\right.$, $\left[4,\left.6\right)\right.$, $\left[6,\left.8\right)\right.$, and $\left[8,10\right]$. The qualitative indicators $I_4$ and $R_5$ are included in this study, with all others being quantitative. The risk evaluation classifications for each secondary indicator are presented in

Table 2. Evaluation grading for secondary indicators.

Indicator	I	II	III	IV	V
$D_1$	$\left[0,\left.50\right)\right.$	$\left[50,\left.150\right)\right.$	$\left[150,\left.250\right)\right.$	$\left[250,\left.500\right)\right.$	$\left[500,+\infty \right]$
$D_2$	$\left[7.5,\left.+\infty \right)\right.$	$\left[6,\left.7.5\right)\right.$	$\left[4.5,\left.6\right)\right.$	$\left[3.1,\left.4.5\right)\right.$	$\left[-\infty ,3\right]$
$D_3$	$\left[-\infty ,\left.-1.908\right)\right.$	$\left[-1.908,\left.-0.482\right)\right.$	$\left[-0.482,\left.0\right)\right.$	$\left[0,\left.2.15\right)\right.$	$\left[2.15,+\infty \right]$
$D_4$	$\left[3000,\left.+\infty \right)\right.$	$\left[1700,\left.3000\right)\right.$	$\left[1000,\left.1700\right)\right.$	$\left[500,\left.1000\right)\right.$	$\left[0,500\right]$
$D_5$	$\left[0,\left.35\right)\right.$	$\left[35,\left.40\right)\right.$	$\left[40,\left.45\right)\right.$	$\left[45,\left.50\right)\right.$	$\left[50,+\infty \right]$
$P_1$	$\left[0,\left.0.5\right)\right.$	$\left[0.5,\left.1\right)\right.$	$\left[1,\left.2\right)\right.$	$\left[2,\left.4\right)\right.$	$\left[4,+\infty \right]$
$P_2$	$\left[0,\left.0.005\right)\right.$	$\left[0.005,\left.0.01\right)\right.$	$\left[0.01,\left.0.05\right)\right.$	$\left[0.05,\left.0.1\right)\right.$	$[0.1,+\infty )$
$P_3$	$\left[0,\left.20\right)\right.$	$\left[20,\left.40\right)\right.$	$\left[40,\left.60\right)\right.$	$\left[60,\left.80\right)\right.$	$\left[80,100\right]$
$P_4$	$\left[0,\left.20\right)\right.$	$\left[25,\left.35\right)\right.$	$\left[35,\left.45\right)\right.$	$\left[45,\left.55\right)\right.$	$\left[55,100\right]$
$P_5$	$\left[0,\left.0.15\right)\right.$	$\left[0.15,\left.0.16\right)\right.$	$\left[0.16,0.1\left.7\right)\right.$	$\left[0.17,\left.0.18\right)\right.$	$\left[0.18,0.19\right]$
$S_1$	$\left[85,\left.100\right)\right.$	$\left[70,\left.85\right)\right.$	$\left[55,\left.70\right)\right.$	$\left[30,\left.55\right)\right.$	$\left[0,30\right]$
$S_2$	$\left[0.95,\left.1\right)\right.$	$\left[0.9,\left.0.95\right)\right.$	$\left[0.85,\left.0.9\right)\right.$	$\left[0.8,\left.0.85\right)\right.$	$\left[0,0.8\right]$
$S_3$	$\left[95,\left.100\right)\right.$	$\left[85,\left.95\right)\right.$	$\left[75,\left.85\right)\right.$	$\left[60,\left.75\right)\right.$	$\left[0,60\right]$
$S_4$	$\left[90,\left.100\right)\right.$	$\left[80,\left.90\right)\right.$	$\left[70,\left.80\right)\right.$	$\left[60,\left.70\right)\right.$	$\left[0,60\right]$
$I_1$	$\left[0,\left.5\right)\right.$	$\left[5,\left.15\right)\right.$	$\left[15,\left.30\right)\right.$	$\left[30,\left.50\right)\right.$	$\left[50,\left.+\infty \right)\right.$
$I_2$	$\left[0,\left.30\right)\right.$	$\left[30,\left.40\right)\right.$	$\left[40,\left.50\right)\right.$	$\left[50,\left.60\right)\right.$	$\left[60,\left.+\infty \right)\right.$
$I_3$	$\left[95,100\right]$	$\left[90,\left.95\right)\right.$	$\left[80,\left.90\right)\right.$	$\left[70,\left.80\right)\right.$	$\left[0,\left.70\right)\right.$
$I_4$	$\left[80,\left.100\right)\right.$	$\left[60,\left.80\right)\right.$	$\left[40,\left.60\right)\right.$	$\left[20,\left.40\right)\right.$	$\left[0,20\right]$
$R_1$	$\left[0.5,+\infty \right]$	$\left[0.25,\left.0.5\right)\right.$	$\left[0.1,\left.0.25\right)\right.$	$\left[0.05,\left.0.1\right)\right.$	$\left[0,\left.0.05\right)\right.$
$R_2$	$\left[40.2,100\right]$	$\left[36.7,\left.40.2\right)\right.$	$\left[33.8,\left.36.7\right)\right.$	$\left[28.6,\left.33.8\right)\right.$	$\left[28.6,\left.33.8\right)\right.$
$R_3$	$\left[85,\left.100\right)\right.$	$\left[80,\left.85\right)\right.$	$\left[75,\left.80\right)\right.$	$\left[70,\left.75\right)\right.$	$\left[65,70\right]$
$R_4$	$\left[95,100\right]$	$\left[90,\left.95\right)\right.$	$\left[85,\left.90\right)\right.$	$\left[80,\left.85\right)\right.$	$\left[0,\left.80\right)\right.$
$R_5$	$\left[80,\left.100\right)\right.$	$\left[60,\left.80\right)\right.$	$\left[40,\left.60\right)\right.$	$\left[20,\left.40\right)\right.$	$\left[0,20\right]$
$R_6$	$\left[85,100\right]$	$\left[75,\left.85\right)\right.$	$\left[50,\left.75\right)\right.$	$\left[30,\left.50\right)\right.$	$\left[0,\left.30\right)\right.$

Note(s): Source: Technical Specification for Ecological Environment Status Evaluation (HJ 192-2015).

.

2.4. Risk Assessment Model

Unlike traditional water engineering projects such as reservoir operation and single-river channel regulation, inter-basin water diversion projects are characterized by large spatial spans, the coupling of multiple risks, limited sample data, and high-dimensional indicators. To scientifically evaluate the ecological and environmental risks of inter-basin water diversion projects, this study developed an evaluation model based on the Sparrow Search Algorithm with Projection Pursuit (SWSSA-PP). The model encompasses the construction of the projection pursuit model, with optimization via the improved Sparrow Search Algorithm, projection value calculation, and risk level determination.

2.4.1. Projection Tracking Model Construction

(1)

Data collection and preprocessing

Let the indicator value sample set be $\left\{x^{\ast }(i,j)\right|i=1,2,\dots ,n;j=1,2,\dots ,p\}$, where $x^{\ast }(i,j)$ represents the jth indicator value of the ith sample. n and p denote the number of samples and indicators, respectively. To mitigate the impact of varying indicator data dimensions on results, standardization is required. For benefit and cost-type indicators, the standardization formulas are given by Equations (9) and (10).

$$x^{\ast }(i,j)={\displaystyle \frac{x(i,j)-minx\left(j\right)}{maxx\left(j\right)-minx\left(j\right)}},$$

(9)

$$x^{\ast }(i,j)={\displaystyle \frac{maxx\left(j\right)-x(i,j)}{maxx\left(j\right)-minx\left(j\right)}}.$$

(10)

After preprocessing, the set ${\left[x_{ij}^{\ast }\right]}_{n\times 24}$ is obtained. Given the limited number of subjects in this study, direct data mining calculations exhibit poor stability. Therefore, a random sample generation method based on risk level classification was employed to augment the dataset. Uniformly distributed random sampling was applied to the value ranges of each indicator, generating 200 combinations to form the synthetic data. The synthetic samples simulate typical combinations of indicators across different risk levels, covering the feature space of each level. Their value ranges align with the level intervals of real data. Together with the evaluation data, they form the computational data ${\left[x_{ij}^{\ast }\right]}_{\left(n+1000\right)\times 24}$ for data mining and exploration of weight information.

(2)

Projection Indicator Function Construction

Integrate the dataset into a one-dimensional projection with projection direction $\alpha =\{a\left(1\right),a\left(2\right),\cdots ,a\left(n\right)\}$, where the projection values are $Z\left(i\right)$ and satisfy ${\sum }_{j=1}^m{\alpha }^2\left(j\right)=1$.

$$Z\left(i\right)={\displaystyle {\sum }_{j=1}^p} a\left(j\right)x\left(i,j\right).$$

(11)

Here, $Z\left(i\right)$ is the comprehensive evaluation value of the sample in this study, directly representing the level of ecological and environmental risk of the water diversion project. The larger the projection value, the higher the risk.

The projection indicator function $Q\left(a\right)$ is defined as follows:

$$Q\left(a\right)=S_z{D}_z,$$

(12)

$$S_z=\sqrt{{\displaystyle \frac{{\displaystyle {\sum }_{i=1}^m}{\left(z\left(i\right)-E\left(z\right)\right)}^2}{m-1}}},$$

(13)

$$D_z={\displaystyle {\sum }_{i=1}^m}{\displaystyle {\sum }_{j=1}^m}\left(R-r\left(i,j\right)\right)·\mu \left(R-r\left(i,j\right)\right),$$

(14)

where $S_z$ denotes the inter-class distance; $D_z$ represents the local density of $Z\left(i\right)$; $E\left(z\right)$ is the mean of $\left\{Z\left(i\right)|i=1,2,\dots ,n\right\}$; $R$ is the $D_z$ window radius, where $r\left(i,j\right)=\left|z\left(i\right)-z\left(j\right)\right|$; and $\mu \left(R-r\left(i,j\right)\right)$ is the unit step function:

$$\mu \left(R-r\left(i,j\right)\right)=\left\{\begin{array}{c}1,R-r\left(i,j\right)\ge 0\\ 0,R-r\left(i,j\right)<0\end{array}\right..$$

(15)

(3)

Projection Indicator Function Optimization

Building upon prior research findings [47,48], the optimal projection direction ${\alpha }^{\ast }$ is computed by maximizing the projection indicator function. The optimal projection function is defined as follows:

$$maxQ\left(a\right)=S_z{D}_z.$$

(16)

2.4.2. Improving the Sparrow Search Algorithm

The Sparrow Search Algorithm (SSA) [49], proposed in 2020, is a novel optimization algorithm characterized by few parameters, a simple structure, and a relatively fast convergence speed. The real sample size of this study is relatively small, for which stable convergence of the conventional SSA is hard to attain within a limited iteration budget, and the algorithm is easily trapped in local optima during later stages as a result of reduced population diversity. Consequently, the incorporation of multiple improvement strategies is necessary. The multi-strategy improved SSA proposed by Liu [50] is referenced in this study, by which population diversity is effectively increased, the trade-off between global search and local exploitation is balanced, and it is found to be especially applicable to small-sample high-dimensional optimization problems, specifically as follows:

(1)

Increase the proportion of adaptive populations

The fixed proportion of individuals within each category in a sparrow population may reduce the efficiency of global search. This study employs the following formula to enhance the ratio coefficient between producers and followers [50]:

$$\left\{\begin{array}{c}a=0.15\times {\displaystyle \frac{-2t}{T_{max}}}-0.1k+0.1\\ N_{PD}=aN\\ N_{SD}=(1-a)N\end{array}\right.,$$

(17)

where $N_{PD}$ and $N_{SD}$ represent the numbers of producers and followers, respectively; $a$ denotes the ratio of $N_{PD}$ to $N_{SD}$; $t$ and $T_{max}$ denote the current and maximum iteration counts; $N$ denotes the population size; and $k$ denotes a random number in the range [0, 1].

(2)

Integrating the Whale Algorithm with bubble net predation strategy

When sparrows sense danger, they relocate to safe zones. The shrinking enclosure mechanism and spiral position update mechanism of the whale optimization algorithm are employed to update the sparrow producer positions, enhancing SSA’s resistance to local optima [49]. The formula for updating producer positions is [50]:

$$X_{i,j}^{t+1}=\left\{\begin{array}{cc}{X}_{best}^t-A|C{X}_{best}^t-X_{i,j}^t|,& \mathrm{if}{R}_2<ST\\ X_{best}^t+D{e}^{kl}COS\left(2\pi l\right),& \mathrm{if}{R}_2\ge ST\end{array}\right.,$$

(18)

where $X_{i,j}^t$ denotes the position of the ith sparrow in the jth dimension, $D=X_{best}^t-X_{i,j}^t$; $lϵ[-1,1]$; $k$ is a constant; $A=2a{r}_1-a$; $C=2r_2$; $a=2-2{\displaystyle \frac{t}{T_{max}}}$; $r_1,r_2$ are random numbers in $[-1,1]$; $X_{best}^t$ denotes the optimal position of the producer at the current iteration; and $R_2$ and $\mathrm{S}\mathrm{T}$ represent the warning threshold and safety threshold, respectively.

(3)

Fusion-Enhanced Seagull Optimization Algorithm Operator

The spiral attack strategy of the Seagull Optimization Algorithm simulates the hierarchical structure and foraging behavior of seagulls [51], addressing the local search limitations of SSA. The follower position formula is updated to [50]:

$$\left\{\begin{array}{l}x=r\mathrm{c}\mathrm{o}\mathrm{s}\theta \\ y=r\mathrm{s}\mathrm{i}\mathrm{n}\theta \\ \begin{array}{l}z=r\theta \\ r=\mathrm{\mu }{\mathrm{e}}^{\theta v}\\ \begin{array}{l}A=f_{\mathrm{c}}-\left[t\left({\displaystyle \frac{f_{\mathrm{c}}}{T_{\mathrm{m}\mathrm{a}\mathrm{x}}}}\right)\right]\\ B=2A^2{r}_{\mathrm{d}}\\ \begin{array}{l}{G}_{\mathrm{s}}\left(t\right)=A{p}_{\mathrm{s}}\left(t\right)\\ M_{\mathrm{s}}\left(t\right)=B\left[P_{\mathrm{b}\mathrm{s}}\left(t\right)-P_{\mathrm{s}}\left(t\right)\right]\\ P_{\mathrm{s}}\left(t\right)=\left|G_{\mathrm{s}}\left(t\right)+M_{\mathrm{s}}\left(t\right)\right|xyz+P_{\mathrm{b}\mathrm{s}}\left(t\right)\end{array}\end{array}\end{array}\end{array}\right.,$$

(19)

where $r$ denotes the radius of the spiral motion; $\theta ϵ[0,2\pi ]$; $\mu$ and $v$ are constants; $f_{\mathrm{c}}$ linearly decreases from 2 to 0; $r_{\mathrm{d}}$ is a random number in [0, 1]; $P_{\mathrm{s}}\left(t\right)$ and $P_{\mathrm{b}\mathrm{s}}\left(t\right)$ represent the current position and optimal position of the seagull; and $G_{\mathrm{s}}\left(t\right)$ and $M_{\mathrm{s}}\left(t\right)$ denote collision-free new positions and optimal positions, respectively.

The linear decay rate of the convergence factor $A$ is excessively slow. This study employs a quadratic decay strategy to adjust the convergence factor, with the specific formula as follows [50]:

$$A=1-{\left({\displaystyle \frac{t}{T_{max}}}\right)}^2.$$

(20)

Adaptive weight is a crucial parameter requiring adjustment in the algorithm [52]. It is set larger in early iterations to broaden the global search scope; in later iterations, it is reduced to enhance the algorithm’s local exploitation capability. The formula is as follows [50]:

$$\omega ={\omega }_{min}+\left({\omega }_{max}-{\omega }_{min}\right)\mathrm{s}\mathrm{i}\mathrm{n}\left(\pi +{\displaystyle \frac{t\pi }{2T_{max}}}\right).$$

(21)

The large stride length of the Leviathan flight algorithm facilitates expanding the search range in the early stages, reducing the risk of getting stuck in local optima; a smaller stride length in the later stages improves the quality of the algorithmic solution. The updated follower position formula is [50]:

$$P_{\mathrm{s}}\left(t\right)=\left|G_{\mathrm{s}}\left(t\right)+M_{\mathrm{s}}\left(t\right)\right|xyz{L}_{\left(\lambda \right)}+L_{\left(\lambda \right)}\omega P_{\mathrm{b}\mathrm{s}}\left(t\right),$$

(22)

where $L_{\left(\lambda \right)}$ denotes the random search path; s represents the random stride, where $s={\displaystyle \frac{{\sigma }_{\mu }\mu }{{\left|\nu \right|}^{\frac{1}{\beta }}}}$, with $\mu$ and $\nu$ following a normal distribution, ${\sigma }_{\mu }={\left({\displaystyle \frac{T\left(1+\beta \right)\mathrm{s}\mathrm{i}\mathrm{n}\left(\pi \frac{\beta }{2}\right)}{T\left[\left(1+\beta \right)/2\right]\beta \times 2^{\left(\beta -1\right)/2}}}\right)}^{{\displaystyle \frac{1}{\beta }}}$.

Assuming that 10–20% of sparrows are at risk, with initial positions randomly generated, the position update formula for the early-warning sparrow is [50]:

$$X_{i,j}^{t+1}=\left\{\begin{array}{c}{X}_{best}^t+\beta \left|X_{i,j}^t-X_{best}^t\right|,\mathrm{if}{f}_i>f_g\\ X_{i,j}^t+K\left({\displaystyle \frac{\left|X_{i,j}^t-X_{worst}^t\right|}{\left(f_i-f_w\right)+\epsilon }}\right),\mathrm{if}{f}_i=f_g\end{array}\right.,$$

(23)

where $\beta$ is the step size control parameter; $\mathrm{K}$ is the sparrow step size control parameter; $f_i$ is the current sparrow’s fitness values; $f_g$ and $f_w$ represent the global optimal and worst fitness values, respectively; and $\epsilon$ is a constant chosen to be as small as possible.

2.4.3. Improved Sparrow Search Algorithm for Projection-Based Path Tracing (SWSSA-PP) Model

Step 1. Initialize SWSSA parameters and the sparrow population.

Based on the optimization requirements, key parameters are set and an initial sparrow population is generated, where each individual corresponds to a projection direction. The population size N is set within the range of 30 to 50; the iteration count t and maximum iteration limit $T_{\mathrm{m}\mathrm{a}\mathrm{x}}$ are controlled between 100 and 500 to prevent premature convergence; the producer ratio $N_{PD}$ is set to 0.2 and dynamically adjusted according to adaptive strategies; the warning threshold is $\left[0,1\right]$, with a safety threshold of 0.8; the Whale Algorithm parameter $a=2-2{\displaystyle \frac{t}{T_{max}}}$; the seagull operator parameters $\mu =1,\nu =0.1$; and the Lévy flight parameter $\beta =1.5$. Dynamically adjust $N_{PD}$ and $N_{SD}$ based on iteration count. Each sparrow individual is represented as a P-dimensional vector ${\overrightarrow{a}}_k=\left(a_{k1},a_{k2},\dots ,a_{kp}\right)$, ($k=1,2,\dots ,N$), which is randomly generated within $\left[-1,1\right]$ and then normalized.

Step 2. Read the indicator data from the evaluation metric system and standardize it using Equations (9) and (10).

Step 3. Calculate the fitness values for the initial population.

Employ the modified SSA to determine the optimal projection direction, using Equation (12) as the fitness function. For each individual ${\overrightarrow{a}}_k$, compute the projection values for all samples according to Equation (11). Calculate $S\left({\overrightarrow{a}}_k\right)$ and $D\left({\overrightarrow{a}}_k\right)$ using Equations (13) and (14); record the global optimal fitness and corresponding ${\overrightarrow{a}}_{best}$ for the initial population.

Step 4. Iterative optimization.

(1) Calculate fitness values for the initial population using Equation (12) and sort them.

(2) Update the positions of producers, followers, and early-warning agents using Equations (18), (22) and (23).

(3) Calculate the fitness $f_i\left(t+1\right)$ for all individuals after updating; if $f_i\left(t+1\right)>f_{best}\left(t\right)$, update the global optimum; and if $f_i\left(t+1\right)=f_{best}\left(t+1\right)$ and ${\overrightarrow{a}}_{best}\left(t+1\right)={\overrightarrow{a}}_i\left(t+1\right)$, then $t=t+1$.

(4) Repeat steps (2) to (3) until $t=T_{max}$, and then output the optimal projection direction and projection value.

The pseudocode is presented as follows (Algorithm 1).

Algorithm 1. The framework of the SWSSA

Input:

T_max: maximum iterations; N: number of sparrows;

SD_ratio: proportion of scouts;

Initialize population and set relevant parameters;

t = 0; SD_num = N × SD_ratio; PD_num = N × 0.2;

Output: x(best), f(g)

1: While (t < T_max)

2: Evaluate fitness values of all sparrows and find the current global best position

3: Sort sparrows by fitness

4: //Adaptive population proportion strategy

5: a = 0.15 × (2 × t/T_max) − 0.1 × rand() + 0.1

6: PD_num = max(1, round(a × N))

7: Follow_num = N-PD_num

8: SD_num = round(N × SD_ratio)

9: //Update producers (the first PD_num best individuals)

10: for i = 1 to PD_num

11: Update the sparrow’s location according to Equation (18)

12: end for

13: //Update followers

14: for i = PD_num + 1 to PD_num + Follow_num

15: Update the sparrow’s location by using Equation (22)

16: end for

17: //Randomly select SD_num scouts to update

18: rand_index = random sample of size SD_num from {1, 2, …, N}

19: for each i in rand_index

20: Update the sparrow’s location

21: end for

22: //Boundary handling and greedy selection

23: Clip all positions to the feasible bounds

24: Keep better solutions

25: t = t + 1

26: end while

27: return x(best), f(g)

2.4.4. Determination of Risk Levels

(1)

Data preparation and initialization

Extract the optimal projection values $Z^{\ast }$ for all samples, the projection value $Z_1^{\ast }$ for the standard sample set, and the projection value $Z_2^{\ast }$ for the evaluation sample set. Integrate these into the input dataset {$Z=Z_1^{\ast },Z_2^{\ast }$}. The projection values derived from the SWSSA-PP model in this study are continuous data after high-dimensional data reduction, and the numerical relationship of “higher risk corresponds to larger projection value” is realized by these values. Based on the numerical distribution of the projection values, objective and automated risk level classification is enabled by K-means clustering. This method is adapted to the mixed dataset of standard and evaluation samples in this study, by which the stability of risk level classification is guaranteed. The steps are as follows.

1. Select one projection value at random from the dataset as the first initial center $c_1$.

2. Calculate the Euclidean distance between the remaining $Z\left(i\right)$ and $c_1$, and then select the second center $c_2$ using the following formula:

$$P\left(Z_i\right)={\displaystyle \frac{{d\left(i\right)}^2}{{\sum }_{j=1}^n{d\left(i\right)}^2}},$$

(24)

where $d\left(i\right)=min\left(\left|Z\left(i\right)-c_1\right|,\left|Z\left(i\right)-c_2\right|\right)$.

3. Repeat step 2 to calculate, in turn, the minimum distance between the projection values and the selected centers $c_1$ and $c_2$. Select $c_3$, $c_4$, and $c_5$ by calculating $P\left(Z_i\right)$, ultimately obtaining the initial center set $\{c_1,c_2,c_3,c_4,c_5\}$.

(2)

Iterative clustering and convergence testing

Calculate the distance between each projection value and the five cluster centers, and then assign $Z\left(i\right)$ to the cluster $G_j$ with the smallest distance, i.e., $Z\left(i\right)\in G_j$. Based on the samples within each cluster, calculate a new cluster center using the following formula:

$$c_j={\displaystyle \frac{1}{n_j}}{\sum }_{Z\left(i\right)\in G_j}Z\left(i\right),$$

(25)

where $n_j$ denotes the number of samples in the jth cluster.

Calculate the maximum change in cluster centers before and after the update $\Delta c$. If $\Delta c<{10}^{-5}$ or the iteration count reaches 100, terminate the iteration and output the cluster centers and the final cluster partition.

(3)

Risk level mapping and classification

Sort the clusters by ascending order of cluster centers to correspond to five risk levels, as detailed in

Table 3. Risk Level Correspondence to Clustering Clusters.

Cluster Group	Risk Level	Projection Value Range
$c_{\left(1\right)}^{\ast }$	I	$\left[min\left(G_1\right),max\left(G_1\right)\right]$
$c_{\left(2\right)}^{\ast }$	II	$\left[min\left(G_2\right),max\left(G_2\right)\right]$
$c_{\left(3\right)}^{\ast }$	III	$\left[min\left(G_3\right),max\left(G_3\right)\right]$
$c_{\left(4\right)}^{\ast }$	IV	$\left[min\left(G_4\right),max\left(G_4\right)\right]$
$c_{\left(5\right)}^{\ast }$	V	$\left[min\left(G_5\right),max\left(G_5\right)\right]$

.

Substitute the evaluation sample projection value $Z_2^{\ast }$ into the clustering model one by one, and then calculate the distance between $Z_2^{\ast }$ and the final $c_j^{\ast }$. The risk level of each sample is then determined according to the risk value intervals corresponding to each risk level.

To further validate the applicability of K-means clustering, it was compared with the natural breakpoint method and fuzzy C-means clustering, which is commonly used in ecological environment assessment. The ratio of the sum of squares within clusters to the sum of squares between clusters (SSE/SST) and the contour coefficient were employed as quantitative evaluation metrics. A lower SSE/SST indicates higher intra-cluster cohesion and greater inter-cluster discrimination, while a contour coefficient closer to 1 signifies superior clustering performance. Detailed validation procedures are presented in Section 2.5.2 on risk level calculation.

2.5. Case Study

2.5.1. Data Collection

Data for $D_1$, $D_2$, $D_3$, $D_5$, $I_2$, $R_1$, $R_2$, and $R_3$ can be obtained from the Hefei Statistical Yearbook (2015–2024), published by the Hefei Municipal Bureau of Statistics. $D_4$, $P_3$, and $I_3$ were sourced from the Hefei Water Resources Bulletin (2015–2024), issued by the Hefei Municipal Water Resources Bureau. $P_1$, $S_1$, $S_3$, and $S_4$ can be accessed from the Hefei Environmental Status Bulletin (2015–2024), released by the Hefei Municipal Bureau of Ecology and Environment. $S_2$, $I_1$, $I_4$, and $P_5$ were acquired through field investigations conducted by the project team. $P_4$ and $R_6$ were derived from the land use status survey data of the Hefei Bureau of Territorial Spatial Planning and the annual statistical bulletin of the garden and greening administration department, respectively. $P_2$ and $R_5$ were primarily determined via expert scoring. Based on these data sources, this formed the evaluation sample set $X_1={\left[X_{ij}\right]}_{10\times 24}$, as detailed in

Table 4. Ecological risk assessment indicator data for the Chuhe Main Canal, 2015–2024.

Indicator	2015	2016	2017	2018	2019	2020	2021	2022	2023	2024
$D_1$	650	680	720	780	820	810	802	850	880	910
$D_2$	9.4	9.0	9.1	9.2	9.7	3.8	9.9	5.9	6.0	6.67
$D_3$	5.11	6.26	7.22	6.02	5.29	1.2	0.5	0.1	0.06	0.04
$D_4$	983	980	976	980	981	975	870	536.6	870	870
$D_5$	51.88	53.28	54.68	53.0	57.42	61.84	62.47	63.1	63.85	64.93
$P_1$	0.016	0.014	0.012	0.007	0.006	0.005	0.004	0.004	0.003	0.002
$P_2$	0.001	0.001	0.001	0.001	0.001	0.001	0.001	0.001	0.001	0.001
$P_3$	55.13	55.1	55.24	55.31	55.15	55.23	55.3	55.26	55.38	55.59
$P_4$	4.203	4.611	4.921	5.134	5.412	5.633	7.001	7.164	7.421	11.31
$P_5$	0.11	0.12	0.131	0.154	0.151	0.167	0.165	0.177	0.174	0.178
$S_1$	86	88	89	92	91	84	83	82	86	82
$S_2$	0.98	0.91	0.88	0.86	0.87	0.84	0.83	0.84	0.82	0.81
$S_3$	91	88	86	81	78	77	75	76	71	68
$S_4$	90.3	88.6	90.1	91.6	84.6	76.4	81.3	77.4	76.1	88.1
$I_1$	6.1	7.2	6.8	9.4	12.6	16.4	20.1	23.1	26.1	28.4
$I_2$	37.6	36.8	36.4	36.1	35.9	34.1	30.9	30.4	30.6	30.4
$I_3$	86	84	82	80	76	77	75	74	72	71
$I_4$	80	78	77	76	74	70	68	66	62	61
$R_1$	1.62	1.63	1.66	1.61	1.7	1.68	1.66	1.68	1.69	1.72
$R_2$	42.1	42.6	42.9	43.1	43.6	43.7	43.9	44.1	44.6	45.8
$R_3$	87.1	87.2	87.6	87.9	88.1	88.6	89.2	89.6	90.4	91.3
$R_4$	74.1	76.8	78.1	79.6	80.6	84.6	86.3	89.3	91.4	95.6
$R_5$	61	66	68	72	74	82	86	87	90	92
$R_6$	45.6	46.1	46.3	46.7	46.5	47.1	47.6	47.9	48.1	48.6

.

Reliability and validity tests were conducted on 24 indicator data points for the Chuhe Main Canal from 2015 to 2024. The results are shown in

Table 5. Reliability and validity test results for the indicator system.

Test Item	Result Value	Judgment Criteria	Conclusion
Cronbach’s α	0.872	>0.8	Good reliability
KMO value	0.791	>0.7	Good validity
Bartlett test	<0.001	$p<0.05$	Significant correlation between indicators

.

As shown in the table above, Cronbach’s α is 0.872, indicating high internal consistency among the indicators. The KMO value is 0.791, and Bartlett’s test is significant, suggesting that the indicator system is suitable for risk assessment modeling.

2.5.2. Risk Level Calculation

(1)

Data normalization processing

For each of the five risk levels, form 200 sets of standard sample data $X_2$, respectively. Combine the standard sample sets and evaluation sample sets to obtain the computational sample set $X=\left\{X_1,X_2\right\}$. Perform normalization processing according to Equations (9) and (10).

(2)

Optimization of the projection indicator function and projection direction construction

Establish the projection indicator function based on computational datasets (16). Initialize key parameters according to the SWSSA-PP model parameter settings. See

Table 6. Key parameters table.

Parameter	Value	Parameter	Value
Population size $N$	40	Whale Algorithm convergence factor $a$	$2-2t/T_{max}$
Maximum iteration count $T_{max}$	500	Seagull Algorithm control parameter $u$	1
Producer proportion $N_{PD}$	0.2	Seagull Algorithm perturbation coefficient $v$	0.1
Safety threshold $ST$	0.8	Lévy stability $\beta$	1.5

for details.

The iterative process is as follows: Each position represents a potential projection direction vector. Forty 24-dimensional projection direction vectors are randomly generated and normalized. Substitute each sparrow position into Equation (16) to calculate the fitness $Q_a$ for each individual, selecting the initial globally optimal projection direction and fitness. Update the positions of producers, followers, and early-warning sparrows using Equations (18), (22) and (23). Update the global optimum solution after each iteration. Repeat the calculation of $Q_a$ and population position updates. Terminate optimization when the change in $Q_{best}$ is less than ${10}^{-5}$ for 10 consecutive iterations or the maximum iteration count is reached, and then output $a^{\ast }$.

(3)

Indicator weight calculation

Using MATLAB 2024 software, the optimal projection direction $a^{\ast }$ = (0.2268, 0.2015, 0.1995, 0.2685, 0.1426, 0.2219, 0.2711, 0.3452, 0.03134, 0.1536, 0.2669, 0.3642, 0.1395, 0.2700, 0.2006, 0.2796, 0.2113, 0.3368, 0.2521, 0.2063, 0.3264, 0.2106, 0.2286, 0.2103). Squaring each element in $a^{\ast }$ and then normalizing yields the weights for each indicator, as detailed in

Table 7. Weight calculation results.

Indicator	Weight	Indicator	Weight	Indicator	Weight	Indicator	Weight
$D_1$	0.042	$P_2$	0.045	$S_3$	0.027	$R_1$	0.041
$D_2$	0.035	$P_3$	0.054	$S_4$	0.044	$R_2$	0.037
$D_3$	0.0363	$P_4$	0.052	$I_1$	0.0353	$R_3$	0.051
$D_4$	0.043	$P_5$	0.043	$I_2$	0.047	$R_4$	0.036
$D_5$	0.033	$S_1$	0.044	$I_3$	0.0375	$R_5$	0.039
$P_1$	0.0383	$S_2$	0.059	$I_4$	0.056	$R_6$	0.0376

. The internal weight distribution of each factor is shown in Figure 5.

(4)

Determine the projected value and establish the risk level

The adjusted risk assessment grading criteria calculated using this research methodology are shown in Figure 6.

The risk value ranges for each risk level are detailed in

Table 8. Classification of ecological and environmental risk levels for water diversion projects.

Risk Level	Interval Value
I	$\left[1.3510,1.7395\right]$
II	$\left[1.7395,2.0732\right]$
III	$\left[2.0732,2.3963\right]$
IV	$\left[2.3963,2.5841\right]$
V	$\left[2.5841,2.8493\right]$

.

At the same time, we can obtain the distribution diagram of the projected values for the standard sample and the evaluation sample, as shown in Figure 7.

Substituting the optimal projection direction and respective indicator weights into the projection indicator function yielded the projected values for the ecological and environmental risk assessment of the Chuhe Main Canal. Based on the DPSIR framework, trend charts for the comprehensive ecological and environmental risk assessment of the Chuhe Main Canal and trend charts for each primary indicator were plotted for each year, as shown in Figure 8 and Figure 9.

Based on the figure above and Table 8, the ecological and environmental risk of the Chuhe Main Canal is observed to show an overall upward trend, with relatively low risk being observed in 2015 and higher risk in 2024. It is indicated that the ecological and environmental risk of the Chuhe Main Canal increases with each passing year. The model calculations are found to be in close alignment with the objective realities reflected in the Hefei Environmental Status Bulletin and Hefei Water Resources Bulletin for the same periods. Located within the Yangtze River Delta urban cluster and subject to intense human disturbance, rapid urbanization, sustained growth in water demand, and persistently high water resource utilization rates were experienced by the Chuhe Main Canal during the study period. Enhanced socioeconomic drivers and increased ecological pressures were identified as contributors to the upward risk trajectory, further validating the model’s effectiveness in identifying specific engineering risk characteristics.

To test the robustness of the weighting results, sensitivity analyses in fields such as water resources and ecological environment commonly employ perturbation ranges of ±5% and ±10%. Among these, ±10% represents a moderate perturbation—sufficient to reflect the impact of weight changes without causing excessive deviation from actual weight estimates, aligning with sensitivity analysis applications [53]. We applied ±10% weight perturbations to the top three ranked indicators and observed their impact on the risk levels for 2015–2024. The results are presented in

Table 9. Weight sensitivity analysis results.

Year	Original Result	${\mathit{P}}_3+10\mathbf{\%}$	${\mathit{P}}_3-10\mathbf{\%}$	${\mathit{S}}_2+10\mathbf{\%}$	${\mathit{S}}_2-10\mathbf{\%}$	${\mathit{I}}_4+10\mathbf{\%}$	${\mathit{I}}_4-10\mathbf{\%}$
2015	I	I	I	I	I	I	I
2016	II	II	II	II	II	II	II
2017	III	III	III	III	III	III	III
2018	IV	IV	IV	IV	IV	IV	IV
2019	IV	IV	IV	IV	IV	IV	IV
2020	IV	IV	IV	IV	IV	IV	IV
2021	IV	IV	IV	IV	IV	IV	IV
2022	IV	IV	IV	IV	IV	IV	IV
2023	IV	IV	IV	IV	IV	IV	IV
2024	IV	IV	IV	IV	IV	IV	IV

.

As shown in the table above, the risk levels remained unchanged across all years, indicating that the weights derived in this study exhibit strong robustness, with evaluation results being minimally affected by weight fluctuations.

Using the dataset from this study as the research subject, a comparison was conducted among five-level risk classification approaches—K-means, the natural breakpoint method, and fuzzy C-means clustering—and these were compared with the risk level classification results for the 2015–2024 evaluation samples of the Chuhe Main Canal. The detailed results are presented in

Table 10. Comparison of quantitative metrics across different classification methods.

Method	SSE/SST	Coefficient of Contour	Computational Efficiency	Result Format	Subjectivity
K-Means Clustering	0.124	0.896	Fast	Clear grading	Low
Natural Break Method	0.368	0.621	Fast	Clear grading	Moderate
Fuzzy C-Means Clustering	0.157	0.823	Slow	Fuzzy membership matrix	High

and

Table 11. Comparison of risk level determination results for the Chuhe Main Canal using different classification methods.

Time	Actual Operating Conditions	K-Means Clustering	Natural Break Method	Fuzzy C-Means Clustering
2015	I	I	I	I
2016	II	II	II	II
2017	III	III	II	III
2018	IV	IV	III	IV
2019	IV	IV	IV	IV
2020	IV	IV	IV	IV
2021	IV	IV	IV	IV
2022	IV	IV	IV	IV
2023	IV	IV	IV	IV
2024	IV	IV	IV	IV

.

As shown in the table above, the optimal quantitative metrics are achieved by K-means clustering, featuring the lowest SSE/SST and the highest silhouette coefficient. It is indicated that its clustering results enable clearer delineation of the projection value ranges across different risk levels, with greater practicality, high computational efficiency, and unambiguous outcomes being demonstrated. With minimal subjective interference, the results are rendered more objective and are found to be closely aligned with engineering realities.

3. Analysis and Discussion of Results

3.1. Analysis of Indicator Weighting Results

In this study, the SWSSA-PP model performs the computation automatically in a data-driven manner, by which the weights of all indicators are derived, subjective intervention is reduced, and the objectivity of the results is ensured. The three highest-weighted indicators are found to be “water resources development and utilization rate”, “biodiversity index”, and “public satisfaction”, with values of 0.054, 0.059, and 0.056, respectively.

The Chuhe Main Canal connects the Yangtze River and the Chuhe River basins, providing functions such as agricultural irrigation, urban water supply, and flood control. Farmland, specialized agricultural zones, industrial areas, and densely populated urban districts are distributed along the canal. Diversified water demands create a rigid reliance on water resources, and significant pressure is imposed on sustaining the supply–demand balance [54]. However, as shown in Table 4, rapid urbanization has accelerated in this study area in recent years. The urbanization rate increased from 51.88% to 64.93% between 2015 and 2024, accompanied by sustained GDP growth and a sharp increase in water demand. The persistently medium-to-high level of the water resources development and utilization rate indicated that the proportion of water withdrawn from the river channel to the natural runoff is excessively large. As shown in Table 2, this value falls under Level III risk. An excessively high water resources development and utilization rate is found to directly result in insufficient ecological baseflow, weakening of the water body’s self-purification capacity, and aggravation of upstream–downstream water use conflicts [55]. This aligns with the actual conditions faced by the Chuhe Main Canal in recent years, including seasonal water shortages and significant pressure to ensure ecological flow. Simultaneously, this indicator influences metrics across four dimensions: driving forces, pressures, state, and impacts. While water resource utilization exhibits a gradual upward trend, aquatic biodiversity loss rates have surged from 6.1% to 28.4%, and public satisfaction has declined from 80% to 61%. Significant correlations have been observed between changes in water resource utilization and both ecological deterioration and social impacts, making this indicator the highest-weighted factor. The biodiversity index directly reflects the integrity of ecosystem structure and function. Drinking water source protection zones and bird habitats are situated along the Chuhe Main Canal. Biodiversity is considered a direct indicator of the project’s disturbance level to regional ecology, making it a key focus for ecological conservation. The artificial modification of main irrigation channels often results in the alteration of aquatic habitats. As water resource development intensifies and land use changes, the risk of habitat fragmentation along these channels increases, leading to a decline in biodiversity indices and directly exacerbating ecosystem vulnerability. Along the Chuhe Main Canal, increased agricultural nonpoint source pollution, coupled with altered hydrodynamic conditions, can readily cause biodiversity decline, leading to ecosystem fragility and susceptibility to algal blooms. The significant weight assigned to this indicator suggests that canalization has led to reduced aquatic vegetation, impeded fish migration, or simplified benthic community structure. Public satisfaction serves as a direct reflection of how ecological risks are transmitted to the societal level, comprehensively indicating the project’s impact on residents’ livelihoods, production activities, and the surrounding environment. As both a water conveyance channel and a vital landscape corridor and waterfront space along its route, the water quality, scenic appeal, and reliability of irrigation water supply of the Chuhe Main Canal directly influence the immediate perceptions of riverside residents and farmers. The high weighting of this indicator points to a gap between public expectations and current management practices under actual operating conditions. For example, even if water quality monitoring data meet standards, perceived turbidity, odors, or short-term scheduling implemented to ensure upstream urban water supply, resulting in insufficient downstream irrigation water during the irrigation season, can all negatively impact public satisfaction.

Key risk factors were identified, and the risk pathways—wherein indicators across DPSIR dimensions are mutually coupled and propagate sequentially—were mapped in this case study. For instance, the “urbanization rate” indicator under the driving force dimension increased from 51.88% to 64.93%, resulting in the “water resource utilization rate” being maintained at a persistently high level and directly intensifying pressure. Furthermore, a significant decline in the biodiversity index was observed due to the continuous accumulation of pressure, leading to the worsening of the ecological condition. This degradation resulted in low public satisfaction, which in turn prompted enhanced response measures, such as improved centralized sewage treatment rates, to be implemented to mitigate the adverse impacts. However, while some pressures can be mitigated by response measures, the new risks arising from the growth of driving forces cannot be fully offset. As clearly shown in Figure 9, risks remained at a persistently high level after 2018. This further confirms that ecological and environmental risks do not exist in isolation but represent a process of dynamic coupling and hierarchical transmission across dimensions, necessitating coordinated management throughout the entire DPSIR process.

By integrating the trends of each indicator from 2015 to 2024 with the corresponding risk level evolution, the quantitative priority order for risk management is established as follows. The water resources development and utilization rate is prioritized first, given its high weight, prolonged Level III risk status, and continuous upward trend. To address this, total water withdrawal is required to be strictly limited, agricultural water-saving irrigation technologies are to be promoted, water resource allocation is to be optimized, and ecological baseflow is to be preferentially safeguarded, so that the utilization rate is gradually reduced to an appropriate threshold. The biodiversity index, having the highest weight and a persistent declining trend, is designated as the second priority, for which degradation is to be prevented by constructing ecological buffer zones and restoring habitats. Public satisfaction, ranking second in weight and showing a year-by-year decrease, is identified as the third priority, necessitating the establishment of a public feedback mechanism and substantial optimization of water supply scheduling. This priority ranking offers a quantitative foundation for differentiated risk management. In the assessment of water resources carrying capacity, priority regulation of the water resources development and utilization rate has been proven to be a key pathway for alleviating the supply–demand contradiction of water resources [56].

3.2. Risk Level Result Analysis

The period from 2015 to 2018 was characterized by medium-to-low risk. During this phase, vigorous ecological conservation efforts were emphasized in the 13th Five-Year Plan, with water resource allocation and pollution prevention measures being implemented concurrently. The industrial structure was primarily characterized by traditional agriculture and low-pollution industries. Driving factors and pressure factors remained at low levels under the guidance of industrial structure optimization policies. The state factors maintained low risk due to advancing wetland conservation along the route. Impact factors also remained stable. Among response factors, the “urban sewage centralized treatment rate” was significantly improved, while the “environmental investment as a percentage of GDP” grew steadily, effectively mitigating potential risks. The period from 2018 to 2024 was characterized by higher risk. A critical juncture in the reform of China’s ecological and environmental protection management system was marked by the year 2018, when the Ministry of Ecology and Environment was formally established. Although regulatory oversight was strengthened institutionally, inadequate environmental infrastructure and insufficient pollution source control persisted in some areas along the canal in the short term. Concurrently, regional development planning was reoriented toward the coordinated advancement of urbanization and industrialization. As Hefei accelerated its urban expansion under the Yangtze River Delta urban agglomeration initiative, the area along the Chuhe Main Canal was designated as a core zone for urban construction and industrial park layout, serving as a key development belt in northern Hefei. During this period, urbanization was significantly accelerated, and a rapid increase in domestic water demand was driven by the sharp growth of the urban population. Low-pollution traditional industries along the canal were gradually transformed into urban-oriented industries, and the scale of industrial parks was expanded. Despite the predominance of low-polluting industries, a sharp increase in water resources development intensity and a substantial rise in industrial wastewater discharge were observed. Furthermore, while cross-regional collaborative governance mechanisms had been strengthened, initial response efficiency remained inadequate. Issues such as delays in handling pollution incidents and unscientific decision-making further amplified risks. As shown in Table 4, between 2015 and 2024, the urbanization rate increased from 51.88% to 64.93%, the water resources development and utilization rate remained above 55% for a prolonged period, the biodiversity index decreased from 0.98 to 0.81, and public satisfaction dropped from 80% to 61%. The actual deterioration of these key indicators directly drove the elevation of the risk level. Furthermore, as demonstrated in Table 9, after a ±10% perturbation was applied to the top three weighted indicators, the risk level for each year remained unchanged, which further confirms the objectivity of the evaluation results.

In response to the ongoing increase in ecological and environmental risks along the Chuhe Main Canal from 2015 to 2024, the following improvements can be implemented in conjunction with indicators significantly impacted by ecological risks and project realities: First, by strictly controlling the exploitation and utilization of water resources, introducing and scaling up water-saving irrigation techniques, and optimizing water allocation to ease the supply–demand imbalance, it is expected that the water resources development and utilization rate will be progressively lowered from over 55% to under 50%. Ecological protection should be strengthened, particularly in nature reserves, with ecological buffer zones being established along riverbanks to reduce soil erosion. By establishing monitoring stations for biodiversity, biological losses resulting from sudden severe pollution accidents can be alleviated through artificial breeding and restocking. Should this measure be effectively implemented, the “biodiversity index” is expected to stop declining and start to rise again. Additionally, in the field of pollution prevention and control, regulations on industrial wastewater discharge compliance rates and centralized sewage treatment rates should be strictly enforced. Vessel navigation management should be strengthened to reduce oil spill risks. If the proportions of cropland and construction land are reasonably controlled, the urbanization rate is appropriately slowed, the tertiary industry is strengthened, stable and continuous environmental investment is ensured, and a public feedback platform is established to further enhance aquatic species protection, then “public satisfaction” will be improved accordingly. Through the effective implementation of these control measures, the ecological risk of the Chuhe Main Canal will not follow a sustained upward trajectory; instead, a period of high-level fluctuation may be followed by a gradual decline.

3.3. Advantages of the Proposed Method

Regarding the indicator system, Richter et al. [3] focused solely on quantifying ecological disturbance risks through hydrological rhythm variations, while Rahel et al. [57] concentrated on the single risk type of biological invasions. A multidimensional integrated analysis was not developed in either study. In this research, the DPSIR model is employed to construct an indicator system. By incorporating core socioeconomic indicators such as regional population density and urbanization rates into the driving force dimension, the root risks underlying water resource supply–demand conflicts are precisely identified. The pressure dimension encompasses indicators such as ${SO}_2$ emission intensity and soil erosion index, quantifying direct ecological stress imposed by project operations. The response dimension integrates metrics such as centralized wastewater treatment rates and aquatic biodiversity protection measures, aligning theoretically with Yang et al.’s [58] emphasis on “ecological protection effectiveness” to evaluate the implementation outcomes of risk control measures. By integrating the complexity and dynamic interconnections of the ecological–environmental system in water diversion projects, the limitations of analyzing isolated risk factors are overcome by this model. Both the spatial distribution of static risks and the dynamic evolution of risks driven by project operations and environmental changes are reflected, thereby achieving better alignment with practical realities. When facing sudden pollution incidents or ecosystem disturbances, targeted adjustments to high–weight indicators and optimized response strategies can be implemented to reduce the system’s ecological–environmental vulnerability, ensuring the long-term stable operation of the project.

Regarding indicator weight calculations, previous ecological and environmental risk assessment studies have frequently employed the Analytic Hierarchy Process (AHP). However, this method is found to be inadequate for quantifying subtle differences between “water resource utilization rates” and “biodiversity indices,” and it is considered insufficient for adapting to risk fluctuations caused by sudden pollution incidents or extreme weather events. Li [59] employed a combined weighting approach using the “AHP–entropy weighting method,” balancing subjective experience with objective data. However, the complex nonlinear relationship between “urbanization rate” and “proportion of farmland and construction land” cannot be effectively addressed by this method. Furthermore, inherent biases in subjective weights within combined weighting methods can affect final outcomes through weight fusion. For instance, when handling small-sample data on “probability of accidental oil spills from vessels,” weight fluctuations are caused by insufficient data dispersion, reducing result reliability. In this study, an improved projection pursuit model is employed to calculate indicator weights, projecting 24-dimensional indicator data into a low-dimensional space. This approach ensures that data structural characteristics are preserved while subjective influence is reduced, enabling precise capture of indicators’ dynamic impact on risk. The top three weighted indicators are identified as “biodiversity index,” “water resource utilization rate,” and “public satisfaction.” These results are consistent with the actual scenario along the Chuhe Main Canal, where drinking water source protection zones are distributed and populations are densely concentrated, resulting in a more reasonable and interpretable weight distribution.

Regarding evaluation models, the Genetic Algorithm-Based Projection Pursuit (GA-PP) model, the traditional PP model, and the SSA-PP model were selected for comparative analysis using the research data in this study. The results are presented in

Table 12. Risk level prediction results for each research method.

Research Methods	2015	2016	2017	2018	2019	2020	2021	2022	2023	2024
SWSSA-PP	I	II	III	IV	IV	IV	IV	IV	IV	IV
GA-PP	I	II	II	III	IV	IV	IV	IV	IV	IV
SSA-PP	I	II	II	III	IV	IV	IV	IV	IV	IV
PP	I	I	II	II	II	III	III	IV	IV	IV
Actual situation	I	II	III	IV	IV	IV	IV	IV	IV	IV

. The iteration convergence curves for the four models are shown in Figure 10.

As shown in Table 12, for the 10 datasets, the assessment results obtained using the model developed in this study were found to be consistent with the actual ecological and environmental risk conditions. In contrast, the accuracy rates for the GA-PP and SSA-PP models were both 80%, while the PP model achieved only 40% accuracy. This demonstrates that the methodology proposed in this study is accurate and reliable for evaluating the ecological and environmental risks associated with water diversion projects.

As can be observed from the iteration diagram, significantly superior optimization performance is achieved by the SWSSA-PP model compared to other algorithms. The fastest convergence speed is exhibited by this model, with the global optimum being reached near 50 iterations. Low parameter dependency is demonstrated by the model, effectively avoiding convergence to local optima, and the highest accuracy is delivered. Objective optimization is achieved by the GA-PP model, but its global exploration capability is limited and it is prone to premature convergence, with the global optimum being reached after approximately 200 iterations and accuracy being slightly lower than that of the SWSSA-PP model. Faster convergence than GA-PP is observed for the SSA-PP model, but it is characterized by insufficient population diversity, being prone to convergence at local optima and inadequate computational accuracy, with convergence being reached near 160 iterations. No optimization function is incorporated into the traditional PP model, and thus no optimization capability is possessed.

Water diversion projects encompass multidimensional indicators spanning socioeconomic and natural ecological dimensions, featuring high-dimensional data with complex nonlinear correlations. While dimensionality reduction can be achieved by the PP model, the risk of losing structural data features is posed by its lack of optimization algorithms. Global search capability is enhanced by GA-PP, but slow convergence is observed, leading to imprecise identification of nonlinear relationships between indicators. Optimization performance is effectively improved by the SSA-PP model, but susceptibility to population diversity remains. By integrating the Whale Algorithm’s enclosure mechanism with the seagull algorithm’s spiral search strategy, the complex nonlinear relationships among indicators in high-dimensional spaces are more effectively captured by the SWSSA-PP model. Variations in ecological sensitivity are observed across different sections along water diversion routes. As urbanization rates rise and water demand increases, precise capture of dynamic risk changes by evaluation models is required to avoid inaccurate risk predictions caused by slow algorithm convergence or local optima. Adaptive weights and a Lévy flight strategy are introduced by the SWSSA-PP model to broaden solution space exploration in early iterations while enabling refined convergence later. Premature termination is prevented by its optimized convergence factor design, with global optimality being achieved after approximately 50 iterations. Significant outperformance of other models in convergence speed is demonstrated, while local optima are effectively avoided.

Ecological and environmental risks are interconnected and interact, forming complex risk transmission mechanisms. Identification of key risk indicators and formulation of corresponding measures are required for water diversion project risk management. While indicator weights can be calculated by the GA-PP and SSA-PP models, adequate identification of critical risks is not achieved. For instance, the weight calculated by the GA-PP model for the water resource utilization rate is only 0.041, which is far below actual conditions. Differences in indicator weights cannot be effectively distinguished by the PP model. Projection directions are optimized by the SWSSA-PP model through enhancement of the Sparrow Search Algorithm. While data structure is preserved, projection direction vectors are utilized to visually reflect each indicator’s impact on overall risk, enabling precise weight calculations. In summary, superior predictive accuracy and convergence efficiency are demonstrated by the SWSSA-PP model. Its adaptability to high-dimensional nonlinear data, mechanisms for capturing dynamic risk evolution, enhanced interpretability, and robustness render it more suitable for ecological and environmental risk assessment in water diversion projects.

The DPSIR-SWSSA-PP model developed in this study is recognized as representing not only theoretical innovation but also practical application value in engineering practice. The proposed model addresses the problems of fragmented indicator systems, insufficient capability for processing high-dimensional nonlinear data, and strong subjective interference in traditional risk assessment. By this model, key influencing factors of ecological and environmental risks in inter-basin water diversion projects, such as “water resources development and utilization rate” and “biodiversity index”, can be systematically identified, thereby providing managers with clear directions for risk control. In addition, although the model improves assessment accuracy, a certain increase in model complexity is inevitably introduced. The physical interpretation of the projection direction vector is less straightforward than that of the original indicators, and the enhanced algorithm adds extra parameters and computational steps. However, from a policy formulation standpoint, the model exhibits reasonable interpretability, as detailed below. After squaring and normalization of each component of the optimal projection direction, the resulting values serve as indicator weights, where a larger weight reflects a greater contribution to the overall risk, allowing key risk factors to be effectively identified. The projection values are classified via K-means clustering, which avoids subjective classification boundaries, and the resulting clusters are shown to align with actual engineering risk levels. Furthermore, by combining the DPSIR framework with the projection values, the progressive transmission of risk from “driving force” to “response” can be distinctly demonstrated. In practical management, the model can be integrated with a monitoring platform, by which data are regularly updated on a quarterly or monthly basis, and the SWSSA-PP calculation is automatically performed and compared against the risk level intervals. Once the risk level reaches Level IV or above, an early warning is triggered by the system, the top three weighted indicators are automatically identified, and differentiated management suggestions are then formulated in a timely manner according to regional characteristics. In response to risk events, including sudden pollution and climate change, a higher data update frequency is required so that abrupt indicator changes are reflected in the projection values in a timely manner. Warning thresholds are established for key indicators based on historical variation magnitudes under extreme events. Once real-time monitoring data exceed these thresholds, an independent early warning is triggered, enabling effective capture of the disturbance effects of sudden risks and providing a quantitative foundation for water resources scheduling, ecological conservation, and pollution control.

3.4. Risk Management Strategy

A quantitative assessment of the ecological and environmental risks of the Chuhe Main Canal was conducted in this study using the DPSIR-SWSSA-PP model, and key risk factors and their evolution trends were identified. To develop more targeted management strategies, the Middle Route of the South-to-North Water Diversion Project and the Hanjiang-to-Weihe River Water Diversion Project were selected as typical cases. A qualitative descriptive comparative analysis was performed between these projects and the case study of this research, based on the published academic literature and engineering operation reports. This analysis is not to be equated with the quantitative evaluation results of the main case, and is primarily employed as a horizontal reference for management strategy development. The details are presented in

Table 13. Comparison of ecological and environmental risk characteristics between the Chuhe Main Canal and typical water diversion projects.

	This Study	South-to-North Water Diversion Project	Diverting Han River Water to Weihe River
Project characteristics	Multi-functional, regional small-to-medium scale	Pure water supply; national-level, large-scale project	Provincial large scale, complex geological areas
Key factors	Water resource utilization rate, biodiversity	Water supply security rate, effectiveness of cross-regional collaborative governance	Geological safety, soil erosion
Primary risks	Socioeconomic-driven ecological degradation	Supply–demand imbalances and pollution risks	Ecological risks triggered by engineering projects
Risk evolution	Levels I to IV show a continuous increase	Level II remains stable	Construction phase IV, operation phase III
Management strategy	Limit total water withdrawal, establish ecological buffer zones	Implement river chief system, cross-provincial water quality compensation	Real-time environmental monitoring during construction period

.

As shown in the table above, the South-to-North Water Diversion Project is characterized by large-scale operations and complex inter-provincial coordination, with greater emphasis being placed on water supply reliability and cross-regional governance efficiency. Cross-regional collaborative governance is adopted as its primary management strategy. The Han-to-Wei Water Diversion Project, due to its passage through ecologically sensitive areas, is subject to ecological disturbances during construction and altered hydrological conditions during operation. Key attention is directed toward construction-related disturbances, with management prioritizing project safety and in situ protection of ecologically sensitive zones through the implementation of avoidance measures and real-time monitoring of geological hazards.

Given the sustained increase in ecological and environmental risks along the Chuhe Main Canal from 2015 to 2024, the DPSIR transmission mechanism and key indicators are integrated in this study. Drawing on large-scale management experiences such as ecological compensation from the South-to-North Water Diversion Project and ecological monitoring from the Han-to-Wei Water Diversion Project, a phased, regionally tailored, and operationally feasible risk management strategy is proposed to mitigate ecological and environmental risks. The specific measures are outlined as follows:

(1)

Phased Risk Management Strategy

Based on the evolution characteristics of risk levels, risk management for the Chuhe Main Canal is divided into short-term and medium-to-long-term phases. Short-term stage: the sustained rapid escalation of risk is required to be halted, and the risk level is to be maintained within Level IV. The primary tasks involve the implementation of a dynamic permitting mechanism for the water resources development and utilization rate, as well as the imposition of total withdrawal caps in areas where over-extraction occurs. Additionally, enhanced targeted management of pollution sources is required. For response indicators such as “centralized sewage treatment rate” and “industrial wastewater compliance rate,” it is required that industrial parks along the channel achieve stable water quality compliance at key monitoring sections within three years. Concurrently, ecological monitoring stations should be established in sections with significant biodiversity decline to ensure timely detection of and effective response to sudden pollution incidents. For the medium-to-long term, the goal is to gradually control risk levels below Level III. Initial steps include optimizing regional industrial and water usage structures, guiding areas along the canal toward low-water-consumption and low-pollution industries, and striving to keep water resource utilization rates below 50% within a decade. Subsequently, ecological buffer zones of at least 100 m should be established on both sides of the canal to enhance green coverage and improve connectivity between soil, water, and biological habitats. Finally, drawing on the experience of the “river chief system,” a “channel chief system” should be established along the Chuhe Main Canal to achieve coordinated implementation of water resource management and pollution prevention.

(2)

Regional Control Strategy

Due to significant variations in topography, land use, and ecological sensitivity along the Chuhe Main Canal, zoned management is implemented based on regional types and key indicators. In densely populated areas with high water demand and concentrated pollution discharge, priority is given to the control of the indicators $P_3$, $I_4$, and $R_3$. This is achieved by promoting water-efficient communities and water recycling technologies, strictly controlling urban sewage, and safeguarding the ecological environment. Water source protection zones are characterized by high water quality sensitivity and biodiversity, necessitating stringent management of the indicators $S_1$, $S_2$, and $P_2$. Risks can be mitigated by prohibiting vessel traffic and reducing oil spill risks, thereby preventing ecological hazards. Ecologically sensitive zones are subject to habitat degradation and poor biological connectivity, requiring focused management of the indicators $S_2$, $R_2$, and $R_6$. Environmental risks can be lowered by restricting infrastructure development and implementing ecological corridor projects. Agricultural irrigation zones are affected by severe soil erosion, necessitating focused management of the indicators $P_4$, $P_5$, and $I_1$. Ecological pollution can be reduced by promoting ecological ditch pollution interception technology alongside drip irrigation and micro-irrigation water-saving techniques. Through these zoned management measures, the ecological environment of the Chuhe Main Canal is more effectively managed.

4. Conclusions

In this study, a comprehensive evaluation model designated as DPSIR-SWSSA-PP was developed and subsequently applied to the ecological risk assessment of the Chuhe Main Canal over the period 2015–2024. The main findings are summarized as follows.

(1) A five-tier indicator system comprising five primary indicators and 24 secondary indicators was constructed based on the DPSIR framework. Indicators that closely reflect project-specific realities, such as “probability of ship-caused oil spill accidents” and “completeness of aquatic organism protection measures”, were incorporated. By this means, the limitations of traditional single-dimensional risk assessment were overcome.

(2) The SWSSA-PP model was proposed, which exhibits faster convergence speed and higher optimization accuracy when processing high-dimensional nonlinear data, and significantly outperforms GA-PP and PP. As revealed by the weight results, “water resources development and utilization rate”, “biodiversity index”, and “public satisfaction” are identified as the top three key indicators.

(3) Empirical evidence indicates that the ecological risk rose progressively from Level I to Level IV, in line with the actual project evolution. Thus, the following risk management priorities are established. The first priority is given to the water resources development and utilization rate, necessitating strict limits on total water withdrawal, promotion of agricultural water-saving irrigation, and preferential protection of ecological baseflow. Second priority is assigned to the biodiversity index, requiring the construction of ecological buffer zones and habitat restoration. Third priority is given to public satisfaction, calling for the establishment of a public feedback mechanism and optimization of water supply scheduling.

The main limitations of this study are as follows. The indicator system is found to be overly complex, with some indicators, such as the completeness of aquatic organism protection measures, difficult to accurately quantify, thereby increasing computational complexity and limiting the transferability of the system to other engineering contexts. Direct application of this system may lead to biased evaluations. Furthermore, the qualitative indicators are dependent on expert scoring, by which a degree of subjectivity is inevitably introduced. It is also difficult to fully unify the evaluation criteria for water diversion projects across different regions, so the comparability of the results is insufficient. Finally, only a single case study is analyzed, and projects under varying climatic, topographic, and diversion-scale conditions are not considered. Consequently, the key risk factors and their evolutionary patterns may differ substantially among projects. In addition, no multi-case validation is performed to optimize the model parameters, so the generalizability of the proposed evaluation framework is limited. Future optimization directions are proposed as follows: For ecological indicators with high data acquisition difficulty, remote sensing imagery should be integrated to construct proxy variable inversion models. Simultaneously, machine learning techniques should be combined to dynamically screen annually sensitive indicators, enabling dynamic updates to the indicator system. System dynamics should be incorporated to simulate the changing trends of ecological and environmental risks in water diversion projects under different management strategies. Comparative studies should be conducted on more projects with varying climates and scales to build a tiered and categorized risk evaluation benchmark database, thereby enhancing the universality of the evaluation framework.