Comparative Evaluation of Genetic Programming and Neural Network as Potential Surrogate Models for Coastal Aquifer Management

Determining the optimal rates of groundwater extraction for the sustainable use of coastal aquifers is a complex water resources management problem. It necessitates the application of a 3D simulation model for coupled flow and transport simulation together with an optimization algorithm in a linked simulation-optimization framework. The use of numerical models for aquifer simulation within optimization models is constrained by the huge computational burden involved. Approximation surrogates are widely used to replace the numerical simulation model, the widely used surrogate model being Artificial Neural Networks (ANN). This study evaluates genetic programming (GP) as a potential surrogate modeling tool and compares the advantages and disadvantages with the neural network based surrogate modeling approach. Two linked simulation optimization models based on ANN and GP surrogate models are developed to determine the optimal groundwater extraction rates for an illustrative coastal aquifer. The surrogate models are linked to a genetic algorithm for optimization. The optimal solutions obtained using the two approaches are compared and the advantages of GP over the ANN surrogates evaluated.     

Stochastic and Robust Multi-Objective Optimal Management of Pumping from Coastal Aquifers Under Parameter Uncertainty

Combined simulation-optimization approaches have been used as tools to derive optimal groundwater management strategies to maintain or improve water quality in contaminated or other aquifers. Surrogate models based on neural networks, regression models, support vector machies etc., are used as substitutes for the numerical simulation model in order to reduce the computational burden on the simulation-optimization approach. However, the groundwater flow and transport system itself being characterized by uncertain parameters, using a deterministic surrogate model to substitute it is a gross and unrealistic approximation of the system. Till date, few studies have considered stochastic surrogate modeling to develop groundwater management methodologies. In this study, we utilize genetic programming (GP) based ensemble surrogate models to characterize coastal aquifer water quality responses to pumping, under parameter uncertainty. These surrogates are then coupled with multiple realization optimization for the stochastic and robust optimization of groundwater management in coastal aquifers. The key novelty in the proposed approach is the capability to capture the uncertainty in the physical system, to a certain extent, in the ensemble of surrogate models and using it to constrain the optimization search to derive robust optimal solutions. Uncertainties in hydraulic conductivity and the annual aquifer recharge are incorporated in this study. The results obtained indicate that the methodology is capable of developing reliable and robust strategies for groundwater management.     

Coastal Aquifer Management Based on the Joint use of Density-Dependent and Sharp Interface Models

In pumping optimization of coastal aquifers, the evaluation of the objective function and constraints using density-dependent models is overwhelmed by complex and time-consuming numerical simulations. To address those cases where the available density-dependent model runs are very limited, due to excessive computational burden, an efficient optimization strategy is developed. The proposed methodology uses an efficient sharp interface model jointly with a complex density-dependent model in an evolutionary optimization algorithm. While most evaluations are based on the sharp interface model, the density-dependent model is selectively called to evaluate promising solutions and to improve the predictions of the sharp interface model through the adaptive modification of the saltwater-freshwater density ratio. The method is tested for pumping optimization problems in confined and unconfined coastal aquifers with multiple pumping wells. The optimal solutions are compared to those obtained by density-dependent as well as by sharp interface optimization alone. Under a very restrictive computational budget, the best feasible solution is attained in less than 25 density-dependent model runs for two optimization problems of 10 and 20 decision variables. The results indicate that this optimization method leads to good feasible solutions and that an improved estimation of optimal pumping rates can be achieved within a limited computational budget. The method could also stand as an efficient preliminary exploration of the optimal search space, to provide good feasible starting points for the implementation of more comprehensive methods of coastal aquifer management.

     

A Probabilistic Multiperiod Simulation–Optimization Approach for Dynamic Coastal Aquifer Management

Combined simulation–optimization (CSO) schemes are common in the literature to solve different groundwater management problems, and CSO is particularly well-established in the coastal aquifer management literature. However, with a few exceptions, nearly all previous studies have employed the CSO approach to derive static groundwater management plans that remain unchanged during the entire management period, consequently overlooking the possible positive impacts of dynamic strategies. Dynamic strategies involve division of the planning time interval into several subintervals or periods, and adoption of revised decisions during each period based on the most recent knowledge of the groundwater system and its associated uncertainties. Problem structuring and computational challenges seem to be the main factors preventing the widespread implementation of dynamic strategies in groundwater applications. The objective of this study is to address these challenges by introducing a novel probabilistic Multiperiod CSO approach for dynamic groundwater management. This includes reformulation of the groundwater management problem so that it can be adapted to the multiperiod CSO approach, and subsequent employment of polynomial chaos expansion-based stochastic dynamic programming to obtain optimal dynamic strategies. The proposed approach is employed to provide sustainable solutions for a coastal aquifer storage and recovery facility in Oman, considering the effect of natural recharge uncertainty. It is revealed that the proposed dynamic approach results in an improved performance by taking advantage of system variations, allowing for increased groundwater abstraction, injection and hence monetary benefit compared to the commonly used static optimization approach.

     

Simulation-Optimization Modeling for Sustainable Groundwater Development: A Moroccan Coastal Aquifer Case Study

A transient simulation model characterizing groundwater flow in the coastal aquifer of Rhis-Nekor was constructed and calibrated. The flow model was then used in conjunction with a genetic algorithm based optimization model to explore the optimal pumping schemes that meet current and future water demands while minimizing the risks for several adverse environmental impacts, such as saltwater intrusion prevention, avoiding excessive drawdown, as well as controlling waterlogging and salinity problems. Modeling results demonstrate the importance of this combined simulation-optimization methodology for solving groundwater management problems associated with the Rhis-Nekor plain.     

Optimal Management of Coastal Aquifers Using Linked Simulation Optimization Approach

Saltwater intrusion management models can be used to derive optimal and efficient management strategies for controlling saltwater intrusion in coastal aquifers. To obtain physically meaningful optimal management strategies, the physical processes involved need to be simulated while deriving the management strategies. The flow and transport processes involved in coastal aquifers are difficult to simulate especially when the density-dependent flow and transport processes need to be modeled. Incorporation of this simulation model within an optimization-based management model is very complex and difficult. However, as an alternative, it is possible to link a simulation model externally with an optimization-based management model. The GA-based optimization approach is especially suitable for externally linking the numerical simulation model within the optimization model. Further efficiency in computational procedure can be achieved for such a linked model, if the simulation process can be simplified by approximation, as very large number of iterations between the optimization and simulation model is generally necessary to evolve an optimal management strategy. A possible approach for approximating the simulation model is to use a trained Artificial Neural Network (ANN) as the approximate simulator. Therefore, an ANN model is trained as an approximator of the three dimensional density-dependent flow and transport processes in a coastal aquifer. A linked simulation – optimization model is then developed to link the trained ANN with the GA-based optimization model for solving saltwater management problems. The performance of the developed optimization model is evaluated using an illustrative study area. The evaluation results show the potential applicability of the developed methodology using a GA- and ANN-based linked optimization – simulation model for optimal management of coastal aquifer.     

Planning of Groundwater Supply Systems Subject to Uncertainty Using Stochastic Flow Reduced Models and Multi-Objective Evolutionary Optimization

The typical modeling approach to groundwater management relies on the combination of optimization algorithms and subsurface simulation models. In the case of groundwater supply systems, the management problem may be structured into an optimization problem to identify the pumping scheme that minimizes the total cost of the system while complying with a series of technical, economical, and hydrological constraints. Since lack of data on the subsurface system most often reflects upon the development of groundwater flow models that are inherently uncertain, the solution to the groundwater management problem should explicitly consider the tradeoff between cost optimality and the risk of not meeting the management constraints. This work addresses parameter uncertainty following a stochastic simulation (or Monte Carlo) approach, in which a sufficiently large ensemble of parameter scenarios is used to determine representative values selected from the statistical distribution of the management objectives, that is, minimizing cost while minimizing risk. In particular, the cost of the system is estimated as the expected value of the cost distribution sampled through stochastic simulation, while the risk of not meeting the management constraints is quantified as the expected value of the intensity of constraint violation. The solution to the multi-objective optimization problem is addressed by combining a multi-objective evolutionary algorithm with a stochastic model simulating groundwater flow in confined aquifers. Evolutionary algorithms are particularly appropriate in optimization problems characterized by non-linear and discontinuous objective functions and constraints, although they are also computationally demanding and require intensive analyses to tune input parameters that guarantee optimality to the solutions. In order to drastically reduce the otherwise overwhelming computational cost, a novel stochastic flow reduced model is thus developed, which practically allows for averting the direct inclusion of the full simulation model in the optimization loop. The computational efficiency of the proposed framework is such that it can be applied to problems characterized by large numbers of decision variables.     

Coupled Quantity-Quality Simulation-Optimization Model for Conjunctive Surface-Groundwater Use

Many water resources optimization problems involve conflicting objectives which the main goal is to find a set of optimal solutions on, or near to, Pareto front. In this study a multi-objective water allocation model was developed for optimization of conjunctive use of surface water and groundwater resources to achieve sustainable supply of agricultural water. Here, the water resource allocation model is based on simulation-optimization (SO) modeling approach. Two surrogate models, namely an Artificial Neural Network model for groundwater level simulation and a Genetic Programming model for TDS concentration prediction were coupled with NSGA-II. The objective functions involved: 1) minimizing water shortage relative to the water demand, 2) minimizing the drawdown of groundwater level, and 3) minimizing the groundwater quality changes. According to the MSE and R² criteria, the results showed that the surrogate models for prediction of groundwater level and TDS concentration performed favorably in comparison to the measured values at the number of observation wells. In Najaf Abad plain case study, the average drawdown was limited to 0.18 m and the average TDS concentration also decreased from 1257 mg/lit to 1229 mg/lit under optimal conditions.     

Pumping Optimization of Coastal Aquifers Assisted by Adaptive Metamodelling Methods and Radial Basis Functions

The application of metamodelling frameworks is a popular approach to handle the computational cost arising from complex computer simulations and global optimization algorithms in simulation-optimization routines. In this paper, Radial Basis Functions (RBF) are used as metamodels for the computationally expensive variable-density flow and salt transport numerical simulations, in a pumping optimization problem of coastal aquifers. While RBF metamodels have been fairly utilized in many engineering optimization problems, their use is very limited in coastal aquifer management. Two adaptive metamodelling frameworks are employed, that is, the adaptive-recursive approach and the metamodel-embedded evolution strategy. In both frameworks, cubic RBF models are used to approximate the constraint functions imposed on the coastal aquifer pumping optimization problem. The optimal pumping rates are first calculated based on the variable-density and salt transport numerical models of seawater intrusion. The resulting optimal solutions and the computational times are set as benchmark values in order to assess the performance of the metamodelling optimization strategies. Results indicate that the metamodel-embedded evolution framework outperformed in terms of computational efficiency the adaptive-recursive approach while it successfully located the region of the global optimum. Furthermore, with the metamodel-embedded evolution strategy the computational time of the variable-density-based optimization was reduced by 96 %.     

An Artificial Intelligence Approach for the Stochastic Management of Coastal Aquifers

Aquifer recharge rates and patterns are often uncertain, especially in arid areas due to sporadic and erratic rainfall. Therefore, determining the optimal groundwater abstraction using classical approaches such as Monte Carlo Simulation (MCS) requires a large number of groundwater simulations and exorbitant computational efforts. The problem becomes even more complex and time consuming for regional coastal aquifers whose domains must be discretized using high-resolution meshes. In fact, even fast evolutionary multi-objective optimization techniques generally require a large number of simulations to determine the Pareto-front among the objectives. This study explores the performance of a Decision Tree (DT) approach for the generation of the Pareto optimal solutions of groundwater extraction. This paper applies the DTs for the optimal management of the Al-Khoud coastal aquifer in Oman. The learning process of the developed DT-based model uses the output of a numerical simulation model to assess the aquifer response based on different abstraction policies. The trained DT network then utilizes the NSGA-II to determine the Pareto-optimal solutions. The simulation show that the general flux pattern in the study area is toward the sea and the hydraulic head following a similar pattern in both best and worst recharging scenarios downstream of the studied recharging dam. Statistical tests showed a good correlation between the DT-based and simulation-based results and demonstrate the capability of the DT approach to obtain high-quality solutions by incorporating a large number of recharge scenarios. Moreover, the required runtime of the DT-based approach is extremely low (5 min) compared to that of the simulation-based method (several days). This means that including additional Monte-Carlo simulations can be readily done in few minutes using the obtained DTs, instead of the long computational time needed by the simulation-based approach.     

Development of an Improved Pollution Source Identification Model Using Numerical and ANN Based Simulation-Optimization Model

The identification of unknown pollution sources is an important and challenging task for the engineers working on pollution management of a groundwater aquifer. The locations and transient magnitude of unknown contaminant sources can be identified using inverse optimization technique. In this approach, the absolute difference between the simulated and the observed contaminant concentration at the observation locations of the aquifer is minimized by using an optimization algorithm. The simulated concentrations is calculated using the aquifer simulation model. As such, there is a need to incorporate the aquifer simulation model with the optimization model. Thus the performance of the model is highly related to the aquifer simulation model. The incorporation of the sophisticated numerical simulation model will give better performance, but the model will be computationally expensive. On the other hand, the model will be computationally less expensive if an approximate simulation model is used in place of the numerical simulation model. However, in this case, the predictive performance of the model will decline. For achieving efficiency in both computational time as well as in predicting the performance, this study presents a new genetic algorithms based simulation-optimization method incorporating both the numerical and the approximate simulation models. The efficiency and field applicability of the model is demonstrated using illustrative study areas. The performance evaluation of the model shows that the proposed model has the potential for real-world field applications.     

云神经网络确定含水层参数研究

高效、精确的含水层参数求解方法一直是水文地质研究领域的重要研究内容之一。实践中通常利用非稳定流抽水试验资料通过配线法确定含水层参数,但是随着计算机应用的普及,已有人开发出几种在非稳定流试验条件下求解含水层水文地质参数的快速、精确的计算机智能优化算法。在此基础上尝试建立了云神经网络模型(Cloud Neural Net,CNN),并将其应用于石家庄市元氏县3个单孔非稳定流抽水试验,对承压含水层参数进行计算,模型计算结果与当地的水文地质条件较为符合,且比传统方法及单纯的人工神经网络模型所得结果更加精确。因此云神经网络模型为研究区地下水资源评价、地下水数值模拟以及溶质运移模拟提供了另一种重要手段。     

A Stochastic Optimization Approach in the Design of an Aquifer Remediation under Hydrogeologic Uncertainty

A stochastic optimization approach is presented for the remediation design of a contaminated aquifer with limited hydrogeologic information. Stochastic simulation using the Monte Carlo technique, produces a series of equally probable realisations of the spatially varying random hydraulic conductivity field. The stochastic flow and transport simulation model is coupled, using the response matrix approach, with a nonlinear optimization algorithm. The whole process is integrated into an algorithm which is effectively applied in the case study of the Kalamaria aquifer, Chalkidiki, Greece. The stochastic optimization procedure is followed by a reliability analysis, giving useful information to the decision makers concerning the effectiveness of the optimal results.     

Coupled Sharp-interface and Density-dependent Model for Simultaneous Optimization of Production Well Locations and Pumping in Coastal Aquifer

The main management challenge in coastal aquifers is to prevent saltwater intrusion, ensuring ample freshwater supply. Saltwater intrusion happens due to unregulated pumping from production wells. Therefore, it is essential to have an effective management policy, which ensures the requisite amount of freshwater to be withdrawn from coastal aquifers without causing saltwater intrusion. A methodology for optimizing production well locations and maximizing pumping from production wells is presented to achieve these conflicting objectives. The location of production wells directly affects the amount of freshwater pumped out of the coastal aquifer. Simultaneous optimization of production well locations and pumping from the same is achieved by linking mathematical simulation models with the optimization algorithm. A new methodology using coupled sharp-interface and density-dependent simulation models is developed to find optimal well locations and optimize the amount of freshwater pumped from the coastal aquifer. The performance of the developed methodology is evaluated for saltwater intrusion in the coastal city of Puri, India. The performance evaluation results show the developed methodology's applicability for managing saltwater intrusion while maximizing freshwater pumping in coastal aquifers under constraints of well location.

     

Evaluation of Analytical Methods to Study Aquifer Properties with Pumping Tests in Coastal Aquifers with Numerical Modelling (Motril-Salobreña Aquifer)

Two pumping tests were performed in the unconfined Motril-Salobreña detrital aquifer in a 250 m-deep well 300 m from the coastline containing both freshwater and saltwater. It is an artesian well as it is in the discharge zone of this coastal aquifer. The two observation wells where the drawdowns are measured record the influence of tidal fluctuations, and the well lithological columns reveal high vertical heterogeneity in the aquifer. The Theis and Cooper-Jacob approaches give average transmissivity (T) and storage coefficient (S) values of 1460 m²/d and 0.027, respectively. Other analytical solutions, modified to be more accurate in the boundary conditions found in coastal aquifers, provide similar T values to those found with the Theis and Cooper-Jacob methods, but give very different S values or could not estimate them. Numerical modelling in a synthetic model was applied to analyse the sensitivity of the Theis and Cooper-Jacob approaches to the usual boundary conditions in coastal aquifers. The T and S values calculated from the numerical modelling drawdowns indicate that the regional flow, variable pumping flows, and tidal effect produce an error of under 10 % compared to results obtained with classic methods. Fluids of different density (freshwater and saltwater) cause an error of 20 % in estimating T and of over 100 % in calculating S. The factor most affecting T and S results in the pumping test interpretation is vertical heterogeneity in sediments, which can produce errors of over 100 % in both parameters.     

基于遗传算法的变密度条件下地下水模拟优化模型

将遗传算法和变密度地下水流及溶质运移模拟程序SEAWAT耦合起来,开发了一个新的用于地下水模拟优化管理的通用程序——SWTGA。以求解变密度条件下地下水优化管理问题,从而为地下水管理决策者提供科学依据和技术支持。设计SWTGA时,建立了适用于变密度条件下地下水优化管理常见问题的目标函数的一般形式,同时设定了常用的约束条件。最后将SWTGA程序应用于一个理想滨海含水层中地下水开采方案的优化设计,寻优之后获得了最佳开采方案,与未优化开采方案的对比显示优化结果合理可行,验证了SWTGA模拟优化程序的有效性和可靠性。     

基于RBF代理模型和粒子群算法的水交换优化研究

良好的水交换对改善水环境、提高水域周边景观效果等具有十分重要的作用。而引水置换的方法对促进水交换效果显著,但利用数值模拟进行水交换研究耗时长、效率低,且人为改变参数的局限性大,不利于寻找最优的换水方案。为解决这一问题,基于径向基函数(简称RBF)代理模型建立水交换优化模型,并通过粒子群算法求最优解。以某人工岛游艇别墅区港池初拟方案为例,验证该方法的可行性和优越性。算例结果表明:（1）构建的基于RBF代理模型的水交换优化模型精度较高;（2）基于RBF代理模型的水交换优化模型计算1次所需时间量级为秒,而传统数值模拟计算的量级为小时;（3）通过粒子群算法,对建立的基于RBF代理模型的水交换优化模型求解,得到研究区域的最优换水方案。上述最优方案的结果与MIKE21水动力和对流扩散模型的计算结果相符。     

Multi-Objective Simulation-Optimization Model for Long-term Reservoir Operation using Piecewise Linear Hedging Rule

An efficiently parameterized and appropriately structured piecewise linear hedging rule is formulated and included within a multi-objective simulation-optimization (S-O) framework that seeks to obtain Pareto-optimal solutions for the long-term hedged operation of a single water supply reservoir. Two conflicting objectives, namely, “minimize the total shortage ratio” and “minimize the maximum shortage” are considered in the S-O framework, while explicit specification of constraints is avoided in the optimization module. Evolutionary search based non-dominated sorting genetic algorithm is used as the driver, which is linked to the simulation engine that invokes the piecewise linear hedging rule within the S-O framework. Preconditioning of the multi-objective stochastic search of the time-varying piecewise linear hedging model is effected by feeding initial feasible solutions sampled from the Pareto-optimal front of a simple constant hedging parameter model, which has resulted in significant improvement of the Pareto-optimality and the computational efficiency.

     

Simulation and Multi-Objective Management of Coastal Aquifers in Semi-Arid Regions

Groundwater is the main water resource in many semi-arid coastal regions and water demand, especially in summer months, can be very high. Groundwater withdrawal for meeting this demand often causes seawater intrusion and degradation of water quality of coastal aquifers. In order to satisfy demand, a combined management plan is proposed and is under consideration for the island of Santorini. The plan involves: (1) desalinization (if needed) of pumped water to a potable level using reverse osmosis and (2) injection into the aquifer of biologically-treated waste water. The management plan is formulated in a multi-objective, optimization framework, where simultaneous minimization of economic and environmental costs is desired, subject to a constraint so that cleaned water satisfies demand. The decision variables concern the well locations and the corresponding pumping and recharging rates. The problem is solved using a computationally efficient, multi-objective, genetic algorithm (NSGAII). The constrained multi-objective, optimization problem is transformed to an unconstrained one using a penalty function proportional to constraint violation. This extends the definition of the objective function outside the domain of feasibility. The impact of prolonged droughts on coastal aquifers is investigated by assuming various scenarios of reduced groundwater recharge. Water flow and quality in the coastal aquifer is simulated using a three-dimensional, variable density, finite difference model (SEAWAT). The method is initially applied to a test aquifer and the trade-off curves (Pareto fronts) are determinedl for each drought scenario. The trade-off curves indicate an increase on the economic and environmental cost as groundwater recharge reduces due to climate change.     

A Cost-Effective Method to Control Seawater Intrusion in Coastal Aquifers

Intrusion of seawater into coastal aquifers is considered one of the most important processes that degrade water-quality by raising the salinity to levels exceeding acceptable drinking standards. Therefore saltwater intrusion should be prevented or at least controlled to protect groundwater resources. This paper presents a cost-effective method to control seawater intrusion in coastal aquifers. This methodology ADR (Abstraction, Desalination and Recharge) includes; abstraction of saline water and recharge to the aquifer after desalination. A coupled transient density-dependent finite element model is developed for simulation of fluid flow and solute transport and used to simulate seawater intrusion. The simulation model has been integrated with an optimization model to examine three scenarios to control seawater intrusion including; abstraction, recharge and a combination system, ADR. The main objectives of the models are to determine the optimal depths, locations and abstraction/recharge rates for the wells to minimize the total costs for construction and operation as well as salt concentrations in the aquifer. A comparison between the combined system (ADR) and the individual abstraction or recharge system is made in terms of total cost and total salt concentration in the aquifer and the amount of repulsion of seawater achieved. The results show that the proposed ADR system performs significantly better than using abstraction or recharge wells alone as it gives the least cost and least salt concentration in the aquifer. ADR is considered an effective tool to control seawater intrusion and can be applied in areas where there is a risk of seawater intrusion.