Next Article in Journal
Differences of Occurrence, Distribution, and Factors Influencing Antibiotic Resistance Genes Between Freshwater and Seawater in China
Next Article in Special Issue
Developing a Multi-Objective Optimization Scheduling Method for the Yangtze to Huaihe River Water Diversion Project Considering Lake Regulation and Storage
Previous Article in Journal
Source, Transport, and Fractionation of Rare Earth Elements in Fluvial Sediments from a Typical Small Urban Basin (East Tiaoxi River, Eastern China)
Previous Article in Special Issue
Multi-Objective Model for Efficient, Equitable, and Sustainable Water Allocation Under Uncertainty: A Case Study of Namhan River Basin, South Korea
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Sustainable Water Optimization Tool (SUWO): An Optimization Framework for the Water–Energy–Food–Ecosystem Nexus

Environmental Engineering Department, Istanbul Technical University, 34469 Maslak, Istanbul, Türkiye
*
Authors to whom correspondence should be addressed.
Water 2025, 17(9), 1280; https://doi.org/10.3390/w17091280
Submission received: 18 March 2025 / Revised: 18 April 2025 / Accepted: 21 April 2025 / Published: 25 April 2025
(This article belongs to the Special Issue Optimization-Simulation Modeling of Sustainable Water Resource)

Abstract

:
Sustainable water management requires integrated approaches balancing competing demands and environmental sustainability. This study introduces the Sustainable Water Optimization Tool (SUWO), an open-source, Python-based simulation-optimization framework for basin-scale surface-water-resources management. SUWO employs the water–energy–food–ecosystem (WEF-E) nexus approach, utilizing a multi-objective genetic algorithm (MOGA) to generate Pareto-optimal solutions and facilitate a trade-off analysis among water uses through simulations of reservoir operations, hydro-energy production, irrigation, and flow regulation. SUWO integrates scenario analysis with multi-criteria decision making (MCDM), enabling the evaluation of various management, climate, and environmental scenarios. The framework was applied to the Sakarya River Basin (SRB) in Türkiye, a rapidly developing region pressured by water infrastructure development, hydroelectric power plants (HEPPs), and irrigation expansion. The SUWO-SRB model showed that while Non-dominated Sorting Genetic Algorithm II (NSGA-II) generally exhibited superior performance, NSGA-III presented a competitive alternative. The optimization results were analyzed across four management scenarios under varying hydrological conditions and environmental management classes (EMCs) for the near future. The model results highlight WEF-E nexus trade-offs. Maximizing energy production often impacts irrigation and the ecosystem, while prioritizing sustainable irrigation can reduce energy output. Dry conditions reduce hydropower and irrigation capacity, emphasizing water scarcity vulnerabilities. Ecological deviation negatively correlates with anthropogenic factors.

1. Introduction

Water, energy, food, and ecosystems are critical and interconnected systems, collectively forming the water–energy–food–ecosystem (WEF-E) nexus. These resources must be sustainably managed to meet growing global demands [1]. Rising population, climate change, and economic development have intensified pressure on water resources, necessitating integrated and optimized management approaches [2,3]. Traditional water resources management often focuses on single objectives, leading to inefficiencies and conflicts between sectors such as energy, agriculture, and environmental conservation [4]. Integrated Water Resources Management (IWRM), aligned with the WEF-E nexus, offers a system-based, multi-objective framework to address these complexities effectively [5].
Numerous studies have applied optimization to water-related challenges, covering various domains within the WEF-E nexus. These include trade-offs between competing sectors [6,7], transboundary water allocation [8,9], reservoir operations [10,11], environmental sustainability [12,13,14], irrigation efficiency [15], water quality and salinity control [16,17], and flood risk management [18,19].
Over recent decades, optimization methods have evolved significantly, transitioning from traditional approaches such as linear programming (LP) and dynamic programming (DP) to more advanced metaheuristic algorithms. In modern water resources management, evolutionary algorithms like genetic algorithms (GAs) and swarm intelligence methods such as Particle Swarm Optimization (PSO) have proven particularly effective. These algorithms are well suited for solving complex, nonlinear, multi-objective problems by balancing economic, environmental, and social goals, especially under conditions of uncertainty [20].
Building on the strengths of these metaheuristic methods, MOGAs have gained significant traction, particularly in optimizing cascade hydropower systems and integrated water resources management. One notable application is a basin-scale simulation-optimization model for the conjunctive use of surface and groundwater that was developed using an NSGA-II-based framework. This model demonstrated improved long-term water reliability and sustainability across scenarios involving climate variability and increasing demand [21]. In addition, an approach utilizing GAs was employed to optimize cascade reservoir operations, explicitly integrating environmental flow requirements across multiple environmental management classes. This approach proved that maintaining ecological health could be achieved with minimal trade-offs in hydropower generation, highlighting the dual objectives of energy production and environmental sustainability [22]. Further research into the Karun hydropower cascade system provided additional evidence of the effectiveness of NSGA-II. Optimization for both energy generation and environmental needs resulted in significant improvements in operational efficiency and sectoral trade-offs [23]. Additionally, NSGA-II was applied to cascade hydropower operations under varying ecological flow scenarios, and the results indicated that ecological targets could be achieved with only minor reductions in energy output. These findings demonstrate the potential of MOGA to optimize both ecological and energy production goals, even under fluctuating environmental conditions [24]. A comparison of different MOGA variants, specifically NSGA-II, NSGA-III, and RVEA, highlighted the strengths of each algorithm in cascade reservoir optimization. NSGA-III was found to perform best in minimizing flow deviations and reducing the reservoir water footprint, while NSGA-II delivered higher energy output, offering decision-makers a clear understanding of algorithm-specific trade-offs and guiding the selection of the most appropriate method for a given scenario [25]. To support decision making, the Multi-Objective Trade-off Index (MTI) was introduced to quantify trade-offs among power generation, water use, and ecological integrity within the WEF-E nexus of cascade hydropower systems. Developed using a multi-objective optimization framework, the MTI offers a clear and transferable method for evaluating competing objectives. These findings highlight the value of MOGA-based optimization in achieving balanced water resources management across hydro-energy, irrigation, and ecosystem sectors [26].
Several widely used water management models integrate optimization to support planning and operational decision making. HEC-ResPRM (Prescriptive Reservoir Model) [27] and MODSIM [28] use network flow optimization to determine efficient reservoir release strategies. RiverWare [29] incorporates goal programming and supports multi-objective optimization for complex system operations. WEAP [30] offers scenario-based planning with optional linear and mixed-integer optimization features. MIKE HYDRO [31] combines simulation with rule-based and optimization-driven approaches for basin-scale management. eWater Source [32] is widely used for national and regional water planning in Australia and has supported river basin studies in South Asia and the Mekong Basin. Region-specific models like CALVIN for California [33] and HIDROTURK for Türkiye [34,35] integrate optimization to support localized water planning. Some hydrological models, such as SWAT [36] and HYPE [37], can support optimization workflows when coupled with external tools.
Open-source tools are gaining traction through collaborative web-based platforms like GitHub [38], offering transparency, flexibility, and cost-effectiveness for research and education. Several models have been widely adopted in institutional and operational contexts. Pywr [39] provides a scriptable, modular structure for water balance modeling and allocation using linear programming, making it ideal for operational simulations and decision making under uncertainty. VIC-ResOpt [40] extends the VIC hydrological model with rule-based and optimized reservoir operation capabilities, supporting climate-informed planning. In the area of advanced optimization and decision support, tools like PLATEMO [41] and the M3O Toolbox [42] provide environments for benchmarking and developing multi-objective algorithms. M3O Toolbox is particularly suited to reservoir policy optimization, offering methods such as EMODPS and model predictive control. More specialized tools such as iRONS [43] and PySedSim [44] offer capabilities for sediment transport, dam infrastructure modeling, and river network simulation. Other open-source packages provide flexible support for reservoir and rule-based optimization. ResSimOpt [45], ReservoirManagement [46], WSIMOD [47], and Reservoir tool [48] enable scenario testing and basic optimization workflows, typically in research or educational contexts, while agricultural models such as AquaCrop [49] and utilities like PyETO [50] are valuable for irrigation planning and evapotranspiration estimation.
Despite these advances, simulation-optimization frameworks still face gaps and challenges. Proprietary tools often lack transparency and adaptability, while some open-source options are too specialized for broad, basin-scale planning. The integration of ecological processes, scenario analysis, and multi-criteria decision making is still underrepresented in many tools. To overcome these limitations, this study introduces Sustainable Water Optimization Tool (SUWO), an open-source, Python-based simulation-optimization framework designed to support IWRM in complex, multi-objective contexts across diverse river basins. SUWO incorporates reservoir operations, irrigation management, hydropower generation, and ecological flow requirements. It uses genetic algorithms to generate Pareto-optimal solutions, supports scenario-based planning, and evaluates ecological impacts through Environmental Management Classes (EMCs).
The modular and adaptable design of SUWO enables its application in various geographic and institutional contexts, making it a versatile decision-support tool for basin-wide, multi-sectoral water planning. By allowing seamless integration with external datasets and models, SUWO offers a transparent and extensible alternative to more rigid systems. As a demonstration of its capabilities, this study applies SUWO to the SRB in Türkiye, a basin facing increasing water stress due to rapid development, agricultural expansion, and climate variability. The framework is used to evaluate optimization results under multiple hydrological and environmental scenarios, comparing NSGA-II and NSGA-III in generating trade-off solutions among hydropower, irrigation, and ecosystem objectives. Through this application, SUWO showcases its potential as a scalable and effective tool for advancing sustainable water resources management, enabling ecosystem integration, trade-off analysis, and adaptive scenario-driven planning.

2. Methodology

2.1. SUWO Framework Overview

Adaptive water management is a cyclical process that integrates optimization with real-world implementation, fostering continuous learning and improvement in water resources management. The process begins by defining environmental objectives and making management decisions, followed by optimizing strategies and conducting trade-off analysis. Finally, the outcomes are implemented and monitored [51]. SUWO specifically focuses on the optimization and trade-off analysis steps, operating within a simulated system to identify optimal strategies and assess their potential impacts before real-world implementation.
SUWO comprises four key modules: (I) initialization, (II) optimization, (III) simulation, and (IV) analysis, as described in Figure 1. These modules are interconnected, allowing for an iterative feedback process that continuously refines and improves the model and its outputs.
  • The Initialization Module gathers and validates hydrological data, including runoff, inflow, and evaporation, while incorporating reservoir characteristics, operational rules, hydropower parameters, irrigation needs, and ecological requirements. It defines water flow pathways and supports scenario analysis, such as drought and climate change impacts.
  • The Optimization Module formulates water management objectives and constraints, optimizing water release for hydropower, irrigation, and ecological requirements. A multi-objective algorithm balances conflicting goals through iterative feedback with the simulation module.
  • The Simulation Module computes water balances, allocations, hydropower generation, and irrigation demand based on input data. It models water flow across the river network, considering operational rules, losses, and ecological requirements.
  • The Analysis Module evaluates optimization and simulation results using performance metrics and Pareto front analysis. It supports scenario analysis and post-processing, refining outputs to assess hydropower, irrigation, food production, and environmental flows.
Within these modules, the SUWO workflow outlines key process steps, as depicted in the simplified flowchart within Figure 2 and Table 1.
The SUWO process begins with data management and preparation using an Excel VBA template and the export_data.sub script. These data are then loaded and preprocessed by load_data.py for the optimization model. The main.py script defines the model parameters and configures the optimization problem, including objectives, constraints, and algorithm settings. The optimization itself is executed by problem.py, utilizing the Pymoo library. Next, simulation.py conducts water management simulations to evaluate the performance of potential solutions. The results are exported and imported by simulation.py for further analysis. Finally, analysis.py employs visualization libraries to generate Pareto fronts, performance metrics, and comparative scenarios, facilitating a comprehensive understanding of optimal water management strategies.
The flowchart provides a general overview of the SUWO workflow, implemented using Python (3.7) and leveraging libraries such as Pymoo, matplotlib, seaborn, and pandas for visualization and data analysis, and NumPy for numerical computations. Custom function libraries streamline specific tasks, including simulations, post-processing of results, and file management. To enhance usability, SUWO incorporates user-friendly features, including an Excel-based data input tool for intuitive data entry and management, bulk processing capabilities for large datasets, and seamless integration with Python scripts for advanced analysis and automation. The actual implementation involves more detailed scripts and functionalities, which can be found in the project’s GitHub repository.

2.2. Water Resources Simulation

This study employs a network-based approach to model water resources systems, conceptualizing them as dynamic networks comprised of nodes and links. This representation facilitates the expression of system dynamics through state-space equations, grounded in the fundamental principle of mass conservation. This approach enables a spatially explicit representation of the water system, capturing the interconnectedness of various components, such as reservoirs, distribution nodes, and demand points, and illustrating the flow dynamics across different routes [52,53].
To simulate the temporal evolution of the system, a defined time step is employed, capturing the dynamics of hydrological processes, water demands, and management operations. The selection of appropriate spatial and temporal scales in water resources optimization is crucial for accurately representing the system and achieving meaningful results [51].
In Figure 3, the SUWO network flow structure accounts for both spatial and temporal variations in water availability. The horizontal dimension depicts the temporal evolution of water resources, with carryover storage connecting consecutive time steps. Vertical connections illustrate the various flow processes within each time period, including releases, diversions, and demand allocations. The diagram also incorporates key infrastructure components, such as reservoirs, distribution nodes, and demand points, as well as irrigation schemes and energy generation considerations.
Water allocation within this framework is governed by a set of predefined rules and priorities, designed to balance the needs of diverse water users, including agriculture, domestic, industrial, and environmental sectors. These rules may be derived from water rights, legal agreements, or overarching management objectives. The allocation process is further constrained by factors such as minimum ecological flow requirements, reservoir storage limits, and the need to respect established water rights.
The SUWO model employs a network-based approach to represent water resource systems, simulating water flow and allocation over discrete time steps. This framework incorporates various inflow components (surface water, groundwater, return flows, precipitation) and outflow components (municipal/industrial use, irrigation, hydropower, evaporation, etc.) to ensure accurate water balance calculations.
The core of the simulation lies in the water balance equation (Equations (1) and (2)), applied at each node and time step:
S i , t = S i , t 1 + I n f l o w i , t O u t f l o w i , t ,   i = 1 ,   ,   n ,       t = 1 ,   ,   T
This can be further expanded as
S i , t = S i , t 1 + I n f i , t + E v a i , t M n × n E n g i , t +   S p l i , t N n × n I r g i , t i = 1 ,   ,   n ,       t = 1 ,   ,   T
In this equation, S i , t represents the storage volume at node i at the beginning of time step t, I n f l o w i , t represents the total inflow to node i at time step t, and O u t f l o w i , t represents the total outflow from node i at time step t. While this equation is universally applicable, the interpretation of the storage term varies depending on the node type. For reservoirs, it represents the actual stored water volume, crucial for regulating water availability. For distribution points, storage is considered negligible, reflecting their primary function as water transfer hubs.
The spatial connectivity of the reservoir network is fully defined by the connectivity matrix. The connectivity matrices M n × n and N n × n are defined such that a value of −1 indicates abstraction, while a value of +1 indicates receiving water from an upstream reservoir. Similarly, the irrigation system connectivity matrix denotes −1 for abstraction and +1 for receiving return water from upstream irrigation.
The water depth, denoted as H i , t , is calculated based on the storage volume using the following equation:
H i , t = α e β log S i ,   t + c
where α , β , and c are empirical coefficients derived from reservoir elevation–area–storage curves.
Evaporation, represented as E v a i , t , is estimated using the following equation:
E v a i , t = γ H i , t   σ   P e t i ,   t     10 - 4
where γ is the evaporation coefficient and P e t i ,   t is the potential evapotranspiration.
Spillway releases, denoted as   S p l i , t , are calculated as follows:
  S p l i , t =   S i , t S i   m a x   i f   S i , t > S i   m a x ,   o t h e r w i s e ,     S p l i , t = 0  
where S i   m a x represents the maximum storage capacity of the reservoir.
Following the water resources simulation process, the methodology employs a multi-objective optimization algorithm to identify optimal solutions, as described in the next section.

2.3. Multi-Objective Optimization

To find optimal water allocation strategies, the model utilizes a multi-objective optimization algorithm. This algorithm adjusts decision variables (e.g., water released for hydropower and irrigation) and evaluates the outcomes using the simulation module. This iterative process aims to balance potentially conflicting objectives, such as maximizing hydropower generation while meeting ecological needs. The flexibility of the model allows for tailoring specific objectives to different case studies, addressing diverse water management challenges effectively.
In the water resources problem, the decision variable x represents water release through the turbines, E n g i , t , and water abstraction for irrigation, I r g i , t , for each node i at each time step t, within the planning horizon T. The total number of decision variables is equal to n var .
x i , t = E n g i , t E n g i , T E n g n , t E n g n , T   I r g i , t I r g i , T I r g n , t I r g n , T
n var =   T   *   ( n Eng + n Irg )
Multi-objective optimization involves minimizing a vector of objectives, denoted as F ( x ) , subject to constraints:
o p t i m i z e   F ( x ) = [ f 1 ( x ) , f 2 ( x ) ,   .   .   .   , f k ( x ) ] T ,   where x S
Here, x represents the decision variables, and S represents the feasible solution space. This study considers three specific objective functions: energy, food, and ecosystem. The energy objective focuses on maximizing hydropower production by minimizing the deficit from its maximum potential. Food security is addressed through the food objective, which aims to minimize irrigation deficits. The ecosystem objective seeks to maintain ecological integrity by minimizing deviations from natural flow regimes.

2.3.1. Energy Objective

The energy objective focuses on maximizing hydropower production by minimizing the deficit from its maximum potential. This is achieved by optimizing the water released through turbines ( E n g ) to generate hydropower, while considering factors such as reservoir head, turbine flow, and efficiency. This is represented by the following equation:
min f e n e r g y x = i = 1 n t = 1 T H P c a p i H P i , t H P c a p i 2
This equation calculates the squared difference between the maximum potential hydropower generation ( H P c a p ) and the actual hydropower generation ( H P ) , normalized by the maximum potential. This formulation aims to maximize hydropower generation by minimizing the shortfall from the ideal production capacity.
Hydropower generation at each hydropower plant is calculated using the following function:
HP i , t = P i , t   *   T i , t HP
P i , t =   η T *   ρ   *   g   *   H i , t   net   *   Q i , t   tur
H i , t   net = H i , t H i , t   tail H i , t   L H i , t   m
Equation (10) calculates the hydropower generation ( HP ) at each hydropower plant, considering the power output ( P ) and turbine efficiency ( η T ). Equation (11) determines the power output based on the water density ( ρ ), acceleration due to gravity ( g ), net head ( H net ), and turbine flow ( Q tur ). Equation (12) calculates the net head by accounting for head ( H ), tailwater elevation ( H tail ), head loss ( H L ), and minor losses ( H m ).

2.3.2. Food Objective

The food objective aims to ensure food security by minimizing irrigation deficits. This is achieved by optimizing the water abstracted for irrigation ( I r g i , t ) while considering crop water requirements, irrigation technologies, and factors such as soil moisture and crop water stress. The objective function for minimizing irrigation deficits is as follows:
min f f o o d x = i = 1 n t = 1 T D i , t I r g i , t D i   m a x 2
In this equation, D i , t   represents the irrigation demand, I r g i , t   represents the irrigation supply at node i at time step t; D   m a x represents the maximum irrigation demand.
Irrigation demand is calculated based on several factors, including crop water requirements, which are determined by crop type, growth stage, and potential evapotranspiration rates. The model also considers different irrigation technologies and their efficiencies, such as surface, sprinkler, and drip irrigation. By incorporating these factors, the model can accurately estimate irrigation demand and optimize water allocation for agriculture.

2.3.3. Ecosystem Objective

The ecosystem objective focuses on maintaining ecological integrity by minimizing deviations from natural flow regimes. This is achieved by comparing the simulated outflow at the basin outlet ( Q o u t ) with the target environmental flow ( Q r e f ), and minimizing the difference between the two. The objective function for minimizing these deviations is
min f e c o s y s t e m x = i = 1 n t = 1 T Q r e f i , t Q o u t i , t Q r e f i   m a x 2
Maintaining natural flow regimes is crucial for preserving biodiversity, supporting ecological processes, and ensuring the long-term health of riverine ecosystems. Altered flow patterns can disrupt habitat connectivity, affect species migration, and degrade water quality, ultimately leading to a decline in ecosystem function and resilience. The environmental flow requirement is the absolute minimum flow needed to sustain basic ecological functions in the river and is treated as a hard constraint (Equation (17)) in the optimization model. The reference or target flow series   ( Q r e f ) represents a more nuanced ecological goal, aiming to mimic the natural flow regime as closely as possible, considering seasonal variations and high flows. The target environmental flow can be determined using various methods, including hydrological, hydraulic, habitat simulation, and holistic methods.
EMCs offer a framework for categorizing and managing river systems based on their ecological condition. These classes, ranging from natural (e.g., EMC-A) to heavily modified or degraded (e.g., EMC-F), reflect the extent of human impact on the system. The classification process often involves assessing various ecological indicators, such as water quality, habitat condition, and biological communities, against predefined criteria or thresholds. EMCs serve as a valuable tool for setting management objectives, prioritizing restoration efforts, and tracking changes in ecological condition over time [54,55].
In this study, the EMCs were determined using hydrological methods, specifically the flow duration curve (FDC) shifting technique. This approach involves shifting the baseline FDC, which represents the natural flow regime, downwards to create new FDCs for each EMC. The shifting is based on 17 fixed percentage points on the probability axis (from 0.01% to 0.1% up to 99.99%). An FDC shift by one step means the flow that was exceeded 99.99 percent of the time in the original FDC will be exceeded 99.9 percent of the time, and so on. For example, EMC-A might be the baseline FDC, EMC-B might be a one-step shift down, and EMC-C might be a two-step shift down.

2.3.4. Constraints

The optimization model incorporates various constraints to ensure that solutions are physically feasible and respect water conservation principles. These constraints include water balance requirements, ensuring that inflow and outflow are balanced at each node. Additional constraints are included for energy generation, discharge flow, reservoir storage limits, target storage, initial storage, irrigation withdrawal limits, and non-negativity, as represented by Equations (15)–(22).
Water balance constraint:
S i , t = S i , t 1 + I n f l o w i , t O u t f l o w i , t
Energy generation constraints:
HP i , t   HP i , t max
Discharge flow constraints:
Q i , t e c o   Q i , t o u t   Q i , t m a x
Reservoir storage limits:
S i   m i n     S i , t   S i   m a x
Target storage constraint:
i , S i , T S i t a r g e t
The initial storage of each reservoir:
i , S i , T = S i i n i t i a l = S i t a r g e t
Irrigation withdrawal limits:
I r g i , t m i n     I r g i , t   I r g i , t m a x
Non-negativity constraints:
S i ,   t ,   I r g i ,   t ,   R i , t ,   HP i , t ,   i     0
The following symbols are used to represent key variables in the optimization model: S i , t is the storage volume at node i at time t, and Q i , t i n   and Q i , t o u t   represent the inflow and outflow at node i at time t, respectively. HP represents hydropower generation, with HP max , being its maximum capacity. The discharge flow is represented by Q o u t , with ecological, Q e c o ,   and maximum, Q m a x ,   limits. The reservoir storage limits are denoted by S   m i n and S i   m a x , while S t a r g e t   is the target storage. The initial storage is represented by S i n i t i a l , and irrigation withdrawals are denoted by I r g .
The ecological limit, represented by the environmental flow requirement Q e c o , is the critical threshold for maintaining essential ecological functions in the river. This limit serves as a hard constraint in the optimization model, ensuring that no solution violates this ecological baseline. Determining the minimum flow requirement involves various methods, each with its own strengths and limitations. Hydrological approaches, often relying on historical flow records, are commonly used for broad-scale planning [56,57,58]. Simplified methods like the Tennant and Tessman methods are suitable when ecological data is limited. The Tennant method utilizes a percentage of annual average runoff, while the Tessman method considers the ratio of monthly to annual average flow. The Q p 90 method, which determines minimum ecological water demand based on the flow frequency of the driest month, offers another approach. The choice of method depends on data availability, scale, and specific ecological considerations.

2.3.5. Penalty Functions

Penalty functions are widely used in multi-objective optimization to enforce constraints by penalizing infeasible solutions and guiding the algorithm toward feasible regions. Effective penalty functions balance exploration and convergence by allowing minor constraint violations, preventing premature convergence and enhancing exploration near constraint boundaries. Selecting appropriate penalty values is critical, as overly large penalties can prematurely push solutions to feasibility, potentially missing the optimal point, whereas excessively small penalties can cause prolonged exploration in infeasible regions [59,60,61].
In the context of reservoir operation, penalty functions are incorporated into the objective function to enforce key constraints such as reservoir storage limits and end-of-period storage targets. The continuity constraint is inherently satisfied through the reservoir storage computation based on the continuity equation, whereas additional constraints are enforced using penalty terms (Equations (23)–(25)). This strategy enhances the applicability of genetic algorithms in handling highly constrained water resources management problems, ensuring feasible and optimal solutions.
g i , t m i n = c m i n m a x   0 ,   S i m i n S i , t 2
g i , t m a x = c m a x m a x   0 , S i , t S i m a x 2
g i t a r g e t = c t a r g e t m a x 0 ,   S i , T S i t a r g e t 2
Deviations from these constraints are penalized using squared differences from the constraint limits, weighted by coefficients ( c m i n ,     c m a x ,   c t a r g e t ) and a scaling factor ( k s c a l e ) to ensure appropriate magnitudes relative to the objective function. These penalty functions guide the optimization algorithm for feasible solutions. Therefore, the penalized objective function is formulated as Equation (26).
min F x = i = 1 n t = 1 T f i , t + k s c a l e i = 1 n t = 1 T g i , t m i n + i = 1 n t = 1 T g i , t m a x + i = 1 n g i t a r g e t

2.4. Optimization Algorithm Implementation

This study uses MOGAs, specifically NSGA-II and NSGA-III, within the SUWO optimization module for water allocation. MOGAs are effective for water resources management due to their ability to handle complex, nonlinear relationships and generate diverse Pareto-optimal solutions. While NSGA-II is well established, NSGA-III’s application in water resources is relatively recent and limited [62]. This research also aims to contribute by applying and comparing NSGA-III in large-scale IWRM optimization, evaluating its suitability through performance and solution analysis.
NSGA-II optimizes solutions through fast non-dominated sorting and crowding distance, ensuring both convergence and diversity in the Pareto front. NSGA-III, designed for many-objective problems, builds upon NSGA-II by introducing reference points, which enhance diversity and convergence in high-dimensional objective spaces.
Both algorithms follow a general evolutionary process, as illustrated in Figure 4. Beginning with an initial population, the algorithms evolve solutions through iterative selection, crossover, and mutation operators. The parent and offspring populations are then combined and ranked based on non-domination and crowding distance, ensuring that the best individuals are selected to form the next generation. This process drives the solutions towards the Pareto-optimal front, balancing convergence with diversity.
NSGA-II, a domination-based algorithm, excels in two-to-three objective problems, utilizing Pareto-dominance and crowding distance for diversity. However, its effectiveness diminishes in many-objective scenarios due to reduced selection pressure. Conversely, NSGA-III employs a reference-point framework, demonstrating superior performance in many-objective problems by maintaining a well-distributed Pareto front [62]. To comprehensively evaluate their suitability, this study compares NSGA-II and NSGA-III using convergence, diversity, non-dominance, and computational efficiency metrics.
Optimization algorithms were integrated into SUWO using the Pymoo Python package [63], ensuring reliability through implementations contributed by the original algorithm’s authors. Pymoo’s comprehensive suite of performance indicators (e.g., hypervolume, IGD+) and analytical tools enable robust evaluation, comparison, and informed decision-making regarding optimization processes and objective trade-offs [64].

3. Case Study

To showcase the practical application and effectiveness of the developed SUWO framework, a case study was conducted in the Sakarya River Basin (SRB) in Türkiye. This basin was strategically selected due to its complex water management challenges, stemming from its vital role in agriculture, hydropower generation, and municipal water supply, all while supporting important ecosystems.

3.1. Study Area: The Sakarya River Basin

The SRB, located in the northwestern Anatolian region of Türkiye, has a drainage area of 58,160 km2, approximately 7% of the country’s total land area. The river discharges into the Black Sea with its hydrology influenced by a predominantly continental climate. The SRB receives an annual average precipitation of 479 mm, below the national average of 574 mm, and generates an average flow of 6.4 billion m3, contributing 3.4% of Türkiye’s total water potential. Peak flow conditions typically occur between March and April, while the lowest flows are observed from August to October. Hydrologically, the basin is divided into six sub-basins: Upper Sakarya, Porsuk, Ankara-Kirmir, Middle Sakarya, Göksu, and Lower Sakarya. The SRB is composed primarily of agricultural areas (53%) and forests/semi-natural areas (44%), with artificial surfaces occupying 3% of the total area.
The basin’s water potential is estimated at 6294 MCM, while sectoral water demand totals 2947 MCM, primarily for agriculture (2034 MCM) and domestic use (689 MCM). Irrigation schemes established by State Hydraulic Works (DSI) cover 125,470 hectares, consuming about 20% of the basin’s water resources. Agricultural water use for all irrigated areas is calculated to be 947 MCM, with projected irrigation expanding to 640,266 hectares, contributing significantly to Türkiye’s agricultural output. There are large irrigation schemes such as Seyitgazi, Cifteler, Eskisehir, and Pamukova. The SRB also hosts 32 dams and 104 ponds, utilized for irrigation, drinking water supply, and energy generation. Water transfers from the basin to Istanbul and Ankara further complicate water management, with derivation systems from the Gerede and Kizilirmak rivers supplying Ankara’s drinking water. Major HEPPs include Sariyar, Gokcekaya, and Yenice, with the recently commissioned Kargi and Gursogut dams and HEPPs joining them forming a cascade system in the Middle Sakarya sub-basin, expected to produce 1602 GWh annually. Detailed information regarding irrigation schemes and HEPPs is provided in Appendix A [65,66,67].
The SRB faces a complex web of interconnected challenges, including natural variability in streamflow, human impacts such as dam construction and excessive water consumption, and climate change. The large cascade dams located on the main stem of the basin significantly alter the flow regime [68]. Low irrigation efficiency and excessive water consumption in agriculture exacerbate water scarcity concerns, highlighting the need for improved irrigation practices and infrastructure [69,70]. The presence of megacities like Ankara and Istanbul within the basin, along with inter-basin water transfers, necessitates careful water allocation strategies to meet urban demands while ensuring sufficient water availability for other uses [71,72,73,74]. Maintaining ecological integrity and addressing water quality issues stemming from pollution are also critical concerns [66]. Climate change further exacerbates these issues through decreased rainfall, increased droughts, and rising temperatures, threatening agricultural productivity and regional water security [75,76,77]. These interconnected challenges necessitate a holistic and sustainable water resources management approach to ensure the long-term health and water security of the SRB.
Various aspects of water resources management in the SRB have been investigated, ranging from basin management to hydrological characteristics, WEF-E nexus, reservoir optimization, and hydro-energy production [65,78,79,80,81]. In this study, the SUWO framework was applied to the SRB as a case study to demonstrate its capabilities within the WEF-E nexus context.

3.2. Data Sources and Processing

The comprehensive model scheme was developed to holistically assess the SRB and enable detailed sub-basin-level analysis, as illustrated in Figure 5. To minimize uncertainties and enhance model performance, node selection was strategically limited to 13 key nodes representing the basin’s hydrological and management characteristics. Node connectivity within the model was defined using a connectivity matrix, ensuring an accurate representation of water flow and allocation dynamics across the basin. The model operates on a one-year planning horizon with a monthly time step, allowing for a detailed temporal assessment of water availability and allocation dynamics.
The updated hydrological and management conditions were calculated and simulated based on data from “The Sakarya River Basin Master Plan Report” [65] supplemented by other relevant reports and literature reviews. These include newly constructed or planned reservoirs and hydropower plants, expanded irrigation schemes, advancements in water management technologies, and projected municipal water demands. Table 2 presents the simulated water components for these key nodes, including natural inflows, hydropower generation, irrigation water use, evaporation losses, and outflows. Natural flow data were analyzed for both dry and wet periods to assess hydrological variability.
The model was designed to prioritize minimum ecological flow requirements and domestic water demands, ensuring essential environmental and societal needs are met. The remaining water potential was then optimally allocated among irrigation and hydropower production, balancing competing water uses within the basin.
The Middle Sakarya features a significant cascade reservoir system, which was modeled by assigning a separate node to each reservoir. Energy production parameters were incorporated, and elevation–volume–area curves were used to simulate evaporation losses. Hydropower generation was estimated through primary and secondary calculations, while reservoir operations were governed by predefined operating curves and ecological flow requirements. The minimum and maximum hydropower production limits were established based on reservoir storage capacities. Notably, the Kargi and Yenice dams function as regulators, without significant water storage.
Irrigation activities were incorporated into the model as sub-basin-level nodes, with water consumption estimated using a weighted average method. The irrigation water demand was calculated based on crop water requirements, effective rainfall, and irrigation efficiency. To account for conveyance, field canal, and application losses, an efficiency factor was applied. The model also integrated monthly irrigation data, including return flows, to ensure comprehensive water balance calculations. Furthermore, minimum and maximum irrigation water limits were established based on historical irrigation amounts over the past five years, providing a realistic framework for evaluating irrigation practices.

3.3. Scenario Development

The SUWO framework facilitates comprehensive scenario analysis to evaluate trade-offs and identify optimal water management strategies across various dimensions. This study focuses on current/near-future projection conditions and encompasses MCDM under representative scenarios, hydrological variability across different flow regimes, and EMCs representing varying levels of ecological integrity.
To support decision making, MCDM is employed to generate pseudo-weight management scenarios. This involves calculating pseudo-weights for each objective function by evaluating the normalized distance of each solution to the worst solution in the objective space [82]. These pseudo-weights provide insights into the trade-offs between different objectives. The details of these MCDM scenarios are further explained in the Results section.
The model is initially run with average values derived from long-term observations of hydrological conditions. It is then further analyzed under dry and wet period flows to investigate the resulting outputs under different flow regimes. Additionally, environmental flow scenarios are examined, using varying flow conditions as a reference to analyze the model’s outputs.
EMC calculations were performed using the natural flow results for each node. These classes can serve as reference flow values in the model. EMC-A and EMC-B represent original and largely natural states. While EMC-A reflects a near-natural flow regime, it may pose challenges for reservoir management in meeting water demands for hydropower generation and other uses. Therefore, EMC-B and EMC-C can be considered as a reference for maintaining downstream water quality and aquatic life [83,84]. EMC-B was selected as the reference flow for normal model scenarios.
For simplification purposes, results obtained from the NSGA-II algorithm are presented in the analysis of hydrological conditions and environmental management scenarios.

3.4. Optimization Algorithm Parameters

The SUWO-SRB model’s optimization process was conducted using the NSGA-II and NSGA-III algorithms, with the configurations used detailed in Table 3. Both algorithms were executed with a population size of 1000 individuals over 1000 generations. Given the model’s complexity, comprising 13 nodes and 2 decision variables per node (water release through the turbines, E n g ; and water abstraction for irrigation, I r g ) across 12 months, the optimization process involved a total of 312 decision variables.
To ensure a comprehensive exploration of the solution space and identify optimal water management strategies, float random sampling was employed as the sampling method for both NSGA-II and NSGA-III. Parent selection for NSGA-II was conducted using tournament selection, whereas NSGA-III utilized an energy-based selection method tailored for three-objective optimization.
Crossover operations were implemented using simulated binary crossover (SBX), with NSGA-II configured with a distribution index (η) of 15 and a probability of 0.9, while NSGA-III applied SBX with a higher distribution index (η) of 30 and a probability of 1. Mutation in both algorithms was performed using polynomial mutation (PM) with a distribution index (η) of 20.
The entire optimization framework was developed using the Pymoo Python package, leveraging its advanced multi-objective optimization capabilities to effectively balance trade-offs between competing water management objectives.

3.5. Model Validation

Model validation is essential to ensure that simulation results align with observed data. This study employs a thorough validation process, comparing model outputs, such as flow, reservoir evaporation, energy production, irrigation water supply, and return flows, against observed values. This step is critical for establishing confidence in the model’s ability to accurately represent the real-world system.
Once the model structure and inputs are confirmed, optimization penalty functions are analyzed to assess their impact on objective functions and determine appropriate values. Penalty coefficients play a key role in controlling water levels and shaping the solution space. Weak penalties may result in unrealistic solutions, while excessive penalties can overly restrict the model. To address this, a sensitivity analysis is conducted to evaluate the effects of different penalty values and determine suitable coefficient ranges. This ensures that the optimization process effectively balances competing objectives while adhering to realistic operational constraints.
The SUWO-SRB model first runs a reference scenario to validate simulation accuracy. Then, a sensitivity analysis is performed using a coefficient matrix to assess the impact of varying penalty values. Suitable coefficient ranges are determined through analysis of bulk model outputs. Penalty weight coefficients and their respective ranges used in the SUWO-SRB model are provided in Table 4. For the SUWO-SRB model, a k s c a l e   of 10−6 was found to be appropriate. The values of c m i n , c m a x , and c t a r g e t coefficients ranged between 0.9 and 2. Increasing the c m i n   coefficient enhanced model flexibility by expanding the search space.

4. Results

4.1. Benchmarking Optimization Algorithms

MOGAs are typically evaluated based on their ability to converge to the true Pareto front, maintain diversity in solutions across the front, and spread solutions across the objective space [85]. The performance of NSGA-II and NSGA-III in solving the SUWO-SRB model was assessed using several metrics. Hypervolume (HV) [86] measures convergence and diversity as the dominated volume. The modified inverted generational distance (IGD+) [87,88] assesses convergence and diversity using the average distance to a reference set. The running IGD ( )   [89] tracks the algorithm’s progress using the IGD between current and later generations. The uniformity performance measure (UPM) [90] quantifies the solution’s distribution uniformity. The spread performance measure (SPM) [86] measures the solution’s spacing. The non-dominant size generation rate (NDSGR) [25] is the ratio of non-dominated solutions. The mean computational time (MCT) measures the computation time. A more detailed explanation of these performance metrics, including formulas, is provided in Appendix B.
A comparative performance analysis of the algorithms applied to the SUWO-SRB model is presented in Table 5. The evaluation is based on multiple runs and encompasses key performance metrics categorized into convergence, diversity, non-dominance, and computational efficiency. Bold values are used to highlight superior performance for each metric.
Further insights into the comparative performance of the algorithms are provided in Figure 6. This figure includes plots of hypervolume, IGD+, and running IGD ( ) over function evaluations, as well as a radar chart summarizing performance across multiple metrics.
An analysis of the HV metric across function evaluations shows that NSGA-III initially reaches a higher HV at a slightly faster rate than NSGA-II, indicating faster early-stage convergence. However, the final hypervolume achieved by NSGA-II (4.053) surpasses that of NSGA-III (3.341), reflecting a 21% improvement in the overall quality and spread of the Pareto front. This suggests that although NSGA-III may offer accelerated convergence in the early stages, NSGA-II ultimately provides a more comprehensive exploration of the objective space.
Both algorithms show a consistent downward trend in the IGD+ as the number of evaluations increases, confirming steady improvements in convergence and diversity. NSGA-II attains a final IGD+ value of 0.044, while NSGA-III concludes at 0.014, translating to a 68% lower error in approximating the ideal front. However, the mean IGD+ across all runs favors NSGA-II, indicating that it maintains more robust performance throughout the optimization process, despite NSGA-III’s stronger final value.
The running IGD (∅) curves for both algorithms demonstrate a gradual decrease until they stabilize after approximately 600,000 function evaluations, beyond which marginal gains diminish. Throughout the optimization, the curves for NSGA-II and NSGA-III closely overlap, suggesting comparable convergence behavior in tracking the evolving Pareto front. The narrow-shaded regions surrounding the curves reflect strong consistency and low variability across multiple runs, with NSGA-III exhibiting slightly greater stability than NSGA-II.
The radar chart offers a holistic visualization of performance across all metrics. NSGA-II dominates in five out of seven categories, particularly excelling in non-dominated solution generation (NDSGR: 0.924 vs. 0.501) and spread performance (SPM: 1.181 vs. 0.796). On the other hand, NSGA-III demonstrates advantages in computational efficiency, with a 1.9% lower runtime, and in producing a more uniform distribution of solutions, as reflected in slightly better UPM scores.
In summary, NSGA-II offers stronger overall performance in convergence, solution diversity, and exploration of trade-offs, making it preferable for problems where solution quality is critical. NSGA-III remains a valuable alternative, especially in scenarios prioritizing faster convergence, tighter solution clusters, and consistent performance. The selection of which algorithm to use should therefore consider the specific optimization goals, system complexity, and trade-off preferences relevant to the application.

4.2. Comparison of Pareto Fronts

The Pareto front visualizations in Figure 7 illustrate the trade-offs between the three objectives deviation from natural flow (ecology), irrigation deficit (food), and potential energy deficiency (energy), all formulated as minimization problems. These visualizations provide critical insights into how NSGA-II and NSGA-III explore the solution space and navigate the WEF-E nexus.
The 3D Pareto front shows that NSGA-II achieves a broader distribution of solutions across the objective space, reflecting its superior capability to identify a more diverse set of trade-offs. By contrast, NSGA-III’s solutions are more tightly clustered, indicating a narrower exploration range, which may benefit decision-makers seeking more consistent or specialized solutions.
The 2D projections of the Pareto front further emphasize the distinct characteristics of each algorithm’s search behavior. In the ecology vs. energy trade-off space, NSGA-II exhibits a noticeably broader range along both axes, highlighting its greater ability to explore diverse strategies that balance ecological flow requirements with hydropower generation. Similarly, in the energy vs. irrigation projection, NSGA-II uncovers a wider array of solutions, offering flexible configurations that optimize irrigation supply without significantly compromising energy output. In contrast, the irrigation vs. ecology projection reveals a higher concentration of NSGA-III solutions within a narrower band, indicating a more focused search and stronger convergence toward specific trade-off regions. This concentrated clustering suggests NSGA-III may be more suitable for applications where uniformity and consistency across objectives are prioritized.
The parallel coordinate plot provides another perspective by mapping each solution’s position across all three objectives. NSGA-II’s solutions demonstrate more variation, reinforcing its advantage in offering a broader portfolio of trade-offs. On the other hand, NSGA-III’s lines cluster tightly, indicating lower diversity but potentially higher consistency across multiple runs.
In summary, NSGA-II consistently provides a wider Pareto front, supporting more flexible decision making by offering diverse solutions that span the full trade-off space. NSGA-III delivers more constrained, compact, and well-clustered solutions, making it a suitable choice when preference is given to uniformity and stability. These distribution patterns are directly tied to the structural differences between the algorithms: NSGA-II relies on crowding distance for diversity, whereas NSGA-III’s reference-point-based selection tends to concentrate solutions around predefined directions.

4.3. Scenario Analysis

4.3.1. MCDM-Based Management Scenarios

To create representative scenarios for further analysis, we define four main objectives with corresponding pseudo-weights that reflect the priorities of each scenario: Balanced Allocation, Maximizing Energy, Sustainable Irrigation, and Eco-Centric Approach. These scenarios, detailed in Table 6, allow for the exploration of different management priorities and their impacts on the WEF-E nexus.
Figure 8 presents a visual comparison of the Pareto fronts generated by the NSGA-II and NSGA-III algorithms for the four representative management scenarios: Balanced Allocation, Maximizing Energy, Sustainable Irrigation, and Eco-Centric Approach. The axes represent the normalized values of the three objective functions energy, irrigation, and ecology, with arrows indicating the preferred direction for each objective.
This section presents the key simulation outcomes under four management scenarios, emphasizing hydropower generation and irrigation water supply across the SRB cascade system. For clarity, solutions derived using the NSGA-II algorithm are presented in this section. As shown in Figure 9, which presents simulated monthly and annual hydropower generation across the SRB cascade system. Monthly production exhibits strong seasonality, with peaks between February and May driven by snowmelt and elevated inflows. Under the Maximizing Energy scenario, generation reaches 230 GWh in April, compared to 95 GWh in August, reflecting a 142% seasonal increase. Annual energy output under this scenario totals 3248 GWh, representing a 38% increase in secondary energy compared to the Balanced scenario. In contrast, the Sustainable Irrigation scenario yields the lowest total output at 2384 GWh due to prioritization of ecological flows and agricultural water use.
Hydropower production also varies across individual HEPPs due to differences in reservoir storage capacity and elevation head. Gokcekaya and Sariyar alone contribute over 45% of the total cascade output, whereas regulator-type facilities like Kargi produce less energy owing to limited storage. The trade-off between water use sectors is evident: the Maximizing Energy scenario reduces irrigation delivery by 12.7% compared to Sustainable Irrigation, which meets full irrigation demand but sacrifices 864 GWh of hydropower. Firm energy generation, which provides base-load reliability, remains relatively constant (1850 GWh/year) across all scenarios. However, secondary energy, reliant on surplus water, fluctuates significantly (ranging from 1060 GWh to 480 GWh; a 120% difference).
Understanding the relationship between irrigation water supply and demand is essential for effective water resources management. Figure 10 illustrates the distribution of simulated monthly irrigation water supply alongside the irrigation demand range across eight key nodes in the SUWO-SRB model. The Sustainable Irrigation scenario, represented by the solid line, serves as a benchmark for evaluating how closely water deliveries align with agricultural requirements over time.
The simulation results show that water availability peaks between May and July across most basins, driven by snowmelt and seasonal rainfall. However, the degree of supply–demand alignment varies significantly among sub-basins. In Porsuk and Ankara, for example, median simulated deliveries during peak months (May–July) satisfy only approximately 76% of demand, indicating recurring shortfalls. In contrast, Pamukova and Goksu meet over 95% of demand during the same period, reflecting more favorable upstream regulation and storage conditions.
Across the system, the Sustainable Irrigation scenario achieves an average annual irrigation delivery of 196.5 MCM, meeting approximately 91% of total irrigation demand (215.8 MCM). However, in water-stressed regions such as Upper Sakarya and Kirmir, the scenario satisfies only 72% of peak monthly demand, underscoring the limitations imposed by existing infrastructure and upstream water allocation priorities.
These quantified mismatches highlight the spatial variability in irrigation performance and emphasize the need for basin-specific interventions. Enhancing reservoir operation strategies, increasing local storage capacity, and improving on-farm water-use efficiency could reduce shortfalls in high-demand months by up to 25%, based on observed gaps. Such adaptive measures are essential for strengthening system resilience and ensuring sustainable irrigation under future hydroclimatic variability.

4.3.2. Hydrological Conditions and Reservoir Operations

To evaluate the operational dynamics of the reservoir system under varying hydrological conditions, three distinct scenarios were defined: normal, wet, and dry years. These classifications were derived from long-term hydrological records using statistical analysis and hydrological indices to ensure representative selection. This approach enables a comprehensive assessment of reservoir behavior under different climatic regimes, facilitating informed and resilient operational planning.
Figure 11 illustrates the optimized monthly reservoir storage distributions across the SRB cascade system, based on solutions derived using NSGA-II. The boxplots capture reservoir storage variability under the three hydrological scenarios, while the gray shaded regions indicate the operational constraints, highlighting the minimum and maximum storage levels for each reservoir.
A comparative analysis reveals substantial variation in reservoir performance and resilience. Gursogut, an upstream reservoir, exhibits the widest storage fluctuation range, with levels varying by approximately 170 MCM between dry and wet years. It approaches maximum capacity in wet conditions and shows steep drawdowns during dry periods. Sariyar displays a distinct operational shift, peaking at around 1400 MCM in wet years, while dipping below 850 MCM in dry years, a swing of over 550 MCM, the largest variation across the cascade. Gokcekaya also reflects sensitivity, with an up to 100 MCM reduction in dry periods compared to wet ones, though it remains more buffered than Sariyar.
In contrast, Kargi and Yenice, situated downstream, maintain consistently high storage levels (above 44 MCM and 58 MCM, respectively), with negligible variation across hydrological scenarios. This operational stability suggests a regulatory function, likely intended for flow stabilization rather than dynamic storage.
The exceptional peak observed in Sariyar during wet years may reflect a modeling artifact stemming from the penalty function’s single-year optimization focus. This outcome indicates that extreme wet conditions can produce reservoir behaviors not easily mitigated within annual planning horizons, reinforcing the need for multi-year storage strategies. To ensure operational continuity, storage levels were constrained to return to their initial values by the end of each simulation year, as seen in all reservoirs.
These results emphasize the importance of spatially differentiated and temporally adaptive reservoir operation rules. The findings suggest that targeted multi-year strategies, especially for key reservoirs like Sariyar and Gursogut, are essential to improve robustness under both drought and flood risks.

4.3.3. Environment Management Classes

EMCs represent distinct ecological flow regimes developed through percentile-based analysis of long-term hydrological data. These six classes, from natural (EMC-A) to critically modified (EMC-F), serve as operational targets for managing river flows to align with ecological integrity goals. Figure 12 presents the simulated monthly outflows at multiple nodes under each EMC, based on solutions derived using NSGA-II, with historical observed flows provided for benchmarking deviations from natural conditions.
Across all nodes, a consistent seasonal pattern emerges, with peak flows occurring from December to May and lower flows from June to November, in line with regional hydrological cycles. Under EMC-A, outflows closely follow natural conditions, with monthly peaks often exceeding 350 MCM in wetter periods. In contrast, EMC-F flows can drop below MCM, reflecting intensified abstraction and regulation. These trends clearly illustrate the degradation gradient from EMC-A to EMC-F in terms of ecological flow availability.
The variability observed across nodes highlights spatial heterogeneity in flow responses to management strategies. While some locations show substantial deviations from natural conditions, others maintain relatively stable patterns, indicating that the impact of flow regulation is site-specific, influenced by location and hydrological context.
The effect of optimization on seasonal flow distribution is critical for evaluating management effectiveness. Figure 13 demonstrates that, despite the total annual discharge remaining within ±5% of observed volumes, the intra-annual distribution is significantly restructured under stricter EMC scenarios. For instance, spring peak flows are reduced by 28% under EMC-E, while summer low flows increase by 15%, reflecting a deliberate reallocation strategy to meet ecological thresholds during sensitive periods.
The shaded envelopes around the simulated flows emphasize the range of variability introduced by optimization, suggesting that while the model enhances flow stability within EMC targets, some hydrological and operational constraints still limit full alignment with environmental objectives. These findings underscore the role of optimization in refining flow regulation, providing actionable insights for adaptive and ecologically sensitive water management in the basin.

4.3.4. Overall Assessment with WEF-E Nexus

After implementing all scenarios, we can compare the SUWO-SRB model results based on the energy, irrigation, and ecosystem objectives. This analysis provides insights into the complex interactions within the WEF-E nexus, highlighting trade-offs and synergies among different water management strategies. Table 7 presents a comprehensive analysis of the water resources management results under various scenarios and hydrological conditions, focusing on key indicators such as hydropower generation, irrigation supply ratio, and ecological deviation.
Hydropower generation quantifies the contribution of water resources to energy production. The irrigation supply ratio, calculated as the total given water divided by the total irrigation demand, evaluates the adequacy of irrigation water supply to meet agricultural needs. The ecological deviation assesses the environmental impact of water management by quantifying the discrepancy between simulated outflow and target environmental flow regimes.
For a visual comparison of various scenarios within the WEF-E nexus, a scoring radar chart is used in Figure 14 to highlight trade-offs across key performance indicators. Each axis represents a WEF-E component, enabling a multidimensional visualization of scenario performance. By analyzing the shapes and areas enclosed by the plotted lines, trade-offs and overall performance across different management scenarios, hydrological conditions, and EMCs can be assessed.
Different management scenarios yielded varying outcomes across the WEF-E nexus. The “Maximizing Energy” scenario, for instance, resulted in the highest hydropower generation at 1909 GWh (NSGA-II) and 1887 GWh (NSGA-III), but reduced the average irrigation supply ratio to 0.87 (NSGA-II) and 0.89 (NSGA-III) compared to the “Sustainable Irrigation” scenario, which achieved a 1.00 (NSGA-II) and 0.99 (NSGA-III) irrigation supply ratio. The “Eco-Centric Approach” focused on minimizing ecological deviation, improving environmental performance but with mixed effects on hydropower and irrigation. Notably, both NSGA-II and NSGA-III exhibited similar trends across these scenarios, despite slight differences in their results.
Hydrological conditions significantly influenced system performance. Wet conditions led to a substantial increase in hydropower generation, reaching 2196 GWh, and an increased irrigation supply ratio of 1.09. However, this also resulted in the highest ecological deviation of 0.49, indicating potential adverse impacts on the ecosystem. In contrast, dry conditions caused a significant reduction in hydropower generation to 1272 GWh and a decrease in the irrigation supply ratio to 0.86, highlighting the vulnerability of these sectors to water scarcity.
A clear trend of increasing ecological deviation is observed as EMCs progress from EMC-A to EMC-F, rising from 0.08 to 5.46, demonstrating a strong negative correlation between ecosystem health and human intervention. Conversely, the irrigation supply ratio exhibits a positive correlation with EMCs, increasing from 0.90 in EMC-A to 1.06 in EMC-F, suggesting that greater human modification is associated with improved irrigation provision. However, hydropower generation follows a negative correlation, decreasing from 1882 GWh in EMC-A to 1763 GWh in EMC-F, highlighting a trade-off between energy production and ecological preservation.

5. Discussion and Conclusions

This study introduces SUWO, an open-source, Python-based framework for basin-scale water resources management under the WEF-E nexus. SUWO integrates the simulation of reservoir operations and water distribution with MOGAs, offering a flexible tool for analyzing trade-offs among conflicting objectives such as hydropower production, irrigation supply, and ecological integrity.
Application of SUWO to the SRB demonstrates the framework’s capacity to generate diverse and balanced management strategies under competing demands. Comparison of NSGA-II and NSGA-III provided insight into algorithmic behavior under varying objective dimensionality. NSGA-II yielded broader Pareto fronts and more non-dominated solutions, while NSGA-III achieved faster convergence and more uniform solution distributions. These results align with recent studies [25,26,62], supporting hybrid or context-sensitive algorithm selection in many-objective contexts. The performance of the algorithms was evaluated using various performance indices, including hypervolume, IGD+, and numeric indices, providing a quantitative comparison of their effectiveness.
Although SUWO is a newly developed and still-evolving tool, a general comparison with existing models can be made based on its core characteristics and intended functionalities. SUWO can be compared with other existing tools and models across four key features: optimization approach, planning focus, support for algorithm development, and simulation integration.
(1) Optimization approach: Compared to traditional models such as WEAP [30], RiverWare [29], and MIKE HYDRO [31], which are typically built on LP formulations, SUWO employs metaheuristic optimization to manage nonlinear, non-convex, and many-objective problem structures. While LP models benefit from computational speed and solver support, they often require simplifications or iterative adjustments to accommodate dynamic ecological or policy constraints [91]. Unlike these traditional models, SUWO’s use of population-based MOGAs offers, by integrating the Pymoo Python package [63], more flexibility in problem representation, particularly in emerging contexts involving many competing objectives across sectors, enabling a more nuanced analysis of trade-offs.
(2) Planning focus: Among the water simulation-optimization tools, Pywr [39] is the most conceptually similar to SUWO. Pywr uses linear programming (LP) to effectively simulate network-based water allocation and supports decision making under deep uncertainty (DMDU). While it excels in operational modeling, its reliance on LP limits its flexibility for addressing multi-objective, cross-sectoral planning challenges. Strategic, multi-objective planning is the core focus of SUWO, which provides flexibility in modeling trade-offs across sectors. SUWO allows for the evaluation of competing objectives such as hydropower, irrigation, and ecological integrity. It supports scenario evaluation and stakeholder-driven workflows, offering a more integrated and long-term approach to water resources management, highlighting SUWO’s advantage in addressing complex, multi-objective planning scenarios compared to LP-based models.
(3) Algorithm development: PLATEMO [41] and M3O Toolbox [42] represent the state of the art in optimization research. PLATEMO supports a wide array of MOGAs, while M3O includes advanced techniques such as Evolutionary Multi-Objective Direct Policy Search (EMODPS), model predictive control (MPC), and dynamic programming (DP). These platforms are excellent for algorithm benchmarking and design but are not built for applied, system-wide planning. SUWO bridges this gap by integrating robust optimization capabilities within a practical framework, making it well suited for basin-scale applications that require actionable decision support.
(4) Simulation integration: Detailed simulation tools such as PySedSim [44], VIC-ResOpt [40], and WSIMOD [47] focus on domain-specific processes like sediment transport, climate-sensitive reservoir operations, and urban–rural water dynamics. While powerful in their respective areas, these tools are not intended for broad, multi-objective planning. SUWO complements them by functioning as a higher-level, planning-oriented platform. It can integrate outputs from these models, apply ecological constraints using EMCs, and evaluate scenarios across sectors—enabling strategic, system-wide decision making.
Technically, SUWO prioritizes modularity and adaptability. Developed in Python, it uses an Excel-based input system that provides transparency and facilitates customization. While a graphical user interface (GUI) is not currently available, the codebase is structured to support future GUI development. Developed initially as a solo effort by an engineer, SUWO represents a prototype framework that aims to lay the foundation for broader collaborative development, rather than a fully mature software product. This open-source architecture encourages contributions and adaptation for basin-specific needs.
Despite its strengths, several limitations exist. SUWO depends on high-quality hydrological, ecological, and infrastructure data, and its calibration process is case-specific. Optimization routines are computationally demanding, particularly in multi-objective or high-resolution scenarios. The EMC-based approach, while practical, does not fully capture species-specific ecological dynamics. Generalizing the model to new basins will require additional validation and structural adaptation.
Nonetheless, SUWO distinguishes itself by occupying a unique position among modeling tools: it is not merely an operational simulator or a low-level optimizer, but a transparent, extensible platform for integrative planning. Its ability to handle multi-sectoral objectives, ecological flows, and stakeholder-driven scenarios positions it as a valuable tool for addressing complex water management challenges in a changing world. By offering many-objective analysis capabilities in a practical and adaptable format, SUWO provides a strong foundation for sustainable basin-scale decision support.

6. Future Research Directions

Future development of SUWO will prioritize enhancing its performance, usability, and applicability across diverse basin contexts to better support integrated WEF-E management. To achieve this, the following key research directions have been identified
Future research will explore the development of WEF-E index-based scenarios, the expansion from multi-objective to many-objective optimization, and the integration of climate scenarios to evaluate long-term system resilience under hydrological uncertainty.
Refining the WEF-E analysis through the development of node-level performance metrics and index scores will support more detailed diagnostics of trade-offs among sectors and inform more targeted management strategies.
Addressing the challenges in handling highly constrained systems by developing an automated calibration tool for penalty coefficients and sensitivity analysis will streamline the currently time-intensive calibration process, which is critical because parameter tuning is a key issue in genetic algorithms, and the effectiveness of optimization relies on the careful calibration of penalty functions and their coefficients. This automation will also enhance SUWO’s usability by making the model more accessible to a wider range of users.
Incorporating a GUI to enable scenario configuration, visualization, and result interpretation for non-technical users will lower the barrier to entry and facilitate wider adoption of the SUWO framework by practitioners.
Further research will explore computational enhancements such as parallel execution and surrogate modeling to reduce runtimes. This is crucial for enabling the application of SUWO to large-scale and complex water resource systems.
Continued collaborative development via GitHub to improve code quality, modularity, and documentation will be essential to expand SUWO’s use in diverse hydroclimatic and institutional contexts.
It is also important to acknowledge that SUWO reflects the effort of a single developer with an engineering background. While the current architecture establishes a functional prototype, future development will benefit significantly from team-based collaboration, including support from software developers and computational modelers. Ultimately, these efforts aim to evolve SUWO into a robust, flexible, and widely applicable decision-support platform for many-objective water resource planning.

Software Availability

  • Name of software: SUWO (Sustainable Water Optimization Tool) (beta)
  • Contact Email: yaykirans@itu.edu.tr
  • Source Language: Python
  • Supported Systems: Windows, Mac, Linux
  • License: MIT
  • Availability: github.com/syaykiran/SUWO-Sustainable-Water-Optimization-Tool (accessed on 18 March 2025).

Author Contributions

Conceptualization, S.Y. and A.E.; Methodology, S.Y. and A.E.; Software, S.Y.; Validation, S.Y. and A.E.; Formal analysis, S.Y. and A.E.; Investigation, S.Y.; Resources, A.E.; Data curation, S.Y.; Writing—original draft, S.Y.; Writing—review & editing, A.E.; Visualization, S.Y.; Supervision, A.E.; Project administration, S.Y. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding authors.

Acknowledgments

The authors would like to thank Melike GUREL and Ali ERTURK for their valuable ideas as PhD Thesis Progress Committee members throughout the study, to Gokhan CUCELOGLU and Nuray AKALIN for their support, and to General Directorate of State Hydraulic Works (DSI) of Republic of Türkiye for providing the data required to conduct this PhD Thesis study.

Conflicts of Interest

The authors declare no conflict of interest.

Abbreviations

The following abbreviations are used in this manuscript:
WEF-EWater–Energy–Food–Ecosystem
IWRMIntegrated Water Resources Management
SUWOSustainable Water Optimization Tool
SRBSakarya River Basin
GAGenetic algorithm
MOGAMulti-objective genetic algorithm
NSGA-IINon-dominated Sorting Genetic Algorithm II
NSGA-IIINon-dominated Sorting Genetic Algorithm III
MCDMMulti-criteria decision making
HEPPHydroelectric power plant
EMCEnvironmental management class
PSOParticle Swarm Optimization
MTIMulti-Objective Trade-off Index
FDCFlow duration curve
LPLinear programming
DPDynamic programming
HVHypervolume
IGDInverted generational distance
IGD+Modified inverted generational distance
UPMUniformity performance measure
SPMSpread performance measure
NDSGRNon-dominated size generation rate
MCTMean computational time
DMDUDecision making under deep uncertainty
EMODPSEvolutionary Multi-Objective Direct Policy Search
MPCModel predictive control
GUIGraphical user interface

Appendix A. Basin Information

Table A1. Characteristics of major irrigation projects and their projected water demands in SRB.
Table A1. Characteristics of major irrigation projects and their projected water demands in SRB.
Sub-BasinIrrigation ProjectNet Irrigation Area (ha)Water Demand (m3/ha)
Upper Sakarya Seyitgazi13,0002889
Cifteler62003844
Yarali36005070
Cildirim28003867
Ilgin Atlanti10,2303410
PorsukEskisehir16,6004312
Kutahya22073422
Inonu18503522
Kuzfindik21793800
AnkaraAkinci28004187
Doganozu24244084
Kirmir Asartepe1500 4832
Middle SakaryaOSV Saricakaya47004027
Karaagac24004371
Pamukova79003911
Kizildamlar10953215
GoksuBogazkoy10,1664121
Babasultan30793312
Table A2. Physical and operational characteristics of hydropower reservoirs in SRB.
Table A2. Physical and operational characteristics of hydropower reservoirs in SRB.
ReservoirDam Height (m)Water Level (m)Water Storage
(106 m3)
Installed
Capacity (MW)
Commission
Year
Gursogut HP596491103712022
Kargi HP67558501002009
Sariyar HP9047514111601956
Gokcekaya HP 1163899532781972
Yenice HP 3327358381999

Appendix B. Performance Metrics

(1)
Hypervolume (HV), also known as the S-metric, measures both convergence and diversity as the volume of the objective space dominated by the Pareto front approximation ( Y N ) and delimited by a reference objective vector ( r R m ), and is calculated as
H V Y N ;   r = λ m y     Y N y ,   r
where Y N is the Pareto front approximation; r R m is a reference objective vector, where all solutions ( y     Y N ) are less than or equal to the reference vector ( y  ≤ r); and λ m is the Lebesgue measure. Higher HV values indicate better performance.
(2)
The modified inverted generational distance (IGD+) assesses the convergence and diversity of a solution set by measuring the average distance between points in a reference set and the closest points in the solution set, employing a modified distance metric. The generational distance (GD) of a set A is defined as the average distance between each point aA and the closest point r in a reference set R. It is calculated as
G D p ( A , R ) = 1 A a A m i n r R   d a , r p 1 / p
where the distance is the Euclidean distance:
d ( a , r ) = k = 1 M a k   r k 2
The inverted generational distance (IGD) is calculated as
I G D p ( A , R ) = G D p ( R , A )
The modified inverted generational distance (IGD+) uses the following modified distance measure:
d + ( r , a ) = k = 1 M m a x r k   a k ,   0 2
A lower IGD+ value indicates better performance.
(3)
Running IGD ( ) assesses an algorithm’s performance trajectory by comparing its current performance to its performance at a subsequent generation (τ) within a sliding window of size ω, where τ denotes the later generation at which the performance plot is updated, and ω denotes the sliding window size used for the termination condition. The running IGD is computed for all past generations ( 0     t     τ ). It employs the IGD metric to measure the difference between the current solution set and a solution set from a later generation:
t τ = I G D P ¯ τ   t ,   P ¯ τ τ
The formula calculates a “progress score” ( t τ ) by comparing the normalized solutions at the current time (t) to the normalized solutions at a later time (τ) using a quality measure called the IGD. The P ¯ τ indicates that the solutions have been adjusted using information up to time τ to make the comparison fair.
(4)
The non-dominant size generation rate (NDSGR) represents the mean ratio of non-dominated size to population size across multiple runs. It is calculated as:
N D S G R = 1 n i = 1 n N n d o m ( i ) N p o p ( i )  
where Nndom(i) is the non-dominated size of the i-th run, Npop(i) is the population size of the i-th run, and n is the total number of runs.
(5)
The uniformity performance measure (UPM) evaluates the uniformity of the solution set based on the variance of distances between solutions. The UPM is calculated as:
U P M = 1 n 1 i = 1 n ( d ¯ d i ) 2 0.5  
d i = min j k = 1 m   f k x i f k x j ; i ,   j = 1 ,   2 ,   ,   n    
where n is the number of solutions, d i is the minimum distance from the i-th solution to the other solutions in the Pareto front, d is the mean of d i , and m is the objective dimension. A smaller UPM value indicates a more uniform distribution of solutions.
(6)
The spread performance measure (SPM) quantifies the spread of solutions in the objective space by calculating the maximum distance between extreme solutions. A larger SPM value indicates a wider distribution of the solution set. The SPM is calculated as:
S P M ( x ) = k = 1 m m a x   a k   b k ; a ,   b x  
where m is the objective dimension, a and b are two solutions, and the maximum distance is calculated across all extreme solutions. A larger SPM value indicates a wider distribution of the solution set.
(7)
The mean computational time (MCT) is the average computational time across multiple runs. The MCT is calculated as:
M C T = 1 n i = 1 n t i
where T is the mean computational time in seconds, t i   is the computational time of the i-th run in seconds, and n is the total number of runs.
(8)
The overall score is calculated to facilitate a fair comparison between different models by normalizing performance metrics to a score between 0 and 1. This approach ensures that higher scores consistently indicate better performance, regardless of whether a metric is a maximization or minimization objective. Given a set of metric values X = {   x ,   x ,   . . . ,   x   } for all models, each metric x is normalized using the following formulas:
For metrics where a higher value is desirable (e.g., Hypervolume, NDSGR), normalization is performed as:
S c o r e ( x ) = x max X
For metrics where a lower value is desirable (e.g., IDG+, UPM), normalization is performed as:
S c o r e ( x ) = min X x

References

  1. Bazilian, M.; Rogner, H.; Howells, M.; Hermann, S.; Arent, D.; Gielen, D.; Steduto, P.; Mueller, A.; Komor, P.; Tol, R.S.; et al. Considering the Water, Energy and Food Nexus: Towards a Resource-Sufficient and Resilient Future. Energy Policy 2011, 39, 7896–7906. [Google Scholar] [CrossRef]
  2. Vörösmarty, C.J.; Green, P.; Salisbury, J.; Lammers, R.B. Global Water Resources: Vulnerability from Climate Change and Population Growth. Sci. New Ser. 2000, 289, 284–288. [Google Scholar] [CrossRef] [PubMed]
  3. Intergovernmental Panel on Climate Change (IPCC). Climate Change 2022—Impacts, Adaptation and Vulnerability. Working Group II Contribution to the IPCC Sixth Assessment Report of the Intergovernmental Panel on Climate Change; Pörtner, H.-O., Roberts, D.C., Tignor, M., Poloczanska, E.S., Mintenbeck, K., Alegría, A., Craig, M., Langsdorf, S., Löschke, S., Möller, V., Okem, A., Rama, B., Eds.; Cambridge University Press: Cambridge, UK; New York, NY, USA, 2023; p. 3056. [Google Scholar] [CrossRef]
  4. Loucks, D.P.; van Beek, E. Water Resource Systems Planning and Management; Springer International Publishing: Cham, Switzerland, 2017; ISBN 978-3-319-44232-7. [Google Scholar]
  5. Grigg, N.S. IWRM and the Nexus Approach: Versatile Concepts for Water Resources Education. J. Contemp. Water Res. Educ. 2019, 166, 24–34. [Google Scholar] [CrossRef]
  6. Dhaubanjar, S.; Davidsen, C.; Bauer-Gottwein, P. Multi-Objective Optimization for Analysis of Changing Trade-Offs in the Nepalese Water–Energy–Food Nexus with Hydropower Development. Water 2017, 9, 162. [Google Scholar] [CrossRef]
  7. Maragò, C.; Guido, R.; Guerriero, F. Multi-Objective Optimization for the Security of Water-Energy-Food Nexus. In Optimization in Green Sustainability and Ecological Transition; Bruglieri, M., Festa, P., Macrina, G., Pisacane, O., Eds.; AIRO Springer Series; Springer Nature Switzerland: Cham, Switzerland, 2024; Volume 12, pp. 289–298. ISBN 978-3-031-47685-3. [Google Scholar]
  8. Digna, R.; Castro-Gama, M.; van der Zaag, P.; Mohamed, Y.; Corzo, G.; Uhlenbrook, S. Optimal Operation of the Eastern Nile System Using Genetic Algorithm, and Benefits Distribution of Water Resources Development. Water 2018, 10, 921. [Google Scholar] [CrossRef]
  9. Kay, M.; Çetin, Ö.; Çapar, G.; Ahi, Y.; Pilevneli, T.; Maden, T.E.; Sarıkaya, M.S.; Akalın, N. Improving Agricultural Water Use Efficiency and Productivity in the Middle East: Pressures, Status, Impacts and Responses; Disseminating Knowledge to Improve Agricultural Water Use Efficiency in the Middle East; Turkish Water Institute (SUEN): Istanbul, Türkiye, 2022; ISBN 978-625-8451-33-7. [Google Scholar]
  10. Lai, V.; Huang, Y.F.; Koo, C.H.; Ahmed, A.N.; El-Shafie, A. A Review of Reservoir Operation Optimisations: From Traditional Models to Metaheuristic Algorithms. Arch. Comput. Methods Eng. 2022, 29, 3435–3457. [Google Scholar] [CrossRef]
  11. Chang, L.-C.; Chang, F.-J.; Wang, K.-W.; Dai, S.-Y. Constrained Genetic Algorithms for Optimizing Multi-Use Reservoir Operation. J. Hydrol. 2010, 390, 66–74. [Google Scholar] [CrossRef]
  12. Bejarano, M.D.; Sordo-Ward, A.; Gabriel-Martin, I.; Garrote, L. Tradeoff between Economic and Environmental Costs and Benefits of Hydropower Production at Run-of-River-Diversion Schemes under Different Environmental flows scenarios. J. Hydrol. 2019, 572, 790–804. [Google Scholar] [CrossRef]
  13. Chen, W.; Olden, J.D. Designing Flows to Resolve Human and Environmental Water Needs in a Dam-Regulated River. Nat. Commun. 2017, 8, 2158. [Google Scholar] [CrossRef]
  14. Tariku, T.B.; Gan, K.E.; Tan, X.; Gan, T.Y.; Shi, H.; Tilmant, A. Global Warming Impact to River Basin of Blue Nile and the Optimum Operation of Its Multi-Reservoir System for Hydropower Production and Irrigation. Sci. Total Environ. 2021, 767, 144863. [Google Scholar] [CrossRef]
  15. Gasanov, M.; Merkulov, D.; Nikitin, A.; Matveev, S.; Stasenko, N.; Petrovskaia, A.; Pukalchik, M.; Oseledets, I. A New Multi-Objective Approach to Optimize Irrigation Using a Crop Simulation Model and Weather History. In Computational Science—ICCS 2021; Paszynski, M., Kranzlmüller, D., Krzhizhanovskaya, V.V., Dongarra, J.J., Sloot, P.M.A., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham, Switzerland, 2021; Volume 12745, pp. 75–88. ISBN 978-3-030-77969-6. [Google Scholar]
  16. Abdullah, A.D.; Castro-Gama, M.E.; Popescu, I.; van der Zaag, P.; Karim, U.; Al Suhail, Q. Optimization of Water Allocation in the Shatt al-Arab River under Different Salinity Regimes and Tide Impact. Hydrol. Sci. J. 2018, 63, 646–656. [Google Scholar] [CrossRef]
  17. Saadatpour, M. An Adaptive Surrogate Assisted CE-QUAL-W2 Model Embedded in Hybrid NSGA-II_ AMOSA Algorithm for Reservoir Water Quality and Quantity Management. Water Resour. Manag. 2020, 34, 1437–1451. [Google Scholar] [CrossRef]
  18. Zargar, M.; Samani, H.M.V.; Haghighi, A. Optimization of Gated Spillways Operation for Flood Risk Management in Multi-Reservoir Systems. Nat. Hazards 2016, 82, 299–320. [Google Scholar] [CrossRef]
  19. Chang, L.-C. Guiding Rational Reservoir Flood Operation Using Penalty-Type Genetic Algorithm. J. Hydrol. 2008, 354, 65–74. [Google Scholar] [CrossRef]
  20. Fonseca, C.M.; Fleming, P.J. Genetic Algorithms for Multiobjective Optimization: Formulation, Discussion and Generalization. In Proceedings of the ICGA-93: Fifth International Conference on Genetic Algorithms, San Mateo, CA, USA, 17–22 July 1993; pp. 416–423. [Google Scholar]
  21. Song, J.; Yang, Y.; Sun, X.; Lin, J.; Wu, M.; Wu, J.; Wu, J. Basin-Scale Multi-Objective Simulation-Optimization Modeling for Conjunctive Use of Surface Water and Groundwater in Northwest China. Hydrol. Earth Syst. Sci. 2020, 24, 2323–2341. [Google Scholar] [CrossRef]
  22. Suwal, N.; Huang, X.; Kuriqi, A.; Chen, Y.; Pandey, K.P.; Bhattarai, K.P. Optimisation of Cascade Reservoir Operation Considering Environmental Flows for Different Environmental Management Classes. Renew. Energy 2020, 158, 453–464. [Google Scholar] [CrossRef]
  23. Jamali, S.; Jamali, B. Cascade Hydropower Systems Optimal Operation: Implications for Iran’s Great Karun Hydropower Systems. Appl. Water Sci. 2019, 9, 66. [Google Scholar] [CrossRef]
  24. Yu, L.; Wu, X.; Wu, S.; Jia, B.; Han, G.; Xu, P.; Dai, J.; Zhang, Y.; Wang, F.; Yang, Q.; et al. Multi-Objective Optimal Operation of Cascade Hydropower Plants Considering Ecological Flow under Different Ecological Conditions. J. Hydrol. 2021, 601, 126599. [Google Scholar] [CrossRef]
  25. Wang, M.; Zhang, Y.; Lu, Y.; Wan, X.; Xu, B.; Yu, L. Comparison of Multi-Objective Genetic Algorithms for Optimization of Cascade Reservoir Systems. J. Water Clim. Change 2022, 13, 4069–4086. [Google Scholar] [CrossRef]
  26. Wu, X.; Yu, L.; Wu, S.; Jia, B.; Dai, J.; Zhang, Y.; Yang, Q.; Zhou, Z. Trade-Offs in the Water-Energy-Ecosystem Nexus for Cascade Hydropower Systems: A Case Study of the Yalong River, China. Front. Environ. Sci. 2022, 10, 857340. [Google Scholar] [CrossRef]
  27. U.S. Army Corps of Engineers. USACE HEC-ResPRM Prescriptive Reservoir Model; Institute for Water Resources Hydrologic Engineering Center: Davis, CA, USA, 2003.
  28. Labadie, J.W.; Baldo, M.L.; Larson, R. MODSIM: Decision Support System for River Basin Management: Documentation and User Manual; Department of Civil Engineering at Colorado State University: Ft. Collins, CO, USA, 2000. [Google Scholar]
  29. Zagona, E.A.; Fulp, T.J.; Shane, R.; Magee, T.; Goranflo, H.M. Riverware: A Generalized Tool for Complex Reservoir System Modeling. J. Am. Water Resour. Assoc. 2001, 37, 913–929. [Google Scholar] [CrossRef]
  30. Sieber, J.; Purkey, D. Water Evaluation And Planning System (WEAP) User Guide; Stockholm Enviroment Institue U.S. Center: Somerville, MA, USA, 2015. [Google Scholar]
  31. Danish Hydraulic Institute (DHI). MIKE HYDRO River, User Guide, 2017; DHI Water & Environment: Hørsholm, Denmark, 2017. [Google Scholar]
  32. eWater Solutions. eWater Source, Version 2023; eWater Ltd.: Canberra, Australia, 2023; Available online: https://ewater.org.au (accessed on 29 March 2025).
  33. Medellín-Azuara, J.; Mendoza-Espinosa, L.G.; Lund, J.R.; Harou, J.J.; Howitt, R.E. Virtues of Simple Hydro-Economic Optimization: Baja California, Mexico. J. Environ. Manag. 2009, 90, 3470–3478. [Google Scholar] [CrossRef]
  34. Özdemi̇r Çalli, K.; Taşci, Y.; Uzun, M.; Karaaslan, Y. Predicting of the Climate-Induced Depletion in Groundwater Level and Storage: A Case in Akarçay Basin. Turk. J. Water Sci. Manag. 2021, 6, 145–176. [Google Scholar] [CrossRef]
  35. Ertürk, A.; Gürel, M.; Ekdal, A.; Cüceloğlu, G.; Karpuzcu, M.E.; Eropak, C.; Çinar, S.; Şanli, N.; Kiymaz, G.; Çeti̇nkaya, S.; et al. A Flexible Water Quality Modelling Simulator Based on Matrix Algebra. Turk. J. Water Sci. Manag. 2021, 6, 31–88. [Google Scholar] [CrossRef]
  36. Neitsch, S.L.; Arnold, J.G.; Kiniry, J.R.; Williams, J.R. Soil and Water Assessment Tool (SWAT) Theoretical Documentation; Texas A&M University: College Station, TX, USA, 2011. [Google Scholar]
  37. Lindström, G.; Pers, C.; Rosberg, J.; Strömqvist, J.; Arheimer, B. Development and Testing of the HYPE (Hydrological Predictions for the Environment) Water Quality Model for Different Spatial Scales. Hydrol. Res. 2010, 41, 295–319. [Google Scholar] [CrossRef]
  38. GitHub, Inc. GitHub. Available online: https://github.com/ (accessed on 24 February 2025).
  39. Tomlinson, J.E.; Arnott, J.H.; Harou, J.J. A Water Resource Simulator in Python. Environ. Model. Softw. 2020, 126, 104635. [Google Scholar] [CrossRef]
  40. Dang, T.D.; Vu, D.T.; Chowdhury, A.F.M.K.; Galelli, S. A Software Package for the Representation and Optimization of Water Reservoir Operations in the VIC Hydrologic Model. Environ. Model. Softw. 2020, 126, 104673. [Google Scholar] [CrossRef]
  41. Tian, Y.; Cheng, R.; Zhang, X.; Jin, Y. PlatEMO: A MATLAB Platform for Evolutionary Multi-Objective Optimization [Educational Forum]. IEEE Comput. Intell. Mag. 2017, 12, 73–87. [Google Scholar] [CrossRef]
  42. Giuliani, M.; Li, Y.; Cominola, A.; Denaro, S.; Mason, E.; Castelletti, A. A Matlab Toolbox for Designing Multi-Objective Optimal Operations of Water Reservoir Systems. Environ. Model. Softw. 2016, 85, 293–298. [Google Scholar] [CrossRef]
  43. Peñuela, A.; Hutton, C.; Pianosi, F. An Open-Source Package with Interactive Jupyter Notebooks to Enhance the Accessibility of Reservoir Operations Simulation and Optimisation. Environ. Model. Softw. 2021, 145, 105188. [Google Scholar] [CrossRef]
  44. Wild, T.B.; Birnbaum, A.N.; Reed, P.M.; Loucks, D.P. An Open Source Reservoir and Sediment Simulation Framework for Identifying and Evaluating Siting, Design, and Operation Alternatives. Environ. Model. Softw. 2021, 136, 104947. [Google Scholar] [CrossRef]
  45. Dobson, B.; Wagener, T.; Pianosi, F. An Argument-Driven Classification and Comparison of Reservoir Operation Optimization Methods. Adv. Water Resour. 2019, 128, 74–86. [Google Scholar] [CrossRef]
  46. Cuvelier, T.; Archambeau, P.; Dewals, B.; Louveaux, Q. Comparison Between Robust and Stochastic Optimisation for Long-Term Reservoir Management Under Uncertainty. Water Resour Manag. 2018, 32, 1599–1614. [Google Scholar] [CrossRef]
  47. Dobson, B.; Liu, L.; Mijic, A. Water Systems Integrated Modelling Framework, WSIMOD: APython Package for Integrated Modelling of Water Quality and Quantityacross the Water Cycle. J. Open Source Softw. 2023, 8, 4996. [Google Scholar] [CrossRef]
  48. Turner, S.W.D.; Galelli, S. Water Supply Sensitivity to Climate Change: An R Package for Implementing Reservoir Storage Analysis in Global and Regional Impact Studies. Environ. Model. Softw. 2016, 76, 13–19. [Google Scholar] [CrossRef]
  49. Steduto, P.; Raes, D.; Hsiao, T.C.; Fereres, E.; Heng, L.; Izzi, G.; Hoogeveen, J. AquaCrop: A New Model for Crop Prediction Under Water Deficit Conditions. Options Méditerranéennes 2009, 80, 285–292. [Google Scholar]
  50. Richards, M. PyETo. Available online: https://github.com/woodcrafty/PyETo (accessed on 8 February 2025).
  51. Derepasko, D.; Guillaume, J.H.A.; Horne, A.C.; Volk, M. Considering Scale within Optimization Procedures for Water Management Decisions: Balancing Environmental Flows and Human Needs. Environ. Model. Softw. 2021, 139, 104991. [Google Scholar] [CrossRef]
  52. Heidari, M.; Chow, V.T.; Kokotović, P.V.; Meredith, D.D. Discrete Differential Dynamic Programing Approach to Water Resources Systems Optimization. Water Resour. Res. 1971, 7, 273–282. [Google Scholar] [CrossRef]
  53. Labadie, J.W. Optimal Operation of Multireservoir Systems: State-of-the-Art Review. J. Water Resour. Plan. Manag. 2004, 130, 93–111. [Google Scholar] [CrossRef]
  54. Smakhtin, V.U.; Eriyagama, N. Developing a Software Package for Global Desktop Assessment of Environmental Flows. Environ. Model. Softw. 2008, 23, 1396–1406. [Google Scholar] [CrossRef]
  55. Sahoo, S.; K Mishra, P.; Behera, S.; Krishan, R. A Comparative Study on Environmental Flows Assessment Methods in Lower Reach of Mahanadi River. Int. J. Eng. Trends Technol. 2016, 32, 82–90. [Google Scholar] [CrossRef]
  56. Li, C.; Kang, L. A New Modified Tennant Method with Spatial-Temporal Variability. Water Resour. Manag. 2014, 28, 4911–4926. [Google Scholar] [CrossRef]
  57. Karakoyun, Y.; Yumurtaci, Z.; Dönmez, A.H. Environmental Flow Assessment for Energy Generation Sustainability Employing Different Hydraulic Evaluation Methods: Çambaşi Hydropower Plant Case Study in Turkey. Clean Technol. Environ. Policy 2016, 18, 583–591. [Google Scholar] [CrossRef]
  58. Sibao, C.; Yanlei, W.; Guangfu, D.; Yuan, W.; Fei, W.; Yan, W.; Chengjian, X.; Xiao, D. Research on the Improved Ecological Water Level Calculation Method Based on Tennant Method. In Proceedings of the 2023 International Conference on Networking, Informatics and Computing (ICNETIC), Palermo, Italy, 29–31 May 2023; pp. 87–91. [Google Scholar]
  59. Coello Coello, C.A. Theoretical and Numerical Constraint-Handling Techniques Used with Evolutionary Algorithms: A Survey of the State of the Art. Comput. Methods Appl. Mech. Eng. 2002, 191, 1245–1287. [Google Scholar] [CrossRef]
  60. Yeniay, Ö. Penalty Function Methods for Constrained Optimization with Genetic Algorithms. Math. Comput. Appl. 2005, 10, 45–56. [Google Scholar] [CrossRef]
  61. Hınçal, O.; Altan-Sakarya, A.B.; Metin Ger, A. Optimization of Multireservoir Systems by Genetic Algorithm. Water Resour. Manag. 2011, 25, 1465–1487. [Google Scholar] [CrossRef]
  62. Ma, H.; Zhang, Y.; Sun, S.; Liu, T.; Shan, Y. A Comprehensive Survey on NSGA-II for Multi-Objective Optimization and Applications. Artif. Intell. Rev. 2023, 56, 15217–15270. [Google Scholar] [CrossRef]
  63. Blank, J.; Deb, K. Pymoo: Multi-Objective Optimization in Python. IEEE Access 2020, 8, 89497–89509. [Google Scholar] [CrossRef]
  64. Blank, J.; Deb, K.; Dhebar, Y.; Bandaru, S.; Seada, H. Generating Well-Spaced Points on a Unit Simplex for Evolutionary Many-Objective Optimization. IEEE Trans. Evol. Comput. 2021, 25, 48–60. [Google Scholar] [CrossRef]
  65. MoAF—Ministry of Agriculture and Forestry, G.D. of W.M. (SYGM). The Sakarya River Basin Drought Management Plan; Ministry of Agriculture and Forestry: Ankara, Türkiye, 2023. [Google Scholar]
  66. TÜBİTAK (The Scientific and Technological Research Council of Türkiye). Watershed Protection Action Plan for Sakarya Basin, Final Report; TÜBİTAK: Istanbul, Türkiye, 2013. [Google Scholar]
  67. Harmancioglu, N.B.; Altinbilek, D. Water Resources of Turkey; Springer Nature: Cham, Switzerland, 2019; ISBN 978-3-030-11728-3. [Google Scholar]
  68. Isik, S.; Dogan, E.; Kalin, L.; Sasal, M.; Agiralioglu, N. Effects of Anthropogenic Activities on the Lower Sakarya River. CATENA 2008, 75, 172–181. [Google Scholar] [CrossRef]
  69. Arslan, F.; Değirmenci, H.; Kartal, S.; Alcon, F. Mapping Performance of Irrigation Schemes in Turkey. Agron. Res. 2020, 18, 202–218. [Google Scholar] [CrossRef]
  70. Cakmak, B.; Kibaroglu, A.; Kendirli, B.; Gokalp, Z. Assessment of the Irrigation Performance of Transferred Schemes in Turkey: A Case Study Analysis. Irrig. Drain. 2010, 59, 138–149. [Google Scholar] [CrossRef]
  71. Saatci, A.M. Solving Water Problems of a Metropolis. J. Water Resour. Prot. 2013, 5, 7–10. [Google Scholar] [CrossRef]
  72. Cuceloglu, G.; Ozturk, I. Development of a Model Framework for Sustainable Water Management Practices: Case Study for the Megacity Istanbul. In Proceedings of the 9th Eastern European Young Water Professionals Conference, Budapest, Hungary, 24–27 May 2017. [Google Scholar]
  73. Gonenc I,E. Interbasin Water Transfer Practices in Turkey. J. Ecosyst. Ecography 2014, 4, 149. [Google Scholar] [CrossRef]
  74. Akalın, N.; Mertoğlu, B.; Ertürk, A. Assessing the Future Water Potential of Istanbul and the Need for Inter-Basin Water Transfer and the Trade-Offs for Water Allocation. Water Supply 2025, ws2025038. [Google Scholar] [CrossRef]
  75. Republic of Turkey the Ministry of Forestry and Water Affairs General Directorate of Water Management. Climate Change Impacts On Water Resources Project, Final Report—Executive Summary; Republic of Turkey the Ministry of Forestry and Water Affairs General Directorate of Water Management: Ankara, Türkiye, 2016.
  76. TSMS (Turkish State of Meteorological Service). Turkish National Climate Reports; TSMS (Turkish State of Meteorological Service): Ankara, Türkiye, 2016.
  77. Yilmaz, F.; Osborn, D.; Tsamados, M. Strengthening Türkiye’s Drought Management: Insights from International Practices (Netherlands, UK and USA). Front. Clim. 2025, 6, 1504779. [Google Scholar] [CrossRef]
  78. Bayazit, M. Systems Approach to Management of the Sakarya River Basin. Int. J. Water Resour. Dev. 1983, 1, 323–330. [Google Scholar] [CrossRef]
  79. Yaykiran, S.; Cuceloglu, G.; Ekdal, A. Estimation of Water Budget Components of the Sakarya River Basin by Using the WEAP-PGM Model. Water 2019, 11, 271. [Google Scholar] [CrossRef]
  80. Alp, E.; Özcan, Z.; Özcan, C.; Özel, B.; Başkan, O.; Demir, Y.; Erdoğan, E.K.; Köksal, M.A. Evaluation of the Water-Energy-Food Nexus with the Focus on Climate Change in Sakarya Watershed, Turkey. In Proceedings of the 93rd Water Environment Federation Technical Exhibition and Conference, USA, Virtual, 5–9 October 2020. [Google Scholar]
  81. Özcan, Z.; Köksal, M.A.; Alp, E. Evaluation of Water–Energy Nexus in Sakarya River Basin, Turkey. In Frontiers in Water-Energy-Nexus—Nature-Based Solutions, Advanced Technologies and Best Practices for Environmental Sustainability; Springer International Publishing: Cham, Switzerland, 2020; pp. 421–424. [Google Scholar]
  82. Reddy, M.J.; Nagesh Kumar, D. Multi-objective Particle Swarm Optimization for Generating Optimal Trade-offs in Reservoir Operation. Hydrol. Process. 2007, 21, 2897–2909. [Google Scholar] [CrossRef]
  83. Abdi, R.; Yasi, M. Evaluation of Environmental Flow Requirements Using Eco-Hydrologic–Hydraulic Methods in Perennial Rivers. Water Sci. Technol. 2015, 72, 354–363. [Google Scholar] [CrossRef]
  84. Uday Kumar, A.; Jayakumar, K.V. Modelling of Environmental Flow Requirements Using Hydraulic and Habitation Models. Ecol. Indic. 2021, 121, 107046. [Google Scholar] [CrossRef]
  85. Audet, C.; Bigeon, J.; Cartier, D.; Le Digabel, S.; Salomon, L. Performance Indicators in Multiobjective Optimization. Eur. J. Oper. Res. 2021, 292, 397–422. [Google Scholar] [CrossRef]
  86. Zitzler, E.; Deb, K.; Thiele, L. Comparison of Multiobjective Evolutionary Algorithms: Empirical Results. Evol. Comput. 2000, 8, 173–195. [Google Scholar] [CrossRef] [PubMed]
  87. Ishibuchi, H.; Masuda, H.; Tanigaki, Y.; Nojima, Y. Modified Distance Calculation in Generational Distance and Inverted Generational Distance. In Evolutionary Multi-Criterion Optimization; Gaspar-Cunha, A., Henggeler Antunes, C., Coello, C.C., Eds.; Lecture Notes in Computer Science; Springer International Publishing: Cham, Switzerland, 2015; Volume 9019, pp. 110–125. ISBN 978-3-319-15891-4. [Google Scholar]
  88. Lopez, E.M.; Coello, C.A.C. An Improved Version of a Reference-Based Multi-Objective Evolutionary Algorithm Based on IGD+. In Proceedings of the Genetic and Evolutionary Computation Conference, Kyoto, Japan, 15–19 July 2018; ACM: New York, NY, USA, 2018; pp. 713–720. [Google Scholar]
  89. Blank, J.; Deb, K. A Running Performance Metric and Termination Criterion for Evaluating Evolutionary Multi- and Many-Objective Optimization Algorithms. In Proceedings of the 2020 IEEE Congress on Evolutionary Computation (CEC), Glasgow, UK, 19–24 July 2020; pp. 1–8. [Google Scholar]
  90. Schott, J.R. Fault Tolerant Design Using Single and Multicriteria Genetic Algorithm Optimization. Master’s Thesis, Massachusetts Institute of Technology, Department of Aeronautics and Astronautics, Cambridge, MA, USA, 1995. [Google Scholar]
  91. Ilich, N. Shortcomings of Linear Programming in Optimizing River Basin Allocation. Water Resour. Res. 2008, 44, 2007WR006192. [Google Scholar] [CrossRef]
Figure 1. Overview of the SUWO Framework: This diagram illustrates the four key stages and their interrelationships.
Figure 1. Overview of the SUWO Framework: This diagram illustrates the four key stages and their interrelationships.
Water 17 01280 g001
Figure 2. SUWO structure workflow, highlighting the key stages and interactions between the scripts.
Figure 2. SUWO structure workflow, highlighting the key stages and interactions between the scripts.
Water 17 01280 g002
Figure 3. SUWO network structure, reservoir storage, and distribution node representation.
Figure 3. SUWO network structure, reservoir storage, and distribution node representation.
Water 17 01280 g003
Figure 4. Flowchart illustrating the general framework of NSGA-II and NSGA-III, highlighting the core components and diversity management mechanisms common to both.
Figure 4. Flowchart illustrating the general framework of NSGA-II and NSGA-III, highlighting the core components and diversity management mechanisms common to both.
Water 17 01280 g004
Figure 5. Schematic representation of the SRB, integrating key nodes, sub-basins, water infrastructure, and the SUWO-SRB model structure.
Figure 5. Schematic representation of the SRB, integrating key nodes, sub-basins, water infrastructure, and the SUWO-SRB model structure.
Water 17 01280 g005
Figure 6. Performance evaluation of NSGA-II and NSGA-III. This figure compares the performance of algorithms using various metrics, including hypervolume; IGD+, and running IGD, and an overall performance radar chart.
Figure 6. Performance evaluation of NSGA-II and NSGA-III. This figure compares the performance of algorithms using various metrics, including hypervolume; IGD+, and running IGD, and an overall performance radar chart.
Water 17 01280 g006
Figure 7. Comparison of Pareto fronts: (a) 3D Pareto front visualization comparing NSGA-II and NSGA-III, highlighting trade-offs among deviation from natural flow, irrigation deficits, and potential production shortfalls; (b) 2D projections of the Pareto front, showing pairwise comparisons of all objectives; (c) parallel coordinate plot illustrating the distribution and relationships among the objectives.
Figure 7. Comparison of Pareto fronts: (a) 3D Pareto front visualization comparing NSGA-II and NSGA-III, highlighting trade-offs among deviation from natural flow, irrigation deficits, and potential production shortfalls; (b) 2D projections of the Pareto front, showing pairwise comparisons of all objectives; (c) parallel coordinate plot illustrating the distribution and relationships among the objectives.
Water 17 01280 g007
Figure 8. Multi-criteria decision making with pseudo-weights for NSGA-II and NSGA-III. This figure shows the distribution of solutions for (a) NSGA-II and (b) NSGA-III across four management scenarios. The arrows within the plots indicate the preferred direction of improvement for each objective function.
Figure 8. Multi-criteria decision making with pseudo-weights for NSGA-II and NSGA-III. This figure shows the distribution of solutions for (a) NSGA-II and (b) NSGA-III across four management scenarios. The arrows within the plots indicate the preferred direction of improvement for each objective function.
Water 17 01280 g008
Figure 9. Hydropower energy generation (GWh) for HEPPs in the SRB cascade system under different management scenarios. (a) Simulated monthly energy generation for individual HEPPs. (b) Comparison of total monthly energy production across the entire cascade system, categorized into firm and secondary energy production. (c) Aggregated annual energy generation.
Figure 9. Hydropower energy generation (GWh) for HEPPs in the SRB cascade system under different management scenarios. (a) Simulated monthly energy generation for individual HEPPs. (b) Comparison of total monthly energy production across the entire cascade system, categorized into firm and secondary energy production. (c) Aggregated annual energy generation.
Water 17 01280 g009
Figure 10. Simulated monthly irrigation water supply and corresponding demand range across sub-basins in the SRB.
Figure 10. Simulated monthly irrigation water supply and corresponding demand range across sub-basins in the SRB.
Water 17 01280 g010
Figure 11. Optimized monthly reservoir storage distributions for the SRB cascade system under different hydrological conditions. Boxplots show storage variations, while shaded gray areas indicate operational constraints, highlighting minimum and maximum limits.
Figure 11. Optimized monthly reservoir storage distributions for the SRB cascade system under different hydrological conditions. Boxplots show storage variations, while shaded gray areas indicate operational constraints, highlighting minimum and maximum limits.
Water 17 01280 g011
Figure 12. Simulated monthly outflow at each node under different EMCs. Observed outflow is included for comparison, illustrating seasonal variations and the impact of flow modifications across management scenarios.
Figure 12. Simulated monthly outflow at each node under different EMCs. Observed outflow is included for comparison, illustrating seasonal variations and the impact of flow modifications across management scenarios.
Water 17 01280 g012
Figure 13. Effect of optimized operations on SRB downstream flow. Gray markers show observed means, green markers indicate EMC-B targets, and purple markers represent simulated means and distributions, highlighting seasonal flow adjustments.
Figure 13. Effect of optimized operations on SRB downstream flow. Gray markers show observed means, green markers indicate EMC-B targets, and purple markers represent simulated means and distributions, highlighting seasonal flow adjustments.
Water 17 01280 g013
Figure 14. Comparison of WEF-E scores across various water management scenarios, including NSGA-II, NSGA-III, hydrological conditions, and EMCs.
Figure 14. Comparison of WEF-E scores across various water management scenarios, including NSGA-II, NSGA-III, hydrological conditions, and EMCs.
Water 17 01280 g014
Table 1. Overview of the process steps in the SUWO workflow.
Table 1. Overview of the process steps in the SUWO workflow.
Process StepScriptDescription
1Data Managementexport_data.subEdit and export input datasets using Data Manager.xlsm.
2Data Loadingload_data.pyPreprocess input datasets for the model.
3Model Parametersmain.pyDefine the model parameters nodes, periods, etc.
4Optimization Setupmain.pyDefine the problem, objectives, and algorithm parameters
5Optimization Executionproblem.py, pymooRun optimization using pymoo and evaluate solutions
6Simulation Functionssimulation.pyPredefined functions for water management simulation
7Result Exportsimulation.pyExport simulation results and optimum solutions
8Result Importsimulation.pyObtain optimization results and history generation data
9Analysis and Visualizationanalysis.py, pymooPareto fronts, performance metrics, simulations, scenarios
10Analysis Functionsanalysis.pyPredefined functions for result analysis and visualization
Table 2. SUWO-SRB model nodes and simulated water components for near-future reference.
Table 2. SUWO-SRB model nodes and simulated water components for near-future reference.
IDNode NameNatural
Inflow
(MCM/Year)
Generated
Energy (GWh/Year)
Irrigation
Water (MCM/Year)
Evaporation
(MCM/Year)
Outflow (MCM/Year)
01 UPSUpper Sakarya 979.0-363.327.951.04
02 PORPorsuk Sub-basin487.9-246.525.315.53
03 ANKAnkara Sub-basin357.6-84.19.340.57
04 GURGursogut Reservoir 87.4287.4-18.8112.86
05 KARKargi Reservoir 2.3270.3-1.4112.74
06 KIRKirmir Sub-basin601.7-64.517.727.79
07 SARSariyar Reservoir 451.6409.4-59.0171.29
08 GOKGokcekaya Reservoir 164.0509.8-16.3180.20
09 YENYenice Reservoir 13.8126.0-3.2179.93
10 MDSMiddle Sakarya 375.1-184.312.1204.08
11 GKSGoksu Sub-basin 649.6-110.15.444.03
12 PMKPamukova Irrigation 35.6-34.40.0247.42
13 LWSLower Sakarya2280.9-41.03.8412.37
Table 3. Parameter settings for NSGA-II and NSGA-III in the SUWO-SRB model.
Table 3. Parameter settings for NSGA-II and NSGA-III in the SUWO-SRB model.
Parameter/AlgorithmNSGA-IINSGA-III
Population Size10001000
Number of Variables312312
Number of Function Evaluations10001000
Sampling MethodFloat random samplingFloat random sampling
SelectionTournament selectionEnergy-based, 3 objectives
CrossoverSBX1, η: 15, probability: 0.9SBX1, η: 30, probability: 1
MutationPM2, η: 20PM2, η: 20
Note: SBX1: simulated binary crossover; PM2: polynomial mutation.
Table 4. Penalty weight coefficients and their respective ranges used in the SUWO-SRB model for optimizing water management strategies.
Table 4. Penalty weight coefficients and their respective ranges used in the SUWO-SRB model for optimizing water management strategies.
CoefficientRangeDescription
k s c a l e 10−7–10−6Overall penalty scaling
c m i n 0.9–2.0Penalty weight for dev. from minimum storage level
c m a x 0.9–2.0Penalty weight for dev. from maximum storage level
c t a r g e t 0.9–10Penalty weight for dev. from target storage level
Table 5. Comparative performance analysis of NSGA-II and NSGA-III algorithms on the SUWO-SRB model. Bold values highlight better performance for each performance metric.
Table 5. Comparative performance analysis of NSGA-II and NSGA-III algorithms on the SUWO-SRB model. Bold values highlight better performance for each performance metric.
Performance MetricCategoryNSGA-IINSGA-IIIIdeal Range
H V , ¯   H V f (Hypervolume)Convergence, Diversity3.341 1, 4.053 22.860 1, 3.786 2Max, (0 → ∞)
, ¯ f (Running Inv. Gen. Dist.) Convergence, Diversity0.184 1, 0.014 20.192 1, 0.013 2Min, (∞ → 0)
I G D + ¯ , I G D f + (Modified Inv. Gen. Dist.)Convergence, Diversity0.155 1, 0.044 20.180 1, 0.014 2Min, (∞ → 0)
UPM (Uniformity Performance Meas.)Diversity0.0100.009Min, (∞ → 0)
SPM (Spread Performance Measure)Diversity1.1810.796Max, (0 → ∞)
NDSGR (N-dom. Size Generation Rate)Non-dominance0.9240.501Max, (0 → 1)
MCT (Mean Computational Time)Computational Efficiency6.5936.468Min, (∞ → 0)
Note: 1 Average value; 2 final value.
Table 6. Management scenarios and pseudo-weights for the WEF-E nexus.
Table 6. Management scenarios and pseudo-weights for the WEF-E nexus.
Management
Scenario
DescriptionEnergyFoodEcology
Balanced AllocationMaintain a balance between three objectives0.330.330.33
Maximizing EnergyFocus on maximizing energy generation while allowing for some degradation in ecology0.800.100.10
Sustainable IrrigationPrioritize sustainable irrigation practices that minimize water usage0.100.800.10
Eco-Centric ApproachPrioritize ecological impact significantly over energy generation and water usage0.100.100.80
Table 7. Optimization results for the SUWO-SRB model under different management scenarios and hydrological conditions, highlighting hydropower generation, irrigation supply ratio, and ecological deviation.
Table 7. Optimization results for the SUWO-SRB model under different management scenarios and hydrological conditions, highlighting hydropower generation, irrigation supply ratio, and ecological deviation.
Scenario Hydropower
Generation
(GWh)
Irrigation
Supply Ratio
Ecological
Deviation
Management
Scenarios 1
(NSGA-II)
Balanced Allocation18360.96 0.13
Maximizing Energy19090.87 0.14
Sustainable Irrigation17491.00 0.08
Eco-Centric Approach17950.92 0.07
Management
Scenarios 2
(NSGA-III)
Balanced Allocation18460.96 0.11
Maximizing Energy18870.89 0.13
Sustainable Irrigation17940.99 0.08
Eco-Centric Approach17810.93 0.08
Hydrological
Conditions 3
Normal Year18390.99 0.12
Wet Year21961.09 0.49
Dry Year12720.86 0.14
Environment
Management Classes 4
A: Natural18820.90 0.08
B: Slightly Modified18460.96 0.11
C: Moderately Modified17931.00 0.49
D: Largely Modified17861.04 1.39
E: Seriously Modified17881.05 2.97
F: Critically Modified17631.06 5.46
Note: 1 Normal year, EMC-B; 2 normal year, EMC-B; 3 EMC-B, NSGA-II, Balanced Alloc.; 4 normal year, NSGA-II, Balanced Allocation.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Yaykiran, S.; Ekdal, A. Sustainable Water Optimization Tool (SUWO): An Optimization Framework for the Water–Energy–Food–Ecosystem Nexus. Water 2025, 17, 1280. https://doi.org/10.3390/w17091280

AMA Style

Yaykiran S, Ekdal A. Sustainable Water Optimization Tool (SUWO): An Optimization Framework for the Water–Energy–Food–Ecosystem Nexus. Water. 2025; 17(9):1280. https://doi.org/10.3390/w17091280

Chicago/Turabian Style

Yaykiran, Salim, and Alpaslan Ekdal. 2025. "Sustainable Water Optimization Tool (SUWO): An Optimization Framework for the Water–Energy–Food–Ecosystem Nexus" Water 17, no. 9: 1280. https://doi.org/10.3390/w17091280

APA Style

Yaykiran, S., & Ekdal, A. (2025). Sustainable Water Optimization Tool (SUWO): An Optimization Framework for the Water–Energy–Food–Ecosystem Nexus. Water, 17(9), 1280. https://doi.org/10.3390/w17091280

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop