Scrutinizing different predictive modeling validation methodologies and data-partitioning strategies: new insights using groundwater modeling case study.

Groundwater salinity is a critical factor affecting water quality and ecosystem health, with implications for various sectors including agriculture, industry, and public health. Hence, the reliability and accuracy of groundwater salinity predictive models are paramount for effective decision-making in managing groundwater resources. This pioneering study presents the validation of a predictive model aimed at forecasting groundwater salinity levels using three different validation methods and various data partitioning strategies. This study tests three different data validation methodologies with different data-partitioning strategies while developing a group method of data handling (GMDH)-based model for predicting groundwater salinity concentrations in a coastal aquifer system. The three different methods are the hold-out strategy (last and random selection), k-fold cross-validation, and the leave-one-out method. In addition, various combinations of data-partitioning strategies are also used while using these three validation methodologies. The prediction model's validation results are assessed using various statistical indices such as root mean square error (RMSE), means squared error (MSE), and coefficient of determination (R2). The results indicate that for monitoring wells 1, 2, and 3, the hold-out (random) with 40% data partitioning strategy gave the most accurate predictive model in terms of RMSE statistical indices. Also, the results suggested that the GMDH-based models behave differently with different validation methodologies and data-partitioning strategies giving better salinity predictive capabilities. In general, the results justify that various model validation methodologies and data-partitioning strategies yield different results due to their inherent differences in how they partition the data, assess model performance, and handle sources of bias and variance. Therefore, it is important to use them in conjunction to obtain a comprehensive understanding of the groundwater salinity prediction model's behavior and performance.

Management of saltwater intrusion using 3D numerical modelling: a first for Pacific Island country of Vanuatu.

Small island countries like Vanuatu are facing the brunt of climate change, sea level rise (SLR), tropical cyclones, and limited or declining access to freshwater. The Tagabe coastal aquifer in Port Vila (the capital of Vanuatu) shows the presence of salinity, indicating saltwater intrusion (SWI). This study aims to develop and evaluate effective SWI management strategies for Tagabe coastal aquifer. To manage SWI, the numerical simulation model for the study area was developed using the SEAWAT code. The flow model was developed using MODFLOW and the transport model was developed using MT3DMS. Whereby SEAWAT solved flow and transport equations simultaneously. The model was calibrated, and different scenarios were evaluated for the management of SWI. The SLR was also considered in the model simulations. The results indicated that increased population, pumping rates, and SLR affect the SWI rates. To manage the SWI, we introduced hydraulic barriers like barrier wells and injection wells which effectively managed SWI in Tagabe coastal aquifer. The results from this study are significantly important whereby, the water managers, site owners, and governing bodies can use the management strategies presented in this study to create policies and regulations for managing SWI rates in Port Vila. Additionally, the water industry, private businesses, and investors who wish to extract groundwater from the Tagabe can use this study as a reference for daily or yearly freshwater production rates without the risk of SWI.

Integrating numerical modelling and scenario-based sensitivity analysis for saltwater intrusion management: case study of a complex heterogeneous island aquifer system.

Population growth, industrialisation and increasing agricultural demands have significantly stressed groundwater resources in Pacific Island countries (PICs). Climate change and sea-level rise also affect the groundwater resources in PICs. These anthropogenic and natural factors give rise to saltwater intrusion (SWI), a major growing environmental problem in the PICs. SWI is a highly non-linear process which makes it more complex to manage. However, with the help of numerical modelling, SWI can be monitored, managed and controlled. In the present study, we used an illustrative study area where the hydrogeological parameters and other boundary conditions used are similar to the PICs aquifer systems in Vanuatu. The scenarios include changing the barrier wells, injection wells, recharge, hydraulic head, hydraulic conductivity and grid size. The numerical simulation model of the study area was developed, and different scenarios were tested using SEAWAT modules. Apart from salt, we also modelled leachate and engine oil present in the investigated study area to see how it affects the freshwater wells over time. The scenario-based sensitivity analysis tests indicate that injection wells, recharge and hydraulic conductivities are highly sensitive, and with the proper modification, SWI can be managed or regulated. The sensitivity of grid size showed that the simulated results varied within the 10% range of different gird sizes. Moreover, it was also found that the rise in sea level or coastal heads by 0.3-1 m does not significantly cause further SWI encroachment in aquifers. The results from this study are very crucial in this modern era when freshwater needs in coastal areas, especially PICs, are rapidly increasing, and fresh groundwater resources are declining. The novel outcome presented in this study opens pathways for further detailed modelling and numerical studies in the field of SWI management strategy development and is, therefore, beneficial for policymakers, groundwater modellers and general scientific communities.

An integrated simulation-optimization framework for assessing environmental flows in rivers.

The present study proposes an integrated simulation-optimization framework to assess environmental flow by mitigating environmental impacts on the surface and ground water resources. The model satisfies water demand using surface water resources (rivers) and ground water resources (wells). The outputs of the ecological simulation blocks of river ecosystem and the ground water level simulation were utilized in a multiobjective optimization model in which six objectives were considered in the optimization model including (1) minimizing losses of water supply (2) minimizing physical fish habitat losses simulated by fuzzy approach (3) minimizing spawning habitat losses (4) minimizing ground water level deterioration simulated by adaptive neuro fuzzy inference system(ANFIS) (5) maximizing macroinvertebrates population simulated by ANFIS (6) minimizing physical macrophytes habitat losses. Based on the results in the case study, ANFIS-based model is robust for simulating key factors such as water quality and macroinvertebrate's population. The results demonstrate the reliability and robustness of the proposed method to balance environmental requirements and water supply. The optimization model increased the percentage of environmental flow in the drought years considerably. It supplies 69% of water demand in normal years, while the environmental impacts on the river ecosystem are minimized. The proposed model balances the portion of using surface water and ground water in water supply considering environmental impacts on both sources. Using the proposed method is recommendable for optimal environmental management of surface water and ground water in river basin scale.

Source Characterization of Multiple Reactive Species at an Abandoned Mine Site Using a Groundwater Numerical Simulation Model and Optimization Models.

The most important first step in the management and remediation of contaminated groundwater aquifers is unknown contaminant source characterization. Often, the hydrogeological field data available for accurate source characterization are very sparse. In addition, hydrogeological and geochemical parameter estimates and field measurements are uncertain. Particularly in complex contaminated sites such as abandoned mine sites, the geochemical processes are very complex and identifying the sources of contamination in terms of location, magnitude, and duration, and determination of the pathways of pollution become very difficult. The reactive nature of the contaminant species makes the geochemical transport process very difficult to model and predict. Additionally, the source identification inverse problem is often non-unique and ill posed. This study is about developing and demonstrating a source characterization methodology for a complex contaminated aquifer with multiple reactive species. This study presents linked simulation optimization-based methodologies for characterization of unknown groundwater pollution source characteristics, i.e., location, magnitude and duration or timing. Optimization models are solved using an adaptive simulated annealing (ASA) optimization algorithm. The performance of the developed methodology is evaluated for different complex scenarios of groundwater pollution such as distributed mine waste dumps with reactive chemical species. The method is also applied to a real-life contaminated aquifer to demonstrate the potential applicability and optimal characterization results. The illustrative example site is a mine site in Northern Australia that is no longer active.

Application of Monitoring Network Design and Feedback Information for Adaptive Management of Coastal Groundwater Resources.

Optimal strategies for the management of coastal groundwater resources can be derived using coupled simulation-optimization based management models. However, the management strategy actually implemented on the field sometimes deviates from the recommended optimal strategy, resulting in field-level deviations. Monitoring these field-level deviations during actual implementation of the recommended optimal management strategy and sequentially updating the management model using the feedback information is an important step towards efficient adaptive management of coastal groundwater resources. In this study, a three-phase adaptive management framework for a coastal aquifer subjected to saltwater intrusion is applied and evaluated for a regional-scale coastal aquifer study area. The methodology adopted includes three sequential components. First, an optimal management strategy (consisting of groundwater extraction from production and barrier wells) is derived and implemented for optimal management of the aquifer. The implemented management strategy is obtained by solving a homogenous ensemble-based coupled simulation-optimization model. Second, a regional-scale optimal monitoring network is designed for the aquifer system considering possible user noncompliance of a recommended management strategy, and uncertainties in estimating aquifer parameters. A new monitoring network design objective function is formulated to ensure that candidate monitoring wells are placed in high risk (highly contaminated) locations. In addition, a new methodology is utilized to select candidate monitoring wells in areas representative of the entire model domain. Finally, feedback information in the form of measured concentrations obtained from the designed optimal monitoring wells is used to sequentially modify pumping strategies for future time periods in the management horizon. The developed adaptive management framework is evaluated by applying it to the Bonriki aquifer system located in Kiribati, which is a small developing island country in the South Pacific region. Overall, the results from this study suggest that the implemented adaptive management strategy has the potential to address important practical implementation issues arising due to noncompliance of an optimal management strategy and uncertain aquifer parameters.

Multi-objective groundwater management strategy under uncertainties for sustainable control of saltwater intrusion: Solution for an island country in the South Pacific.

To date, simulation-optimization (S/O) based groundwater management models have delivered optimal saltwater intrusion management strategies for coastal aquifer systems. At times, however, uncertainties in the numerical simulation model due to uncertain aquifer parameters are not incorporated into the management model. The present study explicitly incorporated aquifer parameter uncertainty into a multi-objective management model for the optimal design of groundwater pumping strategies from the unconfined Bonriki aquifer situated in a small Pacific island country. The objective of the multi-objective management model was to maximise pumping from production wells and minimize pumping from the barrier wells (hydraulic barriers) to ensure that the water quality at different monitoring locations (MLs) were within pre-specified sustainable limits. To achieve the targeted management goal, a coupled flow and transport numerical simulation model of the Bonriki aquifer was developed using the FEMWATER numerical code. The developed three-dimensional numerical model was calibrated and validated using limited available hydrological data. To achieve computational efficiency and feasibility of the management model, the numerical simulation model in the S/O model was replaced with ensembles of Support Vector Machine Regression (SVMR) surrogate models. Each SVMR standalone surrogate model in the ensemble was constructed using datasets from different numerical simulation models with different hydraulic conductivity and porosity values. These ensemble SVMR models were coupled to the multi-objective genetic algorithm optimization model to solve the Bonriki aquifer management problem. The executed optimization model presented a Pareto-front with 600 non-dominated optimal trade-off pumping solutions. The reliability of the management model established after validation of the optimal solution results suggests that the implemented constraints of the optimization problem were satisfied, i.e., the salinity concentrations at respective MLs were within the pre-specified limits. Overall, the results from this study indicated that the developed management model has the potential to address groundwater salinity problems in small island countries.

Sequential optimal monitoring network design and iterative spatial estimation of pollutant concentration for identification of unknown groundwater pollution source locations.

One of the difficulties in accurate characterization of unknown groundwater pollution sources is the uncertainty regarding the number and the location of such sources. Only when the number of source locations is estimated with some degree of certainty that the characterization of the sources in terms of location, magnitude, and activity duration can be meaningful. A fairly good knowledge of source locations can substantially decrease the degree of nonuniqueness in the set of possible aquifer responses to subjected geochemical stresses. A methodology is developed to use a sequence of dedicated monitoring network design and implementation and to screen and identify the possible source locations. The proposed methodology utilizes a combination of spatial interpolation of concentration measurements and simulated annealing as optimization algorithm for optimal design of the monitoring network. These monitoring networks are to be designed and implemented sequentially. The sequential design is based on iterative pollutant concentration measurement information from the sequentially designed monitoring networks. The optimal monitoring network design utilizes concentration gradient information from the monitoring network at previous iteration to define the objective function. The capability of the feedback information based iterative methodology is shown to be effective in estimating the source locations when no such information is initially available. This unknown pollution source locations identification methodology should be very useful as a screening model for subsequent accurate estimation of the unknown pollution sources in terms of location, magnitude, and activity duration.

Uncertainty based optimal monitoring network design for a chlorinated hydrocarbon contaminated site.

An application of a newly developed optimal monitoring network for the delineation of contaminants in groundwater is demonstrated in this study. Designing a monitoring network in an optimal manner helps to delineate the contaminant plume with a minimum number of monitoring wells at optimal locations at a contaminated site. The basic principle used in this study is that the wells are installed where the measurement uncertainties are minimum at the potential monitoring locations. The development of the optimal monitoring network is based on the utilization of contaminant concentration data from an existing initial arbitrary monitoring network. The concentrations at the locations that were not sampled in the study area are estimated using geostatistical tools. The uncertainty in estimating the contaminant concentrations at such locations is used as design criteria for the optimal monitoring network. The uncertainty in the study area was quantified by using the concentration estimation variances at all the potential monitoring locations. The objective function for the monitoring network design minimizes the spatial concentration estimation variances at all potential monitoring well locations where a monitoring well is not to be installed as per the design criteria. In the proposed methodology, the optimal monitoring network is designed for the current management period and the contaminant concentration data estimated at the potential observation locations are then used as the input to the network design model. The optimal monitoring network is designed for the consideration of two different cases by assuming different initial arbitrary existing data. Three different scenarios depending on the limit of the maximum number of monitoring wells that can be allowed at any period are considered for each case. In order to estimate the efficiency of the developed optimal monitoring networks, mass estimation errors are compared for all the three different scenarios of the two different cases. The developed methodology is useful in coming up with an optimal number of monitoring wells within the budgetary limitations. The methodology also addresses the issue of redundancy, as it refines the existing monitoring network without losing much information of the network. The concept of uncertainty-based network design model is useful in various stages of a potentially contaminated site management such as delineation of contaminant plume and long-term monitoring of the remediation process.

Global optimal design of ground water monitoring network using embedded kriging.

We present a methodology for global optimal design of ground water quality monitoring networks using a linear mixed-integer formulation. The proposed methodology incorporates ordinary kriging (OK) within the decision model formulation for spatial estimation of contaminant concentration values. Different monitoring network design models incorporating concentration estimation error, variance estimation error, mass estimation error, error in locating plume centroid, and spatial coverage of the designed network are developed. A big-M technique is used for reformulating the monitoring network design model to a linear decision model while incorporating different objectives and OK equations. Global optimality of the solutions obtained for the monitoring network design can be ensured due to the linear mixed-integer programming formulations proposed. Performances of the proposed models are evaluated for both field and hypothetical illustrative systems. Evaluation results indicate that the proposed methodology performs satisfactorily. These performance evaluation results demonstrate the potential applicability of the proposed methodology for optimal ground water contaminant monitoring network design.