Title: A comparative assessment of adaptive neuro-fuzzy inference system, artificial neural network and regression for modelling stage-discharge relationship

Authors: Mehdi Vafakhah; Ebrahim Kahneh

Addresses: Department of Watershed Management Engineering, Faculty of Natural Resources, Tarbiat Modares University, Iran ' Department of Watershed Management Engineering, Faculty of Natural Resources, Tarbiat Modares University, Iran

Abstract: This study aims to evaluate the adaptive neuro-fuzzy inference system (ANFIS), artificial neural network (ANN), two and multi-linear regression (MLR) for modelling stage-discharge (S-D) relationship and the results are compared with those of the S-D rating curves. S-D relationship was carried out using discharge and water level (stage) data during 2011-2012 water year in Tangeh Lavij hydrometry station with area of 104 km² located in Mazandaran Province. The partial autocorrelation coefficients were estimated for the daily stage and three input combinations based on current and previous daily stage were selected as inputs to the ANFIS, ANN and MLR models. ANN, ANFIS and MLR methods with three input combinations of stage and conventional S-D relationship were used for modelling S-D relationship. The results showed that artificial intelligence methods were found to be superior to regression and conventional S-D relationship and ANFIS are the most suitable method for modelling S-D relationship.

Keywords: artificial neural networks; ANNs: adaptive neuro-fuzzy inference system; ANFIS; multi-linear regression; stage-discharge relationship; flow discharge; rating curve; fuzzy logic; Iran; hydrology; water discharge; river discharge; modelling.

DOI: 10.1504/IJHST.2016.075581

International Journal of Hydrology Science and Technology, 2016 Vol.6 No.2, pp.143 - 159

Received: 24 Nov 2014
Accepted: 02 Sep 2015

Published online: 28 Mar 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article