Groundwater Estimation from Major Physical Hydrology Components Using Artificial Neural Networks and Deep Learning

Hydrogeochemistry and prediction of arsenic contamination in groundwater of Vehari, Pakistan: comparison of artificial neural network, random forest and logistic regression models

Arsenic contamination in the groundwater occurs in various parts of the world due to anthropogenic and natural sources, adversely affecting human health and ecosystems. The current study intends to examine the groundwater hydrogeochemistry containing elevated arsenic (As), predict As levels in groundwater, and determine the aptness of groundwater for drinking in the Vehari district, Pakistan. Four hundred groundwater samples from the study region were collected for physiochemical analysis. As levels in groundwater samples ranged from 0.1 to 52 μg/L, with an average of 11.64 μg/L, (43.5%), groundwater samples exceeded the WHO 2022 recommended limit of 10 μg/L for drinking purposes. Ion-exchange processes and the adsorption of ions significantly impacted the concentration of As. The HCO3- and Na+ are the dominant ions in the study area, and the water types of samples were CaHCO3, mixed CaMgCl, and CaCl, demonstrating that rock-water contact significantly impacts hydrochemical behavior. The geochemical modeling indicated negative saturation indices with calcium carbonate and other salt minerals, encompassing aragonite, calcite, dolomite, and halite. The dissolution mechanism suggested that these minerals might have implications for the mobilization of As in groundwater. A combination of human-induced and natural sources of contamination was unveiled through principal component analysis (PCA). Artificial neural networks (ANN), random forest (RF), and logistic regression (LR) were used to predict As in the groundwater. The data have been divided into two parts for statistical analysis: 20% for testing and 80% for training. The most significant input variables for As prediction was determined using Chi-squared analysis. The receiver operating characteristic area under the curve and confusion matrix were used to evaluate the models; the RF, ANN, and LR accuracies were 0.89, 0.85, and 0.76. The permutation feature and mean decrease in impurity determine ten parameters that influence groundwater arsenic in the study region, including F-, Fe2+, K+, Mg2+, Ca2+, Cl-, SO42-, NO3-, HCO3-, and Na+. The present study shows RF is the best model for predicting groundwater As contamination in the research area. The water quality index showed that 161 samples represent poor water, and 121 samples are unsuitable for drinking. Establishing effective strategies and regulatory measures is imperative in Vehari to ensure the sustainability of groundwater resources.

Long short-term memory models to quantify long-term evolution of streamflow discharge and groundwater depth in Alabama

Long short-term memory (LSTM) models have been shown to be efficient for rainfall-runoff modeling, and to a lesser extent, for groundwater depth forecasting. In this study, LSTMs were applied to quantify the spatiotemporal evolution of surface and subsurface hydrographs in Alabama in the Southeastern United States, where water sustainability has not been fully quantified across spatiotemporal scales. First, the surface water LSTM model with extensive dynamic (precipitation and other weather variables) and static (basin characteristics) inputs predicted the main characteristics of streamflow for six years at 19 gauged basins in Alabama. The model tended to underestimate extremely high streamflow but adding drainage density as an input feature slightly improved the predictions of extreme events. Second, to predict the groundwater depth evolution, a groundwater LSTM (GW-LSTM) model was proposed and applied using static inputs capturing the aquifers' hydrogeological properties and dynamic inputs of meteorological information. Three precipitation scenarios were also explored to evaluate the groundwater hydrograph evolution in the next two decades. The GW-LSTM model predicted the general trend of daily groundwater depth fluctuations (at 21 wells distributed across Alabama from 1990 to 2021) including most extremely high groundwater levels, and recovered groundwater depth for locations withheld from model training and validation. This study, therefore, extended the application of LSTMs in quantifying the spatiotemporal evolution of surface water and groundwater, two manifestations of a single integrated resource.

Ensemble learning for multi-class COVID-19 detection from big data

Coronavirus disease (COVID-19), which has caused a global pandemic, continues to have severe effects on human lives worldwide. Characterized by symptoms similar to pneumonia, its rapid spread requires innovative strategies for its early detection and management. In response to this crisis, data science and machine learning (ML) offer crucial solutions to complex problems, including those posed by COVID-19. One cost-effective approach to detect the disease is the use of chest X-rays, which is a common initial testing method. Although existing techniques are useful for detecting COVID-19 using X-rays, there is a need for further improvement in efficiency, particularly in terms of training and execution time. This article introduces an advanced architecture that leverages an ensemble learning technique for COVID-19 detection from chest X-ray images. Using a parallel and distributed framework, the proposed model integrates ensemble learning with big data analytics to facilitate parallel processing. This approach aims to enhance both execution and training times, ensuring a more effective detection process. The model's efficacy was validated through a comprehensive analysis of predicted and actual values, and its performance was meticulously evaluated for accuracy, precision, recall, and F-measure, and compared to state-of-the-art models. The work presented here not only contributes to the ongoing fight against COVID-19 but also showcases the wider applicability and potential of ensemble learning techniques in healthcare.

Groundwater level response identification by hybrid wavelet-machine learning conjunction models using meteorological data

Due to its heterogeneous and complex nature, groundwater modeling needs great effort to quantify the aquifer, a crucial tool for policymakers and hydrogeologists to understand the variations in groundwater levels (GWL). This study proposed a set of supervised machine learning (ML) models to delineate the GWL changes in the Zarand-Saveh complex aquifer in Iran using 15-year (2005-2020) monthly dataset. The wavelet transform (WT) procedure was also used to improve the GWL prediction ability of ML models for 3-month horizons using input datasets of precipitation, evapotranspiration, temperature, and GWL. The four well-accepted standalone ML methods, i.e., artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), group method of data handling (GMDH), and least square support vector machine (LSSVM), were implemented and compared with the hybrid wavelet conjunction models. The methods were compared based on root mean square error (RMSE), mean absolute error (MAE), correlation coefficient (R), and Nash-Sutcliffe efficiency (NSE). Comparison outcomes showed that the hybrid wavelet-ML considerably improved the standalone model results. The wavelet transform-least square support vector machine (WT-LSSVM) model was superior to other standalone and hybrid wavelet-ML methods to predict GWL. The best GWL predictions were acquired from the WT-LSSVM model with input scenario 5 involving all influential variables, and this model produced RMSE, MAE, R, and NSE as 0.05, 0.04, 0.99, and 0.99 for 1 month ahead of GWL prediction, while the corresponding values were obtained as 0.18, 0.14, 0.95, and 0.90 for 3 months ahead of GWL prediction, respectively.

Estimation and uncertainty analysis of groundwater quality parameters in a coastal aquifer under seawater intrusion: a comparative study of deep learning and classic machine learning methods

Excessive withdrawal of groundwater for agricultural irrigation can cause seawater intrusion into coastal aquifers. Such a case will in turn results in deterioration of irrigation water quality. Determination of irrigation water quality with traditional methods is a time-consuming and costly process. However, machine learning algorithms can be useful tools for modeling and estimating groundwater quality used for irrigation water purposes. In this study, TDS, PS, SAR, and Cl parameters of groundwater were estimated with models based on EC and pH variables. For this purpose, prediction performances of two different deep learning methods (convolutional neural network (CNN) and deep neural network (DNN)) and two different classical machine learning (Random Forest (RF) and extreme gradient boosting (XGBoost)) methods were compared. In addition, predictive uncertainty of the models was determined by quantile regression (QR) analysis. Performance criteria and results of uncertainty analysis revealed that CNN (in testing phase, NSE = 0.95 for TDS, NSE = 0.96 for PS, NSE = 0.67 for SAR and NSE = 0.93 for CI) and DNN (in testing phase, NSE = 0.91 for TDS, NSE = 0.91 for PS, NSE = 0.57 for SAR and NSE = 0.94 for Cl) models had quite a close performance in estimation of TDS, PS, SAR, and Cl parameters and higher than the other two classical machine learning methods. As a result, the CNN model can be considered the best performing model in estimating all quality parameters due to the highest NSE and lowest RMSE values. In addition, the Taylor diagram showed that the values estimated using the CNN model had the highest correlation with the measured data. It was determined that the model with the lowest uncertainty based on the PICP statistics was DNN, followed by the CNN model. However, the CNN model has predicted outliers more accurately. Present findings proved that deep learning models could offer efficient tools for predicting irrigation water quality parameters.

Prediction of hydraulic blockage at culverts from a single image using deep learning

Cross-drainage hydraulic structures such as culverts and bridges in urban landscapes are prone to get blocked by the transported debris (e.g., urban, vegetated), which often reduces their hydraulic capacity and triggers flash floods. Unavailability of relevant data from blockage-originated flooding events and complex nature of debris accumulation are highlighted factors hindering the research within the blockage management domain. Wollongong City Council (WCC) blockage conduit policy is the leading formal guidelines to incorporate blockage into design guidelines; however, are criticized by the hydraulic engineers for its dependence on the post-flood visual inspections (i.e., visual blockage) instead of peak floods hydraulic investigations (i.e., hydraulic blockage). Apparently, no quantifiable relationship is reported between the visual blockage and hydraulic blockage; therefore, many consider WCC blockage guidelines invalid. This paper exploits the power of Artificial Intelligence (AI), motivated by its recent success, and attempts to relate visual blockage with hydraulic blockage by proposing a deep learning pipeline to predict hydraulic blockage from an image of the culvert. Two experiments are performed where the conventional pipeline and end-to-end learning approaches are implemented and compared in the context of predicting hydraulic blockage from a single image. In experiment one, the conventional deep learning pipeline approach (i.e., feature extraction using CNN and regression using ANN) is adopted. In contrast, in experiment two, end-to-end deep learning models (i.e., E2E_ MobileNet, E2E_ BlockageNet) are trained and compared with the conventional pipeline approach. Dataset (i.e., Hydraulics-Lab Blockage Dataset (HBD), Visual Hydraulics-Lab Dataset (VHD)) used in this research were collected from laboratory experiments performed using scaled physical models of culverts. E2E_ BlockageNet model was reported best in predicting hydraulic blockage with $$R^2$$ R 2 score of 0.91 and indicated that hydraulic blockage could be interrelated with the visual features at the culvert.

Investigation of Hyperparameter Setting of a Long Short-Term Memory Model Applied for Imputation of Missing Discharge Data of the Daihachiga River

Missing observational data pose an unavoidable problem in the hydrological field. Deep learning technology has recently been developing rapidly, and has started to be applied in the hydrological field. Being one of the network architectures used in deep learning, Long Short-Term Memory (LSTM) has been applied largely in related research, such as flood forecasting and discharge prediction, and the performance of an LSTM model has been compared with other deep learning models. Although the tuning of hyperparameters, which influences the performance of an LSTM model, is necessary, no sufficient knowledge has been obtained. In this study, we tuned the hyperparameters of an LSTM model to investigate the influence on the model performance, and tried to obtain a more suitable hyperparameter combination for the imputation of missing discharge data of the Daihachiga River. A traditional method, linear regression with an accuracy of 0.903 in Nash–Sutcliffe Efficiency (NSE), was chosen as the comparison target of the accuracy. The results of most of the trainings that used the discharge data of both neighboring and estimation points had better accuracy than the regression. Imputation of 7 days of the missing period had a minimum value of 0.904 in NSE, and 1 day of the missing period had a lower quartile of 0.922 in NSE. Dropout value indicated a negative correlation with the accuracy. Setting dropout as 0 had the best accuracy, 0.917 in the lower quartile of NSE. When the missing period was 1 day and the number of hidden layers were more than 100, all the compared results had an accuracy of 0.907–0.959 in NSE. Consequently, the case, which used discharge data with backtracked time considering the missing period of 1 day and 7 days and discharge data of adjacent points as input data, indicated better accuracy than other input data combinations. Moreover, the following information is obtained for this LSTM model: 100 hidden layers are better, and dropout and recurrent dropout levels equaling 0 are also better. The obtained optimal combination of hyperparameters exceeded the accuracy of the traditional method of regression analysis.

Development of a Deep Learning Emulator for a Distributed Groundwater–Surface Water Model: ParFlow-ML

Integrated hydrologic models solve coupled mathematical equations that represent natural processes, including groundwater, unsaturated, and overland flow. However, these models are computationally expensive. It has been recently shown that machine leaning (ML) and deep learning (DL) in particular could be used to emulate complex physical processes in the earth system. In this study, we demonstrate how a DL model can emulate transient, three-dimensional integrated hydrologic model simulations at a fraction of the computational expense. This emulator is based on a DL model previously used for modeling video dynamics, PredRNN. The emulator is trained based on physical parameters used in the original model, inputs such as hydraulic conductivity and topography, and produces spatially distributed outputs (e.g., pressure head) from which quantities such as streamflow and water table depth can be calculated. Simulation results from the emulator and ParFlow agree well with average relative biases of 0.070, 0.092, and 0.032 for streamflow, water table depth, and total water storage, respectively. Moreover, the emulator is up to 42 times faster than ParFlow. Given this promising proof of concept, our results open the door to future applications of full hydrologic model emulation, particularly at larger scales.

Can Deep Learning Extract Useful Information about Energy Dissipation and Effective Hydraulic Conductivity from Gridded Conductivity Fields?

We confirm that energy dissipation weighting provides the most accurate approach to determining the effective hydraulic conductivity (Keff) of a binary K grid. A deep learning algorithm (UNET) can infer Keff with extremely high accuracy (R2 > 0.99). The UNET architecture could be trained to infer the energy dissipation weighting pattern from an image of the K distribution, although it was less accurate for cases with highly localized structures that controlled flow. Furthermore, the UNET architecture learned to infer the energy dissipation weighting even if it was not trained directly on this information. However, the weights were represented within the UNET in a way that was not immediately interpretable by a human user. This reiterates the idea that even if ML/DL algorithms are trained to make some hydrologic predictions accurately, they must be designed and trained to provide each user-required output if their results are to be used to improve our understanding of hydrologic systems.

Artificial Neural Network Optimized with a Genetic Algorithm for Seasonal Groundwater Table Depth Prediction in Uttar Pradesh, India

Accurate information about groundwater level prediction is crucial for effective planning and management of groundwater resources. In the present study, the Artificial Neural Network (ANN), optimized with a Genetic Algorithm (GA-ANN), was employed for seasonal groundwater table depth (GWTD) prediction in the area between the Ganga and Hindon rivers located in Uttar Pradesh State, India. A total of 18 models for both seasons (nine for the pre-monsoon and nine for the post-monsoon) have been formulated by using groundwater recharge (GWR), groundwater discharge (GWD), and previous groundwater level data from a 21-year period (1994–2014). The hybrid GA-ANN models’ predictive ability was evaluated against the traditional GA models based on statistical indicators and visual inspection. The results appraisal indicates that the hybrid GA-ANN models outperformed the GA models for predicting the seasonal GWTD in the study region. Overall, the hybrid GA-ANN-8 model with an 8-9-1 structure (i.e., 8: inputs, 9: neurons in the hidden layer, and 1: output) was nominated optimal for predicting the GWTD during pre- and post-monsoon seasons. Additionally, it was noted that the maximum number of input variables in the hybrid GA-ANN approach improved the prediction accuracy. In conclusion, the proposed hybrid GA-ANN model’s findings could be readily transferable or implemented in other parts of the world, specifically those with similar geology and hydrogeology conditions for sustainable planning and groundwater resources management.

Analysis of Groundwater Level Variations Caused by the Changes in Groundwater Withdrawals Using Long Short-Term Memory Network

To properly manage the groundwater resources, it is necessary to analyze the impact of groundwater withdrawal on the groundwater level. In this study, a Long Short-Term Memory (LSTM) network was used to evaluate the groundwater level prediction performance and analyze the impact of the change in the amount of groundwater withdrawal from the pumping wells on the change in the groundwater level in the nearby monitoring wells located in Jeju Island, Korea. The Nash–Sutcliffe efficiency between the observed and simulated groundwater level was over 0.97. Therefore, the groundwater prediction performance of LSTM was remarkably high. If the groundwater level is simulated on the assumption that the future withdrawal amount is reduced by 1/3 of the current groundwater withdrawal, the range of the maximum rise of the groundwater level would be 0.06–0.13 m compared to the current condition. In addition, assuming that no groundwater is taken, the range of the maximum increase in the groundwater level would be 0.11–0.38 m more than the current condition. Therefore, the effect of groundwater withdrawal on the groundwater level in this area was exceedingly small. The method and results can be used to develop new groundwater withdrawal sources for the redistribution of groundwater withdrawals.

Evaluating the Impacts of Pumping on Aquifer Depletion in Arid Regions Using MODFLOW, ANFIS and ANN

In arid regions, the groundwater drawdown consistently increases, and even for a constant pumping rate, long-term predictions remain a challenge. The present research applies the modular three-dimensional finite-difference groundwater flow (MODFLOW) model to a unique aquifer facing challenges of undefined boundary conditions. Artificial neural networks (ANN) and adaptive neuro fuzzy inference systems (ANFIS) have also been investigated for predicting groundwater levels in the aquifer. A framework is developed for evaluating the impact of various scenarios of groundwater pumping on aquifer depletion. A new code in MATLAB was written for predictions of aquifer depletion using ANN/ANFIS. The geotechnical, meteorological, and hydrological data, including discharge and groundwater levels from 1980 to 2018 for wells in Qassim, were collected from the ministry concerned. The Nash–Sutcliffe efficiency and mean square error examined the performance of the models. The study found that the existing pumping rates can result in an alarming drawdown of 105 m in the next 50 years. Appropriate water conservation strategies for maintaining the existing pumping rate can reduce the impact on aquifer depletion by 33%.

Emulation of a Process-Based Salinity Generator for the Sacramento–San Joaquin Delta of California via Deep Learning

Salinity management is a subject of particular interest in estuarine environments because of the underlying biological significance of salinity and its variations in time and space. The foremost step in such management practices is understanding the spatial and temporal variations of salinity and the principal drivers of these variations. This has traditionally been achieved with the assistance of empirical or process-based models, but these can be computationally expensive for complex environmental systems. Model emulation based on data-driven methods offers a viable alternative to traditional modeling in terms of computational efficiency and improving accuracy by recognizing patterns and processes that are overlooked or underrepresented (or overrepresented) by traditional models. This paper presents a case study of emulating a process-based boundary salinity generator via deep learning for the Sacramento–San Joaquin Delta (Delta), an estuarine environment with significant economic, ecological, and social value on the Pacific coast of northern California, United States. Specifically, the study proposes a range of neural network models: (a) multilayer perceptron, (b) long short-term memory network, and (c) convolutional neural network-based models in estimating the downstream boundary salinity of the Delta on a daily time-step. These neural network models are trained and validated using half of the dataset from water year 1991 to 2002. They are then evaluated for performance in the remaining record period from water year 2003 to 2014 against the process-based boundary salinity generation model across different ranges of salinity in different types of water years. The results indicate that deep learning neural networks provide competitive or superior results compared with the process-based model, particularly when the output of the latter are incorporated as an input to the former. The improvements are generally more noticeable during extreme (i.e., wet, dry, and critical) years rather than in near-normal (i.e., above-normal and below-normal) years and during low and medium ranges of salinity rather than high range salinity. Overall, this study indicates that deep learning approaches have the potential to supplement the current practices in estimating salinity at the downstream boundary and other locations across the Delta, and thus guide real-time operations and long-term planning activities in the Delta.

Comparing Activation Functions in Modeling Shoreline Variation Using Multilayer Perceptron Neural Network

The study has modeled shoreline changes by using a multilayer perceptron (MLP) neural network with the data collected from five beaches in southern Taiwan. The data included aerial survey maps of the Forestry Bureau for years 1982, 2002, and 2006, which served as predictors, while the unmanned aerial vehicle (UAV) surveyed data of 2019 served as the respondent. The MLP was configured using five different activation functions with the aim of evaluating their significance. These functions were Identity, Tahn, Logistic, Exponential, and Sine Functions. The results have shown that the performance of an MLP model may be affected by the choice of an activation function. Logistic and the Tahn activation functions outperformed the other models, with Logistic performing best in three beaches and Tahn having the rest. These findings suggest that the application of machine learning to shoreline changes should be accompanied by an extensive evaluation of the different activation functions.

Precision Irrigation Strategies for Sustainable Water Budgeting of Potato Crop in Prince Edward Island

Climate change induced uneven patterns of rainfall emphasize the use of supplemental irrigation in rainfed agriculture. The Penman–Monteith method was used to calculate supplemental irrigation for water budgeting of a potato crop in Prince Edward Island, Canada. Cumulative gaps between rainfall and crop evapotranspiration (ETc) during August and September of the study years were due to high crop coefficient factor, justifying the need for supplemental irrigation. Pressurized irrigation systems, including sprinklers, fertigation, and drip irrigation were installed, to evaluate the impact of scheduled supplemental irrigation in offsetting deficits in irrigation water requirements in comparison with conventional practice of rainfed cultivation (control). A two-way ANOVA examined the effect of irrigation methods and year on potato tuber yield, water productivity, tuber quality, and payout. Sprinkler and fertigation systems performed better than drip and control treatments. In terms of payout returns and potato tuber quality (percentage of marketable potatoes), the sprinkler treatment performed significantly better than the other treatments. However, for water productivity, fertigation treatment performed significantly better than control and sprinkler treatments during both years. The use of supplemental irrigation is recommended for profitable cultivation of potatoes in soil, agricultural, and environmental conditions resembling to those of Prince Edward Island.

Computation of Evapotranspiration with Artificial Intelligence for Precision Water Resource Management

Accurate estimation of reference evapotranspiration (ETo) provides useful information for water resource management and sustainable agriculture. This study estimates ETo with recurrent neural networks (RNNs), namely long short-term memory (LSTM) and bidirectional LSTM. Four representative meteorological sites (North Cape, Summerside, Harrington, and Saint Peters) were selected across Prince Edward Island (PEI), Canada to form a PEI dataset from mean values of the four sites’ climatic variables for capturing climatic variability from all parts of the province. Based on subset regression analysis, the highest contributing climatic variables, namely maximum air temperature and relative humidity, were selected as input variables for RNNs’ training (2011–2015) and testing (2016–2017) runs. The results suggested that the LSTM and bidirectional LSTM are suitable methods to accurately (R2 > 0.90) estimate ETo for all sites except Harrington. Testing period (2016–2017) root mean square errors were recorded in range of 0.38–0.58 mm/day for all sites. No major differences were observed in accuracy of LSTM and bidirectional LSTM. Another objective of this study was to highlight the potential gap between ETO and rainfall for assessing agriculture sustainability in Prince Edward Island. Analyses of the data highlighted that the cumulative ETo surpassed the cumulative rainfall potentially affecting yield of major crops in the island. Therefore, agriculture sustainability requires viable options such as supplemental irrigation to replenish the crop water requirements as and when needed.