Comparison of a black-box model to a traditional numerical model for hydraulic head prediction

Tapoglou, E.; Chatzakis, A.; Karatzas, G. P.

Abstract

Two different methodologies for hydraulic head simulation were compared in this study. The first methodology is a classic numerical groundwater flow simulation model, Princeton Transport Code (PTC), while the second one is a black-box approach that uses Artificial Neural Networks (ANNs). Both methodologies were implemented in the Bavaria region in Germany at thirty observation wells. When using PTC, meteorological and geological data are used in order to compute the simulated hydraulic head following the calibration of the appropriate model parameters. The ANNs use meteorological and hydrological data as input parameters. Different input parameters and ANN architectures were tested and the ANN with the best performance was compared with the PTC model simulation results. One ANN was trained for every observation well and the hydraulic head change was simulated on a daily time step. The performance of the two models was then compared based on the real field data from the study area. The cases in which one model outperforms the other were summarized, while the use of one instead of the other depends on the application and further use of the model.

Citation

Tapoglou, E., Chatzakis, A., & Karatzas, G. P. (2016). Comparison of a black-box model to a traditional numerical model for hydraulic head prediction. Global NEST journal, 18(4), 761-770. https://doi.org/10.30955/gnj.002002

• Journal Article Type: Article
• Acceptance Date: Jun 28, 2016
• Online Publication Date: Oct 19, 2016
• Publication Date: Dec 1, 2016
• Deposit Date: Oct 4, 2019
• Publicly Available Date: Nov 4, 2019
• Journal: Global Nest Journal
• Print ISSN: 1790-7632
• Electronic ISSN: 2241-777X
• Peer Reviewed: Peer Reviewed
• Volume: 18
• Issue: 4
• Pages: 761-770
• DOI: https://doi.org/10.30955/gnj.002002
• Keywords: Artificial neural network; Groundwater modeling; Hydraulic head change simulation; Princeton transport code
• Public URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85006437597&partnerID=40&md5=272a5c233b5101cf84f444555de865a8
• Publisher URL: https://journal.gnest.org/publication/gnest_02002
• Contract Date: Nov 1, 2019