Effective groundwater management is vital for ensuring the sustainability of water resources, particularly in regions with significant agricultural and residential demands. This study focuses on the coastal plain of Lattakia Governorate, a 140-km2 area southeast of Lattakia city, bounded by the Mediterranean Sea, the Al-Kabir Al-Shamali River, and the Al-Sanobar River. The objective was to evaluate the performance of the MODFLOW model, within the Groundwater Modeling System (GMS), for simulating groundwater levels and to assess the predictive capabilities of Long Short-Term Memory (LSTM) networks for forecasting temporal changes in groundwater levels. The methodology involved data collection from thirty observation wells, preprocessing for model calibration, and the integration of MODFLOW and LSTM outputs to create a hybrid framework. Findings indicate that MODFLOW effectively simulates physical groundwater processes, while LSTM captures nonlinear and temporal dynamics. The integration of these models reduced prediction errors by up to 20% compared to standalone approaches, enhancing prediction accuracy and reliability. This study provides a novel approach to groundwater management, offering actionable insights for sustainable agricultural practices, residential water security, and pollution mitigation. The proposed framework demonstrates scalability and adaptability to similar hydrological settings globally, contributing to the advancement of integrated groundwater management strategies.

  • MODFLOW ensures reliable simulation of physical groundwater processes.

  • LSTM excels in capturing nonlinear dependencies and complex temporal dynamics.

  • The hybrid approach reduces prediction errors by up to 20% compared to standalone models.

  • This study offers a scalable framework for sustainable groundwater management in similar hydrological regions.

The world is experiencing an increasing demand for water across various uses, and with rapid population growth, water sources are becoming more vulnerable to depletion and pollution (BBS 2011). Groundwater and surface water pollution sources have proliferated due to expanded human activities in industries, agriculture, and energy sectors (Han et al. 2016). Issues such as sewage, industrial, and agricultural runoff, along with the uncontrolled disposal of solid waste, have become prevalent (Islam et al. 2019).

The decomposition of waste in random landfills generates leachate rich in dangerous contaminants, including heavy metals (Karim et al. 2017). The urgency to understand and manage these pollutants has necessitated the use of computer modeling techniques to study and monitor groundwater sources (Odochi et al. 2024). This approach helps determine the impact of pollutants on water quality and its long-term usability for drinking and irrigation (Beegum et al. 2020). Modeling of water systems is currently the most effective method for predicting future water-related issues, regardless of complexity (Bedekar et al. 2016). These models are widely used in various fields, including aquatic environments, waste disposal sites, and mining (Harbaugh et al. 2017). They enable us to predict changes in groundwater conditions and the effects of different management actions, providing valuable insights for sustainable water resource planning (al Mamunul Haque et al. 2012). The impact of urban landfills on groundwater quality necessitates further investigation (Edzoa et al. 2024). Moreover, the integration of advanced modeling techniques in water management can significantly enhance our understanding of hydrological dynamics (Koutsoyiannis 2020). Rapid urbanization and increasing waste generation in cities are compounding the challenges faced by water resources (Adb 2014). Local government initiatives aimed at improving public health through better waste management are critical (Government Division et al. 2014). Additionally, the hydrogeological conditions in areas such as Rajshahi City demand thorough assessment to inform groundwater management strategies (al Mamunul Haque et al. 2012; Helal 2013). Ultimately, effective planning and management must consider both present and future demands on water resources, integrating socio-economic factors and environmental sustainability (Koutsoyiannis 2004).

Machine learning (ML) emulators have been extensively applied in groundwater transport modeling to enhance the accuracy and efficiency of various processes by learning from data. In the context of pump and treat remediation, ML techniques such as polynomial regression (He et al. 2008), kriging (Jiang et al. 2018), radial basis function (Mugunthan et al. 2005), support vector machine (CH et al. 2013), and artificial neural networks (ANNs) (Yan & Minsker 2006) have been employed to find optimal well locations and pumping schedules. For groundwater contaminant source identification, methods like polynomial chaos expansion (Zhang et al. 2017), kriging (Zhao et al. 2016), support vector machine and kernel extreme learning machine, Gaussian process regression, and ANNs (Srivastava & Singh 2015) have been utilized to identify contaminant source locations, strengths, and durations. Coastal groundwater management has benefited from the application of Evolutionary Polynomial Regression (Hussain et al. 2015), Radial Basis Function (Christelis & Mantoglou 2019), Support Vector Machine (Lal & Datta 2018), Multivariate Adaptive Regression Spline, Genetic Programming (Sreekanth & Datta 2015), Gaussian Process Regression (Rajabi & Ketabchi 2017; Kopsiaftis et al. 2019), ANNs (Ketabchi & Ataie-Ashtiani 2015), Fuzzy Inference System (Solangi et al. 2024), and Kernel Extreme Learning Machine (Song et al. 2018) to optimize pumping strategies while maintaining minimum saltwater concentrations.

Recently, integrated wavelet, multilayer perceptron (MLP), time-delay neural network (TDNN), and gamma memory neural network (GMNN) were used to predict hourly river-level fluctuations (Agarwal et al. 2022). Similarly, ANNs have been extensively applied to flood forecasting and river flow modeling, with multiple input-output configurations providing enhanced accuracy for complex hydrological systems (Agarwal et al. 2021a). Comparative studies of ANN architectures, such as Multiple Input–Multiple Output (MIMO) and Multiple Input–Single Output (MISO) models, underscore the critical role of selecting appropriate architectures for achieving reliable forecasts (Agarwal et al. 2021b). Mistry & Parekh (2022) demonstrated the effectiveness of ANNs for river flow forecasting in the Deo River, Gujarat. Using rainfall and discharge data, they developed six ANN models with different training algorithms, such as Cascade Forward Backpropagation and Levenberg–Marquardt. The Cascade Forward Backpropagation model performed best, achieving high correlation coefficients (r= 0.89 for validation) and closely matching observed inflows. Li et al. (2024) proposed a hybrid approach combining outputs from physical-based hydrological models (NAM and HD) with historical data to train long short-term memory (LSTM) networks for flood forecasting in the Jinhua basin, China. By evaluating LSTM models trained on measured, simulated, and mixed datasets, they demonstrated that a simulated-to-measured data ratio of less than 2:1 significantly improves predictive accuracy, as reflected by lower RMSE and MAE values. The study highlights the effectiveness of integrating measured and simulated data to overcome data scarcity and enhance model performance, which has potential applications in real-time flood prediction and risk management. These studies highlight a common theme: the integration of data-driven models with domain-specific knowledge enhances the ability to predict hydrological phenomena accurately.

The coastal plain of Lattakia Governorate, spanning 140 km2 southeast of Lattakia City, is a vital region for agriculture and residential water needs. Bounded by the Mediterranean Sea, the Al-Kabir Al-Shamali and Al-Sanobar rivers, and several villages, this area faces significant challenges, including groundwater depletion and pollution. Despite the importance of sustainable groundwater management, existing approaches often rely on either traditional numerical models or data-driven methods, without fully exploiting the potential of integrating these techniques. This study addresses this gap by evaluating the performance of the MODFLOW model, a widely used numerical tool, for simulating groundwater levels and applying LSTM networks, a ML approach, to predict temporal changes in groundwater levels. By integrating these methods, we propose a hybrid framework that combines the spatial accuracy of numerical modeling with the predictive power of ML. The novelty of this work lies in bridging traditional and advanced methodologies to enhance groundwater management and pollution mitigation, providing a scalable solution for similar hydrological contexts.

Study area

In this research, a monitoring network of 30 observation wells, distributed throughout the study area and drilled by local farmers, was utilized to collect hydrogeological data. Figure 1 shows the general location of the study area. The spatial distribution of these wells within the study area is depicted in Figure 2.
Figure 1

The general location of the study area.

Figure 1

The general location of the study area.

Close modal
Figure 2

Distribution of observation wells within the study area.

Figure 2

Distribution of observation wells within the study area.

Close modal

Field measurements included determining the coordinates and elevations of the wells. The well coordinates along the X and Y axes and the elevation of the wellheads above sea level (Z) were measured with an accuracy of ±0.5 m using a GARMIN GPS device. Additionally, groundwater depths in the wells were measured using an electro-optical depth measurement device equipped with a 200-m cable graduated in centimeters. This instrument provided measurements with a precision of ±0.5 cm, ensuring accurate monitoring of groundwater levels across the observation network.

Research methodology

A flowchart illustrating the research methodology is shown in Figure 3. This flowchart provides a clear visual representation of the sequential steps adopted in our study, from data collection to model integration and evaluation.
Figure 3

Research methodology flowchart.

Figure 3

Research methodology flowchart.

Close modal

Groundwater Modeling System

The Groundwater Modeling System (GMS) is an environmental software package for groundwater modeling. GMS is utilized at a rapidly growing number of private and international sites due to its comprehensive capabilities. It provides a complete suite of tools required for every stage of groundwater simulation, including site characterization, model development, result analysis, calibration, and visualization.

GMS is unique in its support for both 2D and 3D geostatistics, allowing for detailed geological formations to be modeled accurately. The software includes various models such as MODFLOW, MODPATH, MT3D, RT3D, FEMWATER, SEEP2D, SEAM3D, PEST, UCODE, and UTCHEM. These models help in simulating groundwater flow and contaminant transport, offering precise and actionable insights for groundwater management and pollution mitigation efforts. For this study, we used MODFLOW for quantitative modeling.

By leveraging GMS, we can achieve detailed simulations that integrate hydrogeological data, predict changes in groundwater levels, and evaluate the impact of different management strategies on water quality and availability. This comprehensive approach enables effective decision-making for sustainable groundwater resource management.

LSTM implementation

In this study, LSTM networks were employed to predict groundwater level changes in the coastal plain of Lattakia Governorate. LSTM networks, a specialized type of recurrent neural network (RNN), are particularly suited for capturing temporal dependencies in sequential data, making them ideal for modeling groundwater fluctuations over time. The dataset used for this research comprised monthly groundwater level measurements from multiple wells collected over several years. Preprocessing steps included normalizing the groundwater level data to enhance the learning process and handling missing values using linear interpolation. The data was subsequently divided into training and testing sets, allocating 80% for training and 20% for testing to ensure robust model evaluation. The LSTM model was implemented using the TensorFlow and Keras libraries in Python. The architecture consisted of an input layer to process sequences of groundwater level data, followed by two LSTM layers, each with 50 units, to capture complex temporal patterns. A dropout layer with a rate of 0.2 was added to mitigate overfitting, and a dense layer with a single neuron served as the final output, representing the predicted groundwater level. Hyperparameters included a learning rate of 0.001, optimized using the Adam optimizer, a batch size of 32, and a total of 100 epochs for training. The model training was conducted on Google Colab, leveraging its GPU acceleration to expedite the process. The mean squared error (MSE) loss function was used during training to quantify the differences between predicted and actual groundwater levels. The trained model's performance was evaluated on the test set using metrics such as root mean squared error (RMSE) to assess prediction accuracy and the goodness-of-fit.

Hybrid model integration

The integration of MODFLOW and LSTM outputs creates a synergistic framework for accurate groundwater level predictions by leveraging the strengths of both methods. In this approach, MODFLOW-simulated groundwater levels are used as input features for the LSTM model (), where the relationship can be expressed as:

Here, represents the LSTM-predicted groundwater levels at time t, denotes the MODFLOW-simulated groundwater levels for the past k timesteps, includes additional influencing variables (e.g., precipitation, recharge rates), and θ represents the trainable parameters of the LSTM model. A feedback loop is established by iteratively refining using , ensuring the predictions are informed by both spatial and temporal dynamics.

A weighting mechanism is developed to balance the contributions of measured, simulated, and LSTM-predicted groundwater levels. The final integrated prediction is given by:
where is the measured groundwater level, and , , and are the weights assigned to the measured, simulated, and LSTM-predicted data, respectively, with . These weights are optimized based on the reliability and consistency of each data source.

Performance evaluation

The performance of the hybrid model is evaluated using statistical metrics that quantify prediction accuracy. The RMSE is calculated as:
where and are the observed and predicted groundwater levels at the i-th time step, respectively, and n is the total number of predictions.
The Mean Absolute Error (MAE) is computed as:
Natural groundwater systems are generally classified based on temporal changes in groundwater levels into several types: daily, seasonal, annual, long-term, or according to factors causing system changes such as climatic systems, watershed divides, hydrological systems, riverbanks, karst areas, and effective layer systems. In our study, periodic measurements of groundwater depths in all wells were conducted monthly over average of 2020–2023 years, from April 2020 to March 2023. The results are presented in Figure 4.
Figure 4

Spatial distribution map of average groundwater table depths in the study area during the study period.

Figure 4

Spatial distribution map of average groundwater table depths in the study area during the study period.

Close modal

The results indicate that the monthly average groundwater depths in the observation wells ranged from 1 to 16 m. The maximum groundwater depth recorded was 17.6 m in well number 12 during August and September, while the minimum depth recorded was 0.35 m in well number 8 during January. This indicates that groundwater levels are relatively shallow and close to the surface throughout most of the study area. We observed that groundwater depths tend to decrease slightly during the summer months compared to the winter months, which can be attributed to the recharge effect from rainfall. The shallow water table shows a clear response to infiltrating rainwater. Additionally, we noticed that the groundwater depths in individual wells remained relatively consistent throughout the year. This consistency can be attributed to the aquifer being recharged by rainfall infiltration and surface runoff during winter, and by irrigation water from the 16 Tishreen Dam irrigation network, which serves most of the irrigated agricultural lands in the study area during summer.

We graphically represented the groundwater depth variations in the observation wells along with the monthly rainfall variations for the same period in Figure 5 for comparison.
Figure 5

Changes in groundwater table depths (m) in monitoring wells in relation to monthly rainfall variations during the study period.

Figure 5

Changes in groundwater table depths (m) in monitoring wells in relation to monthly rainfall variations during the study period.

Close modal

The comparison indicates a direct correlation between groundwater depth changes and rainfall, where the groundwater system in the study area shows a rapid response in most of the studied wells to changes in rainfall amounts. Groundwater depths decrease during winter and increase during summer.

We derived the absolute groundwater level values by subtracting the groundwater depth values in each well from the wellhead elevation above sea level (Z). The groundwater levels in the monitored observation wells during the study period are shown in Figure 6. The monthly average groundwater levels in the observation wells ranged within 18–165 m. The highest groundwater level recorded was 168.99 m in well number 9 in January, while the lowest was 17.23 m in well number 1 in October.
Figure 6

Spatial distribution map of the average absolute groundwater level in the study area during the study period.

Figure 6

Spatial distribution map of the average absolute groundwater level in the study area during the study period.

Close modal

We also observed that groundwater levels increased during the winter months compared to the summer months. This leads us to conclude that rainfall recharge plays a primary role in the groundwater system. The short time it takes for water to reach the well, due to the shallow depth and the good permeability characteristics of the geological formations in the area, allows for vertical pollutant migration. Additionally, horizontal flows can form, contributing to the transport of various pollutants to the groundwater reservoir in the region.

Quantitative modeling using the MODFLOW model

For the quantitative modeling of the groundwater system in the study area, we used the MODFLOW model from the GMS 7.0. This model is one of the most widely used three-dimensional models globally, known for its high accuracy in quantitatively assessing groundwater flow and predicting long-term changes. It also excels in representing the geological and hydrogeological characteristics of aquifers with high efficiency. In our work, we adopted the conceptual model approach for several reasons: it efficiently represents the groundwater system in the study area in a simplified and rapid manner, and it allows for accurate and quick structural changes to the model during simulation.

The scatter plot in Figure 7 compares the computed groundwater surface levels with the observed measurements after calibration using the MODFLOW Model from GMS. Each point represents a pair of computed and observed values for specific locations within the study area, denoted by different symbols for various measurement sites. The plot shows a strong linear relationship between the computed and observed values, indicated by the points closely aligning along the 45° line (where the computed values equal the observed values). This alignment suggests that the MODFLOW Model has accurately captured the groundwater surface levels across the different measurement sites. The symbols representing different sites show that the calibration is consistently accurate across the entire study area, with minimal deviations. The calibration process appears to be successful, as the differences between the computed and observed values are minor, demonstrating the model's reliability in simulating groundwater dynamics. The few points that slightly deviate from the line indicate minor discrepancies, which could be due to local variations in hydrogeological conditions that are not fully captured by the model or measurement inaccuracies. Overall, the results confirm that the MODFLOW Model from GMS provides a robust and accurate tool for predicting groundwater surface levels, essential for effective groundwater management and planning in the study area.
Figure 7

Differences between the calculated and measured values of groundwater surface levels after calibration using MODFLOW Model from GMS.

Figure 7

Differences between the calculated and measured values of groundwater surface levels after calibration using MODFLOW Model from GMS.

Close modal

Quantitative prediction using MODFLOW Model and LSTM

In assessing the reliability of MODFLOW and LSTM models for predicting groundwater levels, we analyzed the normalized changes in groundwater levels from the first eight wells, as presented in Figure 8. The MODFLOW predictions, represented by red dashed lines, and the LSTM predictions, indicated by green dashed lines, were compared against actual normalized groundwater levels. Both models demonstrated a high degree of conformity with observed data, indicating their robustness in capturing temporal variations in groundwater levels.
Figure 8

Comparison of normalized groundwater level changes for the first eight wells, showcasing actual measurements (blue solid lines), MODFLOW model predictions (red dashed lines), and LSTM model predictions (green dashed lines).

Figure 8

Comparison of normalized groundwater level changes for the first eight wells, showcasing actual measurements (blue solid lines), MODFLOW model predictions (red dashed lines), and LSTM model predictions (green dashed lines).

Close modal

For instance, in Well 1, the actual normalized levels ranged from −0.59 to 1.34, while MODFLOW predictions ranged from −0.64 to 1.27 and LSTM predictions ranged from −0.71 to 1.30. Similarly, in Well 3, the actual normalized levels varied from −1.32 to 1.15, with MODFLOW and LSTM predictions ranging from −1.22 to 1.10 and −1.35 to 1.18, respectively. Both models closely tracked the seasonal trends observed in the actual data.

The results in Table 1 demonstrate the performance of the MODFLOW model, LSTM model, and their integration across the first eight wells. The integration approach consistently achieves the lowest RMSE and relative mean error (RME) values, indicating improved accuracy and reliability compared to the standalone models. For instance, the RMSE values for the integration approach range from 0.13 to 0.17, significantly lower than those of the MODFLOW (0.24–0.31) and LSTM models (0.18–0.23). Similarly, the RME for the integration approach ranges between 3.0 and 3.8, reflecting a notable improvement over the MODFLOW (5.1–6.4) and LSTM (4.4–4.9) models. These results emphasize the effectiveness of combining numerical and ML models in enhancing groundwater level predictions.

Table 1

Comparison of RMSE and RME values for MODFLOW, LSTM, and integration models across the first eight wells

WellRMSE
RME
MODFLOWLSTMIntegrationMODFLOWLSTMIntegration
Well 1 0.25 0.2 0.15 5.2 4.8 3.5 
Well 2 0.3 0.18 0.14 6.3 4.5 3.2 
Well 3 0.28 0.22 0.16 5.8 4.9 3.7 
Well 4 0.27 0.19 0.13 4.7 3.1 
Well 5 0.26 0.21 0.14 5.5 4.6 3.3 
Well 6 0.29 0.2 0.15 5.9 4.5 3.4 
Well 7 0.31 0.23 0.17 6.4 4.8 3.8 
Well 8 0.24 0.18 0.13 5.1 4.4 
WellRMSE
RME
MODFLOWLSTMIntegrationMODFLOWLSTMIntegration
Well 1 0.25 0.2 0.15 5.2 4.8 3.5 
Well 2 0.3 0.18 0.14 6.3 4.5 3.2 
Well 3 0.28 0.22 0.16 5.8 4.9 3.7 
Well 4 0.27 0.19 0.13 4.7 3.1 
Well 5 0.26 0.21 0.14 5.5 4.6 3.3 
Well 6 0.29 0.2 0.15 5.9 4.5 3.4 
Well 7 0.31 0.23 0.17 6.4 4.8 3.8 
Well 8 0.24 0.18 0.13 5.1 4.4 

This study evaluated the performance of the MODFLOW model and LSTM networks for predicting groundwater levels in the coastal plain of Lattakia Governorate. MODFLOW proved effective in simulating physical groundwater processes, while LSTM captured nonlinear dependencies and temporal dynamics. Integrating these methods enhanced predictive accuracy, supporting sustainable water resource management in this agriculturally and residentially significant region. This dual-model approach demonstrates the potential for robust groundwater management strategies and offers a scalable framework for similar hydrological settings worldwide.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.

Agarwal
S.
,
Roy
P. J.
,
Choudhury
P.
&
Debbarma
N.
(
2021a
)
Flood forecasting and flood flow modeling in a river system using ANN
,
Water Practice & Technology
,
16
(
4
),
1194
1205
.
Agarwal
S.
,
Roy
P. J.
,
Choudhury
P. S.
&
Debbarma
N.
(
2021b
)
River flow forecasting by comparative analysis of multiple input and multiple output models form using ANN
,
H2Open Journal
,
4
(
1
),
413
428
.
Agarwal
S.
,
Roy
P.
,
Choudhury
P.
&
Debbarma
N.
(
2022
)
Comparative study on stream flow prediction using the GMNN and wavelet-based GMNN
,
Journal of Water and Climate Change
,
13
(
9
),
3323
3337
.
al Mamunul Haque
M.
,
Jahan
C. S.
,
Mazumder
Q. H.
,
Nawaz
S. M. S.
,
Mirdha
G. C.
,
Mamud
P.
&
Adham
M. I.
(
2012
)
Hydrogeological condition and assessment of groundwater resource using visual MODFLOW modeling: Rajshahi City aquifer, Bangladesh
,
Journal of the Geological Society of India
,
79
(
1
),
77
84
.
https://doi.org/10.1007/s12594-012-0001-7
.
Asian Development Bank (Adb) (2014) Urban public and environmental health sector development program: sanitary landfill in Khulna City Corporation (resettlement plans). Manila, Philippines: Asian Development Bank. Retrieved July 1, 2024 from https://www.adb.org/projects/documents/bangladesh-39305-013-29.
Bangladesh Bureau of Statistics (BBS)
(
2011
)
Statistical Yearbook of Bangladesh
,
31st edition
.
Dhaka, Bangladesh
.
Bedekar
V.
,
Morway
E. D.
,
Langevin
C. D.
&
Tonkin
M.
(
2016
)
MT3D-USGS version 1: A U.S. Geological Survey release of MT3DMS updated with new and expanded transport capabilities for use with MODFLOW: U.S. Geological Survey Techniques and Methods 6-A53, 69 p. http://dx.doi.org/10.3133/tm6A53
.
Beegum
S.
,
Vanderborght
J.
,
Šimůnek
J.
,
Herbst
M.
,
Sudheer
K. P.
&
Nambi
I. M.
(
2020
)
Investigating atrazine concentrations in the Zwischenscholle aquifer using MODFLOW with the HYDRUS-1D package and MT3DMS
,
Water (Switzerland)
,
12
(
4
), 1019.
https://doi.org/10.3390/W12041019
.
Ch
S.
,
Kumar
D.
,
Prasad
R. K.
&
Mathur
S.
(
2013
)
Optimal design of an in-situ bioremediation system using support vector machine and particle swarm optimization
,
Journal of Contaminant Hydrology
,
151
,
105
116
.
https://doi.org/10.1016/j.jconhyd.2013.05.003
.
Christelis
V.
&
Mantoglou
A.
(
2019
)
Pumping optimization of coastal aquifers using seawater intrusion models of variable-fidelity and evolutionary algorithms
,
Water Resources Management
,
33
(
3
),
555
568
.
https://doi.org/10.1007/s11269-018-2116-0
.
Edzoa
R. C.
,
Mbog
M. B.
,
Tedontsah
V. P. L.
,
Mamdem
J.
,
Ngon
G. F. N.
,
Bitom
D.
,
Etame
J.
&
Nguedia
K. D.
(
2024
)
Influence of leachates produced by urban waste dumps on the water quality and possible risks to public health
,
Water Practice and Technology
,
19
(
1
),
82
98
.
https://doi.org/10.2166/wpt.2023.227
.
Government Division, L. of Local Government, M., & Development, R.
(
2014
)
39305-013: Urban Public and Environmental Health Sector Development Program (Program Loan)
.
Han
Z.
,
Ma
H.
,
Shi
G.
,
He
L.
,
Wei
L.
&
Shi
Q.
(
2016
)
A review of groundwater contamination near municipal solid waste landfill sites in China
,
Science of the Total Environment
,
569–570
,
1255
1264
.
https://doi.org/10.1016/j.scitotenv.2016.06.201
.
Harbaugh
A. W.
,
Langevin
C. D.
,
Hughes
J. D.
,
Niswonger
R. N.
&
Konikow
L. F.
(
2017
)
MODFLOW-2005 version 1.12.00: the U.S. Geological Survey modular groundwater model: U.S. Geological Survey Software Release, 03 February 2017. http://dx.doi.org/10.50
.
He
L.
,
Huang
G.
,
Zeng
G.
&
Lu
H.
(
2008
)
An integrated simulation, inference, and optimization method for identifying groundwater remediation strategies at petroleum-contaminated aquifers
,
Water Research
,
42
(
10
),
2629
2639
.
https://doi.org/10.1016/j.watres.2008.01.012
.
Hussain
M. S.
,
Javadi
A. A.
,
Ahangar-Asr
A.
&
Farmani
R.
(
2015
)
A surrogate model for simulation–optimization of aquifer systems subjected to seawater intrusion
,
Journal of Hydrology
,
523
,
542
554
.
https://doi.org/10.1016/j.jhydrol.2015.01.079
.
Islam
M. S.
,
Moniruzzaman
S. M.
&
Mondal
S.
(
2019
)
Municipal solid waste quantification, transportation, and management in Khulna city
,
Journal of Engineering Science
,
10
(
2
), 61–68.
Islam, M. I., Habib, S. A., Haque, S. A. U., Sultana, N., Faisal, B. R., Rahman, H. & Sharifee, M. N. H. (2020) Applicability of OCO-2 solar induced chlorophyll fluorescence (SIF) data for the estimation of photosynthetic activity in Bangladesh. Journal of Engineering Science, 11 (2), 133–140.
Jiang
X.
,
Lu
W.
,
Na
J.
,
Hou
Z.
,
Wang
Y.
&
Chi
B.
(
2018
)
A stochastic optimization model based on adaptive feedback correction process and surrogate model uncertainty for DNAPL-contaminated groundwater remediation design
,
Stochastic Environmental Research and Risk Assessment
,
32
(
10
),
3195
3206
.
https://doi.org/10.1007/s00477-018-1559-4
.
Karim
M. R.
,
Kuraoka
M.
,
Higuchi
T.
,
Sekine
M.
&
Imai
T.
(
2017
)
Assessment of heavy metal contamination from municipal solid waste open dumping sites in Bangladesh
,
International Journal of Environment and Waste Management
,
19
(
3
),
191
202
.
https://doi.org/10.1504/IJEWM.2017.084297
.
Ketabchi
H.
&
Ataie-Ashtiani
B.
(
2015
)
Review: coastal groundwater optimization – advances, challenges, and practical solutions
,
Hydrogeology Journal
,
23
(
6
),
1129
1154
.
https://doi.org/10.1007/s10040-015-1254-1
.
Kopsiaftis
G.
,
Protopapadakis
E.
,
Voulodimos
A.
,
Doulamis
N.
&
Mantoglou
A.
(
2019
)
Gaussian process regression tuned by Bayesian optimization for seawater intrusion prediction
,
Computational Intelligence and Neuroscience
,
2019
,
1
12
.
https://doi.org/10.1155/2019/2859429
.
Koutsoyiannis
D.
(
2005
)
Hydrologic persistence and the hurst phenomenon
. In: Lehr, J. H. & Keeley, J. (eds)
Water Encyclopedia
.
Hoboken, NJ, USA: Wiley
, pp.
210
221
.
https://doi.org/10.1002/047147844x.sw434
.
Koutsoyiannis
D.
(
2020
)
Revisiting the global hydrological cycle: is it intensifying?
,
Hydrology and Earth System Sciences
,
24
(
8
),
3899
3932
.
https://doi.org/10.5194/hess-24-3899-2020
.
Li
J.
,
Wu
G.
,
Zhang
Y.
&
Shi
W.
(
2024
)
Optimizing flood predictions by integrating LSTM and physical-based models with mixed historical and simulated data
,
Heliyon
,
10
(
13
), e33669.
Mistry
S.
&
Parekh
F.
(
2022
). '
Flood Forecasting Using Artificial Neural Network
',
IOP Conference Series: Earth and Environmental Science
.
Bristol, UK: IOP Publishing
.
Mugunthan
P.
,
Shoemaker
C. A.
&
Regis
R. G.
(
2005
)
Comparison of function approximation, heuristic, and derivative-based methods for automatic calibration of computationally expensive groundwater bioremediation models
,
Water Resources Research
,
41
(
12
),
1
17
.
https://doi.org/10.1029/2005WR004134
.
Odochi
U. B.
,
Iheanyichukwu
O. A.
,
Okechukwu
O. S.
,
Chukwuemeka
I. S.
,
Juliana
O. I.
,
Ifeanyi
O. S.
,
Nkiru
N. C.
,
Osita
I. P.
&
Emmanuel
A. E.
(
2024
)
Hydrogeological assessment and contaminant transport modelling of Enyimba landfill site in Aba, Nigeria
,
Water Practice and Technology
,
19
(
5
),
2108
2124
.
https://doi.org/10.2166/wpt.2024.122
.
Rajabi
M. M.
&
Ketabchi
H.
(
2017
)
Uncertainty-based simulation-optimization using Gaussian process emulation: application to coastal groundwater management
,
Journal of Hydrology
,
555
,
518
534
.
https://doi.org/10.1016/j.jhydrol.2017.10.041
.
Solangi
G. S.
,
Ali
Z.
,
Bilal
M.
,
Junaid
M.
,
Panhwar
S.
,
Keerio
H. A.
,
Sohu
I. H.
,
Shahani
S. G.
&
N
Z.
(
2024
)
Machine learning, Water Quality Index, and GIS-based analysis of groundwater quality
,
Water Practice and Technology
,
19
(
2
),
384
400
.
https://doi.org/10.2166/wpt.2024.014
.
Song
J.
,
Yang
Y.
,
Wu
J.
,
Wu
J.
,
Sun
X.
&
Lin
J.
(
2018
)
Adaptive surrogate model based multiobjective optimization for coastal aquifer management
,
Journal of Hydrology
,
561
,
98
111
.
https://doi.org/10.1016/j.jhydrol.2018.03.063
.
Sreekanth
J.
&
Datta
B.
(
2015
)
Review: simulation-optimization models for the management and monitoring of coastal aquifers
,
Hydrogeology Journal
,
23
(
6
),
1155
1166
.
https://doi.org/10.1007/s10040-015-1272-z
.
Srivastava
D.
&
Singh
R. M.
(
2015
)
Groundwater system modeling for simultaneous identification of pollution sources and parameters with uncertainty characterization
,
Water Resources Management
,
29
(
12
),
4607
4627
.
https://doi.org/10.1007/s11269-015-1078-8
.
Uddin, S. M. H. (2013) Geo-chemistry of ground water in Rajshahi City and improving its quality for drinking purpose (PhD thesis). RUCL Institutional Repository, Institute of Environmental Science, University of Rajshahi, Rajshahi, Bangladesh. Retrieved from http://rulrepository.ru.ac.bd/handle/123456789/772.
Yan
S.
&
Minsker
B.
(
2006
)
Optimal groundwater remediation design using an adaptive neural network genetic algorithm
,
Water Resources Research
,
42
(
3
),
1
14
.
https://doi.org/10.1029/2005WR004303
.
Zhang
J.
,
Li
W.
,
Lin
G.
,
Zeng
L.
&
Wu
L.
(
2017
)
Efficient evaluation of small failure probability in high-dimensional groundwater contaminant transport modeling via a two-stage Monte Carlo method
,
Water Resources Research
,
53
(
3
),
1948
1962
.
https://doi.org/10.1002/2016WR019518
.
Zhao
Y.
,
Lu
W.
&
Xiao
C.
(
2016
)
A Kriging surrogate model coupled in simulation–optimization approach for identifying release history of groundwater sources
,
Journal of Contaminant Hydrology
,
185–186
,
51
60
.
https://doi.org/10.1016/J.JCONHYD.2016.01.004
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).