Evaluating the effect of defining management scenarios in water table prediction accuracy using Wavelet-Support Vector Regression (WSVR) hybrid model

Document Type : Research Paper

Authors

Department of Water Science and Engineering, Faculty of Agriculture, Bu Ali Sina University, Hamedan, Iran

Abstract

Due to surface water limitations, groundwater reservoirs are known as the main source of supplying water requirements in most arid countries. Recently, owing to these aquifers’ over-exploitation, in order to apply optimal management of these resources, reliable water table forecasting exceeded in importance. This study is aimed to Determine the best scenario for combining input data in Hamedan-Bahar plain water table forecasting using the wavelet-support vector regression (WSVR) hybrid model. In the first step, Rainfall, temperature, evaporation, and the groundwater level data of seventeen piezometers in 26 years (1991-2017) were collected and completed. Nine scenarios with different lags and combinations were considered to select the model inputs and their lag numbers. Modeling performance in each scenario was evaluated using statistical parameters such as the Pearson correlation coefficient (r), standard error (SE), and root mean square error (RMSE), and using the best scenario, the water table of the area was predicted for ten years ahead. Based on the obtained results, the scenario in which each of the four input parameters was used, with one and two lags, had the highest accuracy. Additionally, the predicted results, using the best scenario, illustrated a noticeable downward trend in the water table of the region in the future.  Therefore, concerning the high sensibility of this plain because of supplying the water demands in drinking water, agriculture, and industry of Hamedan and Bahar, as well as the necessity of more water harvesting in the future, deciding more favorable groundwater management is highly imperative in this area.

Keywords

Main Subjects

Evaluating the effect of defining management scenarios in water table prediction accuracy using Wavelet-Support Vector Regression (WSVR) hybrid model

Extended Abstract

 

Introduction

Nowadays, owing to the fact that we are facing surface water limitations, it is undeniable that most arid and semi-arid countries rely on groundwater reservoirs. Upward trends of groundwater harvesting, over-population, and reducing soil permeability can contribute to decreasing water table so that precipitation is not capable of this damage compensation. Therefore, accurate hydrological studies and management are becoming highly imperative. Thus, in order to decrease these damaging effects and improve the aquifers’ situation, taking some actions such as getting to know and investigating the existing conditions of groundwater resources and forecasting the water table in the future is necessary. Simulating current circumstances of considered phenomena can be provided by indirect methods like artificial intelligence (AL) as an acceptable, fast, and accurate way to simulate and predict different phenomena. In addition, in recent years, wavelet transport has been used by researchers to improve AL’s simulation and modeling results. According to previous studies, the wavelet-support regression (WSVR) hybrid model had high accuracy in forecasting the water table. Hence, regarding using a wide range of data with different lag times and lack of optimal structure for modeling inputs in past studies, in this study, the purpose is to define and determine the best scenario for selecting modeling input data with suitable lag time to predict water table for heightening WSVR result accuracy.

Material and Methods

This study is aimed to determine the most effective scenario for selecting input data and their suitable lag time using the WSVR hybrid model to predict the groundwater level of the Hamedan-Bahar plain. This area is located in Hamedan, at the sea level of 1700-1800 meters, and its aquifer has 483 kilometers area. For this study implementation, efficient data on groundwater level based on previous studies, including precipitation (P), evaporation (E), temperature (T), and water table, were collected from seventeen piezometers, six rain-gage stations, and three climatological stations during 1991-2017. After data preparation and data interpolation in the area, nine scenarios were determined with various combinations and lag times (regarding the degree of correlation of parameters with the water table) in order to define WSVR model inputs. The model used is a mixture of wavelet transform and SVR model. The db2 wavelet and two decomposition levels were adjusted in wavelet transform (to denoise data and decompose each time series to two groups of high and low frequency), and the RBF kernel function was used in the SVR model. Then, by applying this model, this area's water table was simulated.

Results and Discussion

According to the results, overall, the WSVR model had high accuracy for predicting the water table of this area, as it had the lowest Pearson correlation coefficient (r) and maximum standard error (SE) of 0.897 and 0.001 in the whole modeling process among all scenarios, respectively. In addition, it was clarified that the ninth scenario was recognized as the best scenario compared to others, with the highest r in the training and testing stage (0.999 and 0.974, in turn) and the lowest SE (0.0005 in training and 0.00007 in testing) and RMSE (in training 0.129, and 0.827 in testing stage). Eventually, after forecasting this area's water table for the next ten years using this scenario, it can be seen that the depth of water access is increasing.

Conclusion

The results outline that it is necessary to apply precipitation, temperature, evaporation, and water table data, all together, with one and two lags, to predict the groundwater level of this area accurately with the WSVR model. It should also be mentioned that to take advantage of this model, the db2 wavelet with two decomposition levels for each time series in wavelet transform, and the RBF kernel function in the SVR model, should be adjusted in coding the WSVR. Furthermore, according to the predicted water level of this area and its significant downward trend, appropriate management measures should be taken to control the future situation for supplying water requirements.

References

Ansari, M., & Jabbari, I. (2023). Modeling and forecasting of the underground water level of Izadkhasht plain, Fars province. Journal of Geography and Environmental Hazards, (), -.
Behzad, M., Asghari, K., & Coppola Jr, E. A. (2010). Comparative study of SVMs and ANNs in aquifer water level prediction. Journal of Computing in Civil Engineering, 24(5), 408-413.
Bierkens, M. F. (1998). Modeling water table fluctuations by means of a stochastic differential equation. Water Resources Research, 34(10), 2485-2499.
Cohen, A., & Kovacevic, J. (1996). Wavelets: The mathematical background. Proceedings of the IEEE, 84(4), 514-522.
Ebrahimi, H., & Rajaee, T. (2017). Simulation of groundwater level variations using wavelet combined with neural network, linear regression and support vector machine. Global and Planetary Change, 148, 181-191.
Eskandari, A., Faramarzyan yasuj, F., Zarei, H., & Solgi, A. (2019). Evaluation of combined ANFIS with wavelet transform to modeling and forecasting groundwater level, Journal of Watershed Management Research, 9(18), 56-69. (In Persian)
Eskandari, A., Solgi, A., & Zarei, H. (2018). Simulating fluctuations of groundwater level using a combination of support vector machine and wavelet transform. Journal of Irrigation Sciences and Engineering (JISE), 41(1), 165-180. (In Persian)
Ghobadian, R., Bahrami, Z., & Dabagh Bagheri, S. (2016). Applying the management scenarios in prediction of groundwater level fluctuations by using the conceptual and mathematical MODFLOW model (Case study: Khezel-Nahavand Plain). Iranian journal of Ecohydrology, 3(3), 303-319. (In Persian)
Gong, Y., Zhang, Y., Lan, S., & Wang, H. (2016). A comparative study of artificial neural networks, support vector machines and adaptive neuro fuzzy inference system for forecasting groundwater levels near Lake Okeechobee, Florida. Water resources management, 30, 375-391.
Izadi, A.A., Davari, K., Alizadeh, A., Ghahraman, B., & Haghayeghi Moghadam, S.A. (2007). Water table forecasting using artificial neural networks. Iranian Journal of Irrigation and Drainage, 1(2), 59-71. (In Persian)
Khayat, A. (2017). Prediction of grounwater level using the Nuero- Fuzzy method and wavelet analysis under the effects of climate change. Master of Science’s thesis, Faculty of Agriculture, University of Birjand, Iran. (In Persian)
 
Khedri, A., Kalantari, N., & Vadiati, M. (2020). Comparison study of artificial intelligence method for short term groundwater level prediction in the northeast Gachsaran unconfined aquifer. Water Supply, 20(3), 909-921.
Moosavi, V., Vafakhah, M., Shirmohammadi, B., & Behnia, N. (2013). A wavelet-ANFIS hybrid model for groundwater level forecasting for different prediction periods. Water resources management, 27, 1301-1321
Nayak, P. C., Rao, Y. S., & Sudheer, K. P. (2006). Groundwater level forecasting in a shallow aquifer using artificial neural network approach. Water resources management, 20, 77-90.
 
Nourani, V., Hasan Zadeh, Y., Komasi, M., & Sharghi, E. (2008). Precipitation-runoff modeling with wavelet-artificial neural network hybrid model. Fourth National Congress of Civil Engineering. University of Tehran. (In Persian)
Rahnama, N., Zarei, H., & Ahmadi, F. (2023). Application of K-star Algorithm for Groundwater Level Forecasting (Case study: Aspas plain). Iranian Journal of Irrigation & Drainage, 17(2), 305-320.
Rostaminezhad Dolatabad, H., Shahabi, S., & Madadi, M. R. (2023). Evaluating Decision Tree Efficiency in Combination with Wavelet Transform to Predict Groundwater Level Fluctuation. Iranian Journal of Irrigation & Drainage, 17(3), 413-427.
Sadat‑Noori, M., Glamore1, W., Khojasteh, D. (2020). Groundwater level prediction using genetic programming: the importance of precipitation data and weather station location on model accuracy. Environmental Earth Sciences, 79:37.
Saeedi Razavi, B., Arab, A. (2019). Groundwater Level Prediction of Ajabshir Plain using Fuzzy Logic, Neural Network Models and Time Series. Journal of Hydrogeology, 3(2), 69-81. (In Persian)
Shrivastava, N. A., & Panigrahi, B. K. (2014). A hybrid wavelet-ELM based short term price forecasting for electricity markets. International Journal of Electrical Power & Energy Systems, 55, 41-50.
Suryanarayana, C., Sudheer, C., Mahammood, V., & Panigrahi, B. K. (2014). An integrated wavelet-support vector machine for groundwater level prediction in Visakhapatnam, India. Neurocomputing, 145, 324-335.
Tapak, L., Rahmani, A. R., & Moghimbeigi, A. (2013). Prediction the groundwater level of Hamadan-Bahar plain, west of Iran using support vector machines. Journal of research in health sciences, 14(1), 82-87.
Vapnik, V. (1998). Statistical learning theory. (No Title).
Yoon, H., Jun, S. C., Hyun, Y., Bae, G. O., & Lee, K. K. (2011). A comparative study of artificial neural networks and support vector machines for predicting groundwater levels in a coastal aquifer. Journal of hydrology, 396(1-2), 128-138.
Yu, H., Wen, X., Feng, Q., Deo, R. C., Si, J., & Wu, M. (2018). Comparative study of hybrid-wavelet artificial intelligence models for monthly groundwater depth forecasting in extreme arid regions, Northwest China. Water resources management, 32, 301-323.
Zhou, T., Wang, F., & Yang, Z. (2017). Comparative analysis of ANN and SVM models combined with wavelet preprocess for groundwater depth prediction. Water, 9(10), 781.
Iranian Journal of Soil and Water Research
Volume 54, Issue 9
December 2023
Pages 1415-1429

Files

Share

How to cite

Statistics