Prediction of monthly regional groundwater levels through hybrid soft-computing techniques

doi:10.1016/j.jhydrol.2016.08.006

機構典藏 > College of Engineering > Graduate Institute & Department of Water Resources and Environmental Engineering > Journal Article > Item 987654321/109594

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/109594

Title:	Prediction of monthly regional groundwater levels through hybrid soft-computing techniques
Authors:	Fi-John Chang;Li-Chiu Chang;Chien-Wei Huang;I-Feng Kao
Keywords:	Regional groundwater level forecast;Artificial neural networks (ANNs);Self-organizing map (SOM);Nonlinear autoregressive with exogenous inputs (NARX) network;Zhuoshui River basin
Date:	2016-10
Issue Date:	2017-02-24 02:12:12 (UTC+8)
Publisher:	Elsevier BV
Abstract:	Groundwater systems are intrinsically heterogeneous with dynamic temporal-spatial patterns, which cause great difficulty in quantifying their complex processes, while reliable predictions of regional groundwater levels are commonly needed for managing water resources to ensure proper service of water demands within a region. In this study, we proposed a novel and flexible soft-computing technique that could effectively extract the complex high-dimensional input–output patterns of basin-wide groundwater–aquifer systems in an adaptive manner. The soft-computing models combined the Self Organized Map (SOM) and the Nonlinear Autoregressive with Exogenous Inputs (NARX) network for predicting monthly regional groundwater levels based on hydrologic forcing data. The SOM could effectively classify the temporal-spatial patterns of regional groundwater levels, the NARX could accurately predict the mean of regional groundwater levels for adjusting the selected SOM, the Kriging was used to interpolate the predictions of the adjusted SOM into finer grids of locations, and consequently the prediction of a monthly regional groundwater level map could be obtained. The Zhuoshui River basin in Taiwan was the study case, and its monthly data sets collected from 203 groundwater stations, 32 rainfall stations and 6 flow stations during 2000 and 2013 were used for modelling purpose. The results demonstrated that the hybrid SOM-NARX model could reliably and suitably predict monthly basin-wide groundwater levels with high correlations (R2 > 0.9 in both training and testing cases). The proposed methodology presents a milestone in modelling regional environmental issues and offers an insightful and promising way to predict monthly basin-wide groundwater levels, which is beneficial to authorities for sustainable water resources management.
Relation:	Journal of Hydrology 541(B), p.965-976
DOI:	10.1016/j.jhydrol.2016.08.006
Appears in Collections:	[Graduate Institute & Department of Water Resources and Environmental Engineering] Journal Article

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	153	View/Open
Prediction of monthly regional groundwater levels through hybrid soft-computing techniques.pdf		3273Kb	Adobe PDF	1	View/Open

Loading...