RESEARCH ARTICLE Toward multi-day-ahead forecasting of suspended sediment concentration using ensemble models Mohamad Javad Alizadeh 1 & Ehsan Jafari Nodoushan 2 & Naghi Kalarestaghi 3 & Kwok Wing Chau 4 Received: 10 July 2017 /Accepted: 2 October 2017 /Published online: 9 October 2017 # Springer-Verlag GmbH Germany 2017 Abstract This study explores two ideas to made an improve- ment on the artificial neural network (ANN)-based models for suspended sediment forecasting in several time steps ahead. In this regard, both observed and forecasted time series are in- corporated as input variables of the models when applied for more than one lead time. Secondly, least-square ensemble models employing multiple wavelet-ANN models are devel- oped to increase the performance of the single model. For this purpose, different wavelet families are linked with the ANN model and performance of each model is evaluated using error measures. The Skagit River near Mount Vernon in Washington county is selected as the case study. The daily flow discharge and suspended sediment concentration (SSC) in the current day are considered as input variables to predict suspended sediment concentration in the next day. For more lead times, the input structure is updated by adding the fore- cast of SSC in the previous time step. Results of this study demonstrate that incorporating both observed and predicted variables in the input structure improves performance of conventional models in which those only employ observed time series as input variables. Moreover, ensemble model de- veloped for each lead time outperforms the best single wavelet-ANN model which indicates superiority of the en- semble model over the other one. Findings of this study reveal that acceptable forecasts of daily suspended sediment concen- tration up to 3 days in advance can be achieved using the proposed methodology. Keywords Ensemble forecasting . Wavelet-ANN . Suspended sediment concentration . Updating input structure . Several lead times Introduction The suspended sediment load forecasting in rivers is of great importance for environment and water resources engineering. However, it is a complex and nonlinear hydrological phenom- enon in which it depends on a large number of geomorpholo- gy characteristics of the basin and flow conditions among others. Spatial variability of the flow and basin characteristics adds difficulty of the problem. As in most rivers, a major part of the transported load of sediment is in suspension form, many different types of models gaining different methodolo- gies have been developed (Asselman 1999; Van Rijn 1993; Verstraeten and Poesen 2001; Yang 1996). Earlier, mathemat- ical modeling was a common approach for simulation of suspended sediment load. The theoretical equations may not be of much interest due to consisting of a large number of influencing parameters. Recently, artificial intelligence-based models showed great ability for simulating and forecasting of different hydrological phenomena. The main advantage of such models over the mathematical models is they do not need to establish exact relationship between input and target Responsible editor: Philippe Garrigues * Mohamad Javad Alizadeh mjalizadeh@mail.kntu.ac.ir; mohamadjavad@aut.ac.ir 1 Faculty of Civil Engineering, K. N. Toosi University of Technology, Tehran, Iran 2 Department of Civil Engineering, Bijar Branch, Islamic Azad University, Bijar, Iran 3 School of Mathematics, Iran University of Science and Technology, Tehran, Iran 4 Department of Civil and Environmental Engineering, Hong Kong Polytechnic University, Hung Hom, Hong Kong Environ Sci Pollut Res (2017) 24:28017–28025 https://doi.org/10.1007/s11356-017-0405-4