Successive-station monthly streamflow prediction using different artificial neural network algorithms

Abstract

In this study, applicability of successive-station prediction models, as a practical alternative to streamflow prediction in poor rain gauge catchments, has been investigated using monthly streamflow records of two successive stations on Ç oruh River, Turkey. For this goal, at the first stage, based on eight different successive-station prediction scenarios, feed-forward back-propagation (FFBP) neural network algorithm has been applied as a brute search tool to find out the best scenario for the river. Then, two other artificial neural network (ANN) techniques, namely generalized regression neural network (GRNN) and radial basis function (RBF) algorithms, were used to generate two new ANN models for the selected scenario. Ultimately, a comparative performance study between the different algorithms has been performed using Nash-Sutcliffe efficiency, squared correlation coefficient, and root-meansquare error measures. The results indicated a promising role of successive-station methodology in monthly streamflow prediction. Performance analysis showed that only 1-month-lagged record of both stations was satisfactory to achieve accurate models with high-efficiency value. It is also found that the RBF network resulted in higher performance than FFBP and GRNN in our study domain.

Univariate streamflow forecasting using commonly used data-driven models: literature review and case study

“…His results were in agreement with those by Kisi (2007). Mehr et al (2015) utilized ANNs to forecast monthly streamflow in poor raingauge catchments, indicating that only 1-month-ahead prediction performed well.…”

Research and application of an intelligent networking model for flood forecasting in the arid mountainous basins

“…At the same time, flash floods in these areas are likely to be accompanied by a series of geological disasters such as landslides and debris flows (Van Tu, Duc, Tung, & Cong, 2016). Accurate flood forecasting can effectively support the early adoption of flood warning, flood control and other flood mitigation measures so as to reduce flood losses (Guven, 2009; Yaseen, El‐shafie, Jaafar, Afan, & Sayl, 2015), and is also of great significance for regional planning, irrigation water intake, sediment transport and other hydrological applications (Araghinejad, Burn, & Karamouz, 2006; Danandeh Mehr, Kahya, Şahin, & Nazemosadat, 2015; Solomatine & Shrestha, 2009). However, floods in arid mountainous areas remains a weak link in disaster forecasting and prevention throughout the world (Zhang, Lin, et al, 2018; Zhang, Yu, et al, 2018).…”

“…The General Regression Neural Network (GRNN) is also a widely used model in ANN models. It does not need to go through the training iteration process of backpropagation ANN, which extracts function values directly from the input dataset, and has a lower requirement on the number of input learning samples (Danandeh Mehr et al, 2015; Specht, 1991). Many scholars use the GRNN to construct the rainfall–runoff forecasting model, and to forecast the river runoff (Awchi, 2014; Hu, Lam, & Ng, 2005; Kisi, 2008; Singh & Deo, 2007; Turan & Yurdusev, 2009).…”

RBFNN versus FFNN for daily river flow forecasting at Johor River, Malaysia

“…However, the authors proposed on the basis of obtaining results that artificial neural networks are promising for forecasting stream flow. Mehr [21] conducted successive-station monthly streamflow prediction using different artificial neural network algorithms, and the results exhibit that RBFNN model is superior to the other forecasting models. Hosseinzadeh Talaee [22] accomplished streamflow forecasting utilizing multilayer perceptron with different training algorithms, including resilient back propagation, variable learning rate and Levenberg-Marquardt.…”