Explainable ML modeling of saltwater intrusion control with underground barriers in coastal sloping aquifers.

Controlling seawater intrusion (SWI) into freshwater aquifers is crucial for preserving water quality in coastal groundwater management. This research evaluates the performance of three machine learning (ML) models: eXtreme Gradient Boosting (BO-XGB), Light Gradient Boosting Machine (BO-LGB), and Categorical Gradient Boosting (BO-CGB) in predicting the SWI wedge length. A database of 345 numerical simulations was compiled from previous research, and Bayesian Optimization (BO) with fivefold cross-validation was used to fine-tune the models. The inputs included abstraction well distance (Xa), abstraction well depth (Ya), recharge well distance (Xr), recharge well depth (Yr), abstraction rate (Qa), artificial recharge rate (Qr), and SWI wedge length (L). Results show that BO-CGB consistently achieved the best performance, with high R2 values (0.996 in training and 0.969 in testing) and low RMSE values (0.439 m in training and 1.327 m in testing). SHapley Additive exPlanations (SHAP) analysis highlighted that Qa and Qr had the most significant impact on SWI wedge length predictions, followed by Xa and Ya. Partial Dependence Plot (PDP) analysis revealed a strong negative correlation between flow variables Qa and Qr and wedge length, while Xr displayed a more complex, non-linear pattern. BO-CGB emerged as the most reliable model for predicting SWI wedge length. To facilitate practical application, an interactive Graphical User Interface (GUI) was developed, enabling users to input variables and receive instant predictions, enhancing the practical usability of the ML models in managing SWI in coastal aquifers.

Managing saltwater intrusion (SWI) in coastal aquifers is critical for safeguarding freshwater quality and ensuring sustainable water resources. This study evaluates the performance of eight machine learning (ML) models in predicting the SWI wedge length ratio (L/Lo) in sloping coastal aquifers. The assessed models encompassed linear, bagging, boosting, and advanced gradient boosting-based approaches, enabling a comprehensive comparison of their predictive capabilities. First, a numerical dataset of 450 samples was compiled, incorporating key dimensionless input variables such as relative density, hydraulic conductivity ratio, bed slope, and recharge well properties. The dataset was split into training and testing subsets in a 70:30 ratio, and model hyperparameters were optimized using Bayesian Optimization (BO). A thorough evaluation was conducted to identify the best-performing predictive model. Results showed that the Extreme Gradient Boosting (XGB) model demonstrated superior predictive accuracy compared to all other models, achieving low root-mean-square-error (RMSE) values of 0.0216 during training and 0.0331 during testing, along with high R2 scores of 0.9801 and 0.9586, respectively. The Categorical Gradient Boosting (CGB) model also exhibited strong performance, with RMSE values of 0.0271 (training) and 0.0316 (testing). SHapley Additive exPlanations (SHAP) analysis revealed that the relative recharge well rate was the most influential predictor, followed by recharge well distance and depth. To facilitate practical application, desktop and web-based graphical user interfaces (GUIs) were developed, allowing users to input variables and effortlessly predict L/L₀. This study demonstrates the effectiveness of ML models in predicting SWI in sloping coastal aquifers and provides user-friendly tools for engineers and researchers.

Barrier walls are considered one of the most effective methods for facilitating the retreat of saltwater intrusion (SWI). This research plans to examine the effect of using barrier walls for controlling of SWI in sloped unconfined aquifers. The sloping unconfined aquifer is considered with three different bed slopes. The SEAWAT model is implemented to simulate the SWI. For model validation, the numerical results of the seawater wedge at steady state were compared with the analytical solution. Increasing the ratio of flow barrier depth (db/d) forced the saltwater interface to move seaward and increased the repulsion ratio (R). With a positive sloping bed, further embedding the barrier wall from 0.2 to 0.7 caused R to increase from 0.3% to 59%, while it increased from 1.8% to 41.7% and from 3.4% to 46.9% in the case of negative and horizontal slopes, respectively. Embedding the barrier wall to a db/d value of more than 0.4 achieved a greater R value in the three bed-sloping cases. Installing the barrier wall near the saltwater side with greater depth contributed to the retreat of the SWI. With a negative bed slope, moving the barrier wall from Xb/Lo = 1.0 toward the saltwater side (Xb/Lo = 0.2) increased R from 7.21% to 68.75%, whereas R increased from 5.3% to 67% for the horizontal sloping bed and from 5.1% to 64% for the positive sloping bed. The numerical results for the Akrotiri coastal aquifer confirm that the embedment of the barrier wall significantly affects the controlling of SWI by increasing the repulsion ratio (R) and decreasing the SWI length ratio (L/La). Cost-benefit analysis is recommended to determine the optimal design of barrier walls for increasing the cost-effectiveness of the application of barrier walls as a countermeasure for controlling and preventing SWI in sloped unconfined aquifers.

Optimization of milling conditions for AISI 4140 steel using an integrated machine learning-multi objective optimization-multi criteria decision making framework

The quality of groundwater resources in coastal aquifers is affected by saltwater intrusion. Over-abstraction of groundwater and seawater level rise due to climate change accelerate the intrusion process. This paper investigates the effects of aquifer bed slope and seaside slope on saltwater intrusion. The possible impacts of increasing seawater head due to sea level rise and decreasing groundwater level due to over-pumping and reduction in recharge are also investigated. A numerical model (SEAWAT) is applied to well-known Henry problem to assess the movement of the dispersion zone under different settings of bed and seaside slopes. The results showed that increasing seaside slope increased the intrusion of saltwater by 53.2% and 117% for slopes of 1:1 and 2:1, respectively. Increasing the bed slope toward the land decreased the intrusion length by 2% and 4.8%, respectively. On the other hand, increasing the bed slope toward the seaside increased the intrusion length by 3.6% and 6.4% for bed slopes of 20:1 and 10:1, respectively. The impacts of reducing the groundwater level at the land side and increasing the seawater level at the shoreline by 5% and 10% considering different slopes are studied. The intrusion length increased under both conditions. Unlike Henry problem, the current investigation considers inclined beds and sea boundaries and, hence, provides a better representation of the field conditions.

The demand for freshwater is very high in the coastal regions due to the high population density in coastal areas. To meet this demand for freshwater, the coastal aquifers are often heavily pumped without any regulation, resulting in saltwater intrusion. Therefore, the biggest challenge in the management of coastal aquifer is to meet the demand for freshwater by pumping the coastal aquifer without causing saltwater intrusion. In this study, a brief overview of various methods for identification, prediction, and management of saltwater intrusion is presented. Detection of saltwater intrusion is largely hindered due to insufficient spatiotemporal monitoring because of budgetary constraints. Application, merits, and demerits of the newer cost-effective techniques as well as conventional techniques for identifying saltwater intrusion are discussed in this chapter. The application of various prediction models and their computational difficulties is also presented in this study. Finally, advanced techniques for identification and sustainable management practice in saltwater intrusion are discussed. Though significant progress has been made in the recent past in the management of coastal aquifers, they still show gaps in addressing real-life scenarios. An attempt has been made to highlight the suitability of a developed methodology and their respective limitations.

The need for freshwater is emerging as the utmost critical resource issue facing humanity. In several arid and semi-arid parts of the world, groundwater resources are being used as an alternative source of freshwater. Excessive and/or unplanned groundwater withdrawals have a negative impact on the aquifer. Groundwater withdrawn from coastal aquifers are susceptible to contamination by saltwater intrusion. This study investigates the efficiency and viability of using artificial freshwater recharge (AFR) to increase fresh groundwater pumping from production wells for beneficial use. A three dimensional (3D), transient, density dependent, finite element based flow and transport model of an illustrative coastal aquifer is implemented using FEMWATER code. First, the effect of AFR on inland encroachment of saline water is quantified for existing scenarios. Specifically, groundwater head and salinity concentration differences at monitoring locations before and after artificial recharge is presented. Second, a multi-objective management model incorporating groundwater pumping and AFR is implemented to control groundwater salinization in an illustrative coastal aquifer system. To avoid computational burden and ensure computational feasibility, the numerical flow and transport simulation model is substituted by the new support vector regression (SVR) predictive models as approximate simulators in the simulation–optimization framework for developing optimal management strategies. The performance evaluation results indicated that the SVR models were adequately trained and were capable of approximating saltwater intrusion processes in the aquifer. Multi-objective genetic algorithm (MOGA) is used to solve the multi-objective optimization problem. The Pareto-optimal front obtained as solution from the SVR–MOGA optimization model presented a set of optimal solutions needed for the sustainable management of the coastal aquifer. The pumping strategies obtained as Pareto optimal solutions with and without freshwater recharge wells showed that saltwater intrusion is sensitive to the AFR. Also, the hydraulic head lenses created by AFR can be used as one practical option to control saltwater intrusion in coastal aquifers. The developed 3D saltwater intrusion model, predictive capability of the developed SVR models and the feasibility of using the proposed linked multi-objective SVR–MOGA optimization model makes the proposed methodology potentially attractive in solving large scale regional saltwater intrusion management problems.

Prediction of urban surface water quality scenarios using hybrid stacking ensembles machine learning model in Howrah Municipal Corporation, West Bengal

The accurate prediction of the strength enhancement ratio () and strain enhancement ratio (εcc/εco) in FRP-wrapped elliptical concrete columns is crucial for optimizing structural performance. This study employs machine learning (ML) techniques to enhance prediction accuracy and reliability. A dataset of 181 samples, derived from experimental studies and finite element modeling, was utilized, with a 70:30 train-test split (127 training samples and 54 testing samples). Four ML models: Decision Tree (DT), Adaptive Boosting (ADB), Stochastic Gradient Boosting (SGB), and Extreme Gradient Boosting (XGB) were trained and optimized using Bayesian Optimization to refine their hyperparameters and improve performance.Results demonstrate that SGB achieved the best performance for predicting , with an R2 of 0.850, the lowest RMSE (0.190), and the highest generalization capability, making it the most reliable model for strength enhancement predictions. For strain enhancement prediction (εcc/εco), XGB outperformed other models, achieving an R2 of 0.779 with the lowest RMSE (2.162), indicating a better balance between accuracy, generalization, and minimal overfitting. DT and ADB exhibited lower predictive performance, with higher residual errors and lower generalization capacity. Furthermore, Shapley Additive exPlanations analysis identified the FRP thickness-elastic modulus product (tf × Ef) and concrete compressive strength () as the most influential features impacting both enhancement ratios. To facilitate real-world applications, an interactive graphical user interface was developed, enabling engineers to input ten structural parameters and obtain real-time predictions.

Seawater intrusion occurs commonly in coastal aquifers around the world, threatening the availability and usability of fresh groundwater resources for vegetation and human uses. The rapid growth of the world population and urbanization requires sound strategies for protection and management of the freshwater resources, especially coastal groundwater. This goal can only be achieved through proper understanding of the processes that underlie seawater intrusion in coastal aquifers. Major insights have been gained over the past several decades, in particular, roles of the density-driven flow in driving and maintaining the invasive flow of saltwater. However, most studies have linked the seawater intrusion process merely to the level of salt content through the free convection induced by the salinity contrast between groundwater and seawater but ignored their temperature difference. In reality, the thermal contrast between coastal groundwater and marine seawater may range up to 15°C in absolute value with either warmer or colder seawater. Such thermal contrast can alter seawater circulation through the coastal aquifer, which in turn affects the biogeochemical reactions of land-sourced pollutants in the aquifer prior to discharge to the marine ecosystem.This research aimed to investigate the combined effects of salinity and temperature contrasts on the interactions between freshwater and seawater in unconfined coastal aquifers. Using laboratory experiments and numerical simulations, this research explored the coastal groundwater dynamics under various boundary settings with regards to thermal variations, tidal forcing and seasonal changes of seawater temperature. Findings from this research provided insights into the importance of temperature variations on various key processes in coastal aquifers under the condition of either static sea level or tidal oscillation.The effects of temperature contrast were first investigated for the seaward boundary of static level using physical experiments and numerical models in combination with tracer tracking. With the static sea level, the thermal contrast induced long-term impacts on the aquifer and altered background flow patterns and transport activities. The position of the saltwater wedge toe was modified significantly by the presence of temperature gradient either landward or seaward. Colder seawater enhanced the advancement of saltwater while warmer seawater hindered it. More importantly, the seawater circulation pattern changed dramatically in the latter case. A second circulation cell was discovered for the first time near the seaward boundary. The regular landward circulation cell was pushed to the vicinity of the interface where considerably larger velocity was observed. In-depth sensitivity analysis revealed the important role of spatial correlation between temperature-induced and salinity-induced density gradients, especially at the base of the aquifer, in driving the formation of the new cell.Both laboratory experiments and numerical simulation were carried out to investigate the thermal effects under the condition of tidal oscillation. The responses of the saltwater wedge were found to be similar to those under the static sea level, in particular, the retreat and advance of the wedge with warmer and colder seawater, respectively. The mixing zone widened as a result of the tidal fluctuation. Meanwhile, the upper saline plume and the freshwater discharge zone expanded in the warmer seawater case and contracted with colder seawater. The increased seawater temperature also intensified water exchange across aquifer-ocean interface, seawater circulation and the submarine groundwater discharge. Furthermore, tidally induced seawater circulation intensified with increased contribution to the submarine groundwater discharge compared with density-driven seawater circulation. All these characteristics were persistent over a range of tidal amplitudes. These results shed light on the importance of the thermal effects and have important implications for the assessment of the biogeochemical processes in coastal aquifers.The seasonal variations of the temperature contrast were then examined based on numerical simulations. The results showed clearly seasonality of the aquifer – ocean exchange and seawater circulation induced by the seasonal variation of seawater temperature in both cases with the static sea level and tidal conditions. Compared with the cases of the isothermal condition, all fluxes increased during colder months and decreased during warmer months. The periodic oscillation of the thermally induced density gradient resulted in a continuously changing mode of saltwater flow in the saltwater wedge. The flow path and transit time of circulating seawater shortened considerably in comparison with that in the isothermal case. This finding is particularly important for the evaluation of transport of land-sourced contaminants to the marine environment.The insights into the thermal effects on coastal unconfined aquifers gained from laboratory experiments and numerical simulations were applied to calculate a thermal impact factor and a thermal sensitivity index for aquifers along global coastlines based on local conditions of freshwater temperature and temperature contrast. The results suggested that the temperature effect is significant and would either amplify or reduce the impact of sea level rise on the vulnerability of coastal aquifers over a large proportion of the global coastlines.

Accurate prediction of crown convergence in Tunnel Boring Machine (TBM) tunnels is critical for ensuring construction safety, optimizing support design, and improving construction efficiency. This study proposes an interpretable machine learning method based on Bayesian optimization (BO) and SHapley Additive exPlanations (SHAP) for predicting crown convergence (CC) in TBM tunnels. Firstly, a dataset comprising 1,501 samples was constructed using tunnel engineering data. Then, six classical ML models, namely, Support Vector Regression, Decision Tree, Random Forest, Light Gradient Boosting Machine (LightGBM), eXtreme Gradient Boosting, and K-nearest neighbors—were developed, and BO was applied to tune the hyperparameters of each model to achieve accurate prediction of CC. Subsequently, the SHAP method was adopted to interpret the LightGBM model, quantifying the contribution of each input feature to the model’s predictions. The results indicate that the LightGBM model achieved the best prediction performance on the test set, with root mean squared error, mean absolute error, mean absolute percentage error, and determination coefficient values of 0.9122 mm, 0.6027 mm, 0.0644, and 0.9636, respectively; the average SHAP values for the six input features of the LightGBM model were ranked as follows: Time (0.1366) > Rock grade (0.0871) > Depth ratio (0.0528) > Still arch (0.0200) > Saturated compressive strength (0.0093) > Rock quality designation (0.0047). Validation using data from a TBM water conveyance tunnel in Xinjiang, China, confirmed the method’s practical utility, positioning it as an effective auxiliary tool for safer and more efficient TBM tunnel construction.

BackgroundTimely and accurate outcome prediction plays a critical role in guiding clinical decisions for hypertensive ischemic or hemorrhagic stroke patients admitted to the ICU. However, interpreting and translating the predictive models into clinical applications are as important as the prediction itself. This study aimed to develop an interpretable machine learning (IML) model that accurately predicts 28-day all-cause mortality in hypertensive ischemic or hemorrhagic stroke patients.MethodsA total of 4,274 hypertensive ischemic or hemorrhagic stroke patients admitted to the ICU in the USA from multicenter cohorts were included in this study to develop and validate the IML model. Five machine learning (ML) models were developed, including artificial neural network (ANN), gradient boosting machine (GBM), eXtreme Gradient Boosting (XGBoost), logistic regression (LR), and support vector machine (SVM), to predict mortality using the MIMIC-IV and eICU-CRD database in the USA. Feature selection was performed using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm. Model performance was evaluated based on the area under the curve (AUC), accuracy, positive predictive value (PPV), and negative predictive value (NPV). The ML model with the best predictive performance was selected for interpretability analysis. Finally, the SHapley Additive exPlanations (SHAP) method was employed to evaluate the risk of all-cause in-hospital mortality among hypertensive ischemic or hemorrhagic stroke patients admitted to the ICU.ResultsThe XGBoost model demonstrated the best predictive performance, with the AUC values of 0.822, 0.739, and 0.700 in the training, test, and external cohorts, respectively. The analysis of feature importance revealed that age, ethnicity, white blood cell (WBC), hyperlipidemia, mean corpuscular volume (MCV), glucose, pulse oximeter oxygen saturation (SpO2), serum calcium, red blood cell distribution width (RDW), blood urea nitrogen (BUN), and bicarbonate were the 11 most important features. The SHAP plots were employed to interpret the XGBoost model.ConclusionsThe XGBoost model accurately predicted 28-day all-cause in-hospital mortality among hypertensive ischemic or hemorrhagic stroke patients admitted to the ICU. The SHAP method can provide explicit explanations of personalized risk prediction, which can aid physicians in understanding the model.

Regional scale management of coastal aquifers for control of saltwater intrusion is a challenging problem, requiring solution of optimization and simulation models. Simulation of density dependent nonlinear flow and transport processes in a coastal aquifer requires the solution of coupled flow and transport equations. Prescription of optimal spatial and temporal management strategy for coastal aquifers is possible by utilizing a linked simulation-optimization approach. However, such linked models require the iterative and numerical simulation of the flow and transport processes numerous number of times within an optimization algorithm. In order to ensure computational feasibility and efficiency, trained and tested surrogate models with acceptable accuracy and efficiency can be used as approximate simulators within an optimization algorithm. In this study, an efficient surrogate model based on ensemble of Adaptive Neuro-fuzzy Inference System (ANFIS) is developed and evaluated as an approximate simulator of the physical processes of a multi-layered coastal aquifer. The management of coastal aquifers is also multiple-objective in nature. Therefore, the developed surrogate model is linked to a Controlled Elitist Multi-objective Genetic Algorithm (CEMGA). Ensembles of the surrogate models (En-ANFIS) are utilized in order to incorporate uncertainties in prediction using surrogate models. The proposed simulation-optimization framework is implemented in a parallel computing platform to achieve further computational efficiency. The performance of the multi-objective management model is evaluated for an illustrative study area. The evaluation results indicate that ANFIS based ensemble-modelling approach together with CEMGA is able to evolve reliable strategies for this multiple objective management of coastal aquifers.

A meta-ensemble machine learning strategy to assess groundwater holistic vulnerability in coastal aquifers

This study investigates the effectiveness of inclined double cutoff walls installed beneath hydraulic structures by employing five machine learning models: Random Forest(RF), Adaptive Boosting(AdaBoost), eXtreme Gradient Boosting(XGBoost), Light Gradient Boosting Machine(LightGBM), and Categorical Boosting (CatBoost). A comprehensive dataset of 630 samples was gathered from previous studies, including key input variables such as the relative distance between the cutoff wall and the structure's apron width (L/B), the inclination angle ratio between downstream and upstream cutoffs (θ2/θ1), the depth ratio of downstream to upstream cutoff walls (d2/d1), and the relative downstream cutoff depth to the permeable layer depth (d2/D). Outputs considered were the relative uplift force (U/Uo), the relative exit hydraulic gradient (iR/iRo), and the relative seepage discharge per unit structure length (q/qo). The dataset was split with a 70:30 ratio for training and testing. Hyperparameter optimization was conducted using Bayesian Optimization (BO) coupled with five-fold cross-validation to enhance model performance. Results showed that the CatBoost model demonstrated superior performance over other models, consistently yielding high R2 values, specifically surpassing 0.95, 0.93, and 0.97 for U/Uo, iR/iRo, and q/qo, respectively, along with low RMSE scores below 0.022, 0.089, and 0.019 for the same variables. A feature importance analysis is conducted using SHapley Additive exPlanations(SHAP) and Partial Dependence Plot (PDP). The analysis revealed that L/B was the most influential predictor for U/Uo and iR/iRo, while d2/D played a crucial role in determining q/qo. Moreover, PDPs illustrated a positive linear relationship between L/B and U/Uo, a V-shaped impact of d2/d1 on iR/iRo and q/qo, and complex nonlinear interactions for θ2/θ1 across all target variables. Furthermore, an interactive Graphical User Interface(GUI) was developed, enabling engineers to efficiently predict output variables and apply model insights in practical scenarios.