Modelling and Simulation of Univariate Model of ‘Groundwater’ and Multivariate Model of ‘Rainfall, Temperature, Root and Topsoil, Depth to Groundwater Table’ Using Deep Learning and Machine Learning Analysis for Time Series Forecasting of Neural Network Model in Rajshahi Region of Bangladesh

Abstract

This study investigates modeling and simulation of groundwater dynamics using univariate and multivariate time series forecasting. While the univariate analysis focuses on the water table, the multivariate analysis integrates relevant variables such as precipitation, temperature, root and surface soil cover, and depth to the water table. The incorporation of advanced computational systems such as deep learning and machine learning significantly improves the analytical accuracy and model robustness compared to traditional numerical approaches. The main outcomes of this study include an extension of the projection model that can estimate groundwater levels based on existing historical data of relevant variable quantities. The developed model can help policymakers and stakeholders make informed decisions regarding groundwater utilization and conservation.

Introduction

Overview:

This study explores the application of machine learning (ML) and deep learning (DL) models for groundwater level (GWL) forecasting, particularly in the Rajshahi region of Bangladesh. It integrates environmental variables (rainfall, temperature, soil moisture) to enhance predictive accuracy and support water resource management.

Key Objectives:

1. Develop univariate and multivariate models to predict GWL.

2. Use ML/DL algorithms to capture temporal and environmental influences on groundwater.

3. Evaluate models using statistical performance metrics and real-world data.

Data and Methodology:

Data Sources:

• Groundwater data: Bangladesh Water Development Board (BWDB)

• Rainfall, temperature, soil moisture: NASA

• Temporal resolution: Daily to seasonal intervals

Preprocessing:

• Handling missing data, scaling, ensuring stationarity

• Feature engineering with lagged variables and correlation analysis

Algorithms Used:

• ML: SVR, Random Forest (RF), KNN

• DL: LSTM, GRU, RNN, LSTM+GRU hybrid

Frameworks:

• Python (TensorFlow, Keras, PyTorch, Scikit-learn)

• Visualization: Matplotlib, Seaborn, Plotly

• Explainability: SHAP

Performance Evaluation:

Metrics:

• MAE, RMSE, MAPE, R², Cross-Correlation Function (CCF)

Data Splitting:

• 65% training, 35% testing

• Rolling window strategies (adjustable and fixed training sizes)

Model Performance Summary:

Key Findings:

• GWL shows strong seasonal patterns with no long-term rise/fall trend.

• Multivariate models (including rainfall, temperature, soil moisture) improve forecasting accuracy.

• DL models (LSTM, GRU) are better at handling complex temporal dependencies than classical ML models.

• SVR and KNN are less effective for highly variable or non-linear time series data.

Applications:

1. Water Resource Management: Planning irrigation and mitigating shortages

2. Climate Impact Studies: Understanding climate variables' impact on groundwater

3. Agriculture: Supporting crop selection and irrigation timing

Challenges:

• Limited high-resolution and long-term data

• DL model complexity and computational demands

• Stationarity and dynamic relationship issues in multivariate models

Conclusion

Predicting the future of artificial intelligence (AI) and neural networks (NN) involves studying current trends, technological developments, and possible applications to make informed predictions about future developments. Computer models of NNs are useful for testing evolving models of thought, such as deliberation, intelligence, and decision-making. Predicting accurate outcomes can be challenging. Understanding current trends and evolving technologies can provide valuable insights into potential directions for AI development. In this study, we combined the power of myriad machine learning algorithms and SVR, RF, KNN, LSTM, GRU, and LSTM+GRU for modeling and simulation by performing deep learning univariate and multivariate time series forecasting with neural network models. Major division areas of Rajshahi were applied. Considering the underlying mechanisms and values that govern the behavior of these models can be challenging, especially for deep learning architectures with millions of constraints. Train each model using the training data and tune the hyper-constraints using the validation set. Improve model performance using methods such as cross-validation and grid search. Abstract appropriate structures from the raw data that can capture the changing aspects and interactions between variables. These may include lag variables, cyclical indicators, and weather indices. I believe this article is an insightful and comprehensive resource for researchers and experts in the field.

References

[1] Abdollahi, A., Bazrafshan, J., & Razmjooy, N. (2020). A deep learning method for temperature forecasting based on Bi-LSTM. Applied Soft Computing, 97, 106782. [2] Acharyya, A., (2017), Sustainable Development of Groundwater Resources in India: A Relook at Policy Initiatives, Vol-VIII. [3] Adhikari, U., & Ikeda, M. (2020). XGBoost and LSTM model for multivariate time series forecasting of groundwater level. Water, 12(11), 3082. DOI. [4] Aishwarya, R., & Vasudevan, V. (2023). A comprehensive review of multi-source data integration for groundwater modeling. Journal of Hydrology, 617, 128812. DOI. [5] Chengg, Li., Li, Gs., & Castelletti, A. (2020). The Ground Water Level (GWL) forecasting with LSTM networks: A study in agricultural water management. Hydrological Sciences Journal, 65(9), 1505-1516. [6] Fawaz, H. I., Forestier, G., Weber, J., et al. (2019). Deep learning for time series classification: Knowledge Discovery, 33(4), 917-963. [7] Gharbi, S., & Bouaziz, M. (2023). Real-time data assimilation for GWL prediction using machine learning. Water Resources Management, 37(5), 1461-1478. DOI. [8] Karthikeyan, L., Khosa, R., & Singh, V. P. (2020). Deep learning models for soil moisture retrieval from remote sensing data: A review. Environmental Earth Sciences, 79(7), 193. [9] Lundberg, S. M., & Lee, S. I. (2017). A unified approach to interpreting model predictions. In Advances in Neural Information Processing Systems (pp. 4765-4774). [10] Lim, B., & Zohren, S. (2021). The Time series predicting with deep learning (DL): A survey. Philosophical Transactions Royal Society: Mathematical, Physical and Engineering Sciences, 379(2194), 20200209. [11] Mojid, M.A., Parvez, M.F., Mainuddin, M. and Hodgson, G., (2019), Water Table Trend- A Sustainability of GWL Development in North-West Bangladesh, Water, Vol 11, pp-1182. [12] Rojas, C., & Krol, M. (2022). Comparative performance of ML methods for groundwater level forecasting: A systematic review. Hydrological Processes, 36(4), e14549. DOI. [13] Shi, X., Chen, Z., Wang, H., et al. (2015). Convolutional LSTM network: ML approach for precipitation nowcasting. Advances Neural Information Processing Systems, 28, 802-810. [14] Simoni, S., & Manoli, G. (2021). Monte Carlo-based sensitivity analysis in groundwater simulation models. Software, 136, 104979. DOI. [15] Tang, Ji., Zhong, Xe., & Li, Qs. (2021). The Groundwater level prediction using a Transformer-based model. Journal of Hydrology, 594, 125707. DOI. [16] Wang, Z., Song, Y., Wang, H., et al. (2019). Multivariate time series forecasting with deep learning and attention mechanism. Neurocomputing, 361, 246-256. [17] Wani, M. H., & Kumar, S. (2021). Feature sensitivity in multivariate groundwater level forecasting models. Environmental Earth Sciences, 80(10), 1256. DOI. [18] Zarei, A. R., & Sepaskhah, A. R. (2022). Prediction of GWL in an arid region using ML and ensemble methods. Hydrological Sciences Journal, 67(7), 1150-1165. DOI. [19] Zhang, Z., Cui, P., & Zhu, W. (2021). Deep learning on graphs: A survey of the IEEE Transactions on the Knowledge and Data Engineering p.e, 34(1), 249-270. [20] Zhang, D., & Choi, H. (2019). Sensitivity analysis of LSTM models for groundwater level forecasting. Journal of Hydrology, 568, 963-976. DOI.

Copyright

Copyright © 2025 Ashraf Shahriar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.