Using Hybrid ARIMAX-ANN Model for Simulating Rainfall - Runoff - Sediment Process Case Study: Aharchai Basin, Iran

Using Hybrid ARIMAX-ANN Model for Simulating Rainfall - Runoff - Sediment Process Case Study: Aharchai Basin, Iran

Vahid Nourani (Department of Water Resources Engineering, University of Tabriz, Tabriz, Iran), Samira Roumianfar (Department of Water Resources Engineering, University of Tabriz, Tabriz, Iran), and Elnaz Sharghi (Department of Water Resources Engineering, University of Tabriz, Tabriz, Iran)
Copyright: © 2013 |Pages: 17
DOI: 10.4018/jamc.2013040104
OnDemand PDF Download:
No Current Special Offers


The need for accurate modeling of rainfall-runoff-sediment processes has grown rapidly in the past decades. This study investigates the efficiency of black-box models including Artificial Neural Network (ANN) and Autoregressive Integrated Moving Average with eXogenous input (ARIMAX) models for forecasting the rainfall-runoff-sediment process. According to the complex behavior of the rainfall-runoff-sediment time series, they include both linear and nonlinear components; therefore, employing a hybrid model which combines the advantages of both linear and non-linear models improves the accuracy of prediction. In this paper, a hybrid of ARIMAX-ANN model is applied to rainfall-runoff-sediment modeling of a watershed. At the first step of the hybrid modeling, the ARIMAX method is applied to forecast the linear component of the rainfall-runoff process and then in the second step, an ANN model is used to find the non-linear relationship among the residuals of the fitted linear ARIMAX model. Finally, total effective time series of runoff, obtained by the hybrid ARIMAX-ANN model are imposed as input to the proposed ANN model for prediction daily suspended sediment load of the watershed. The proposed model is more appropriate, as it uses the semi-linear relation for prediction of sediment load.
Article Preview

1. Introduction

Accurate modeling of hydrological processes such as rainfall-runoff or prediction of sediment load can provide the basic information for the urban and environmental planning, land use, flood and water resources management of a watershed. Prediction of sediment load is required in a wide spectrum of problem because of its impact on design and management of water construction such as a dam and stable of channels. Also sediment is a major pollutant and a carrier of nutrients, pesticides and other chemicals. The modeling of runoff-sediment yield from watershed is one of the most complex hydrological phenomena to comprehend, as it has tremendous temporal and spatial variability. Hence, currently, researchers pay more attention to apply the simpler, cheaper and easier methods to obtain a relationship between sediment load and water discharge. Therefore, studies have been conducted to reduce the complexities of the problem in terms of developing practical techniques. Models based on their involvement of physical characteristics generally fall into three broad categories: black box or system theoretical models, conceptual models, and physical based models. Due to the large number of obscure parameters involved in the hydrological variable in a watershed, black box lumped modeling may have some advantages over fully distributed modeling (Nourani & Mano, 2007). In the cases with high rate of complexity for which we cannot consider every effective physical parameter, it seems necessary to use black box models which may produce accurate results than physical based models.

In this way, one of important area of black box modeling is time series forecasting technique, in which past observations of the same variable are collected and analyzed to develop a model which is qualified to capture the underlying relationship. Nowadays, conventional black box time series models such as Auto Regressive Integrated Moving Average (ARIMA) are widely used time series models. The popularity of ARIMA model is due to its statistical as well as well-known Box-Jenkins methodology in the model building process (Box and Jenkins, 1970). Although ARIMA models are quite flexible in that they can represent several different types of time series, i.e., pure autoregressive (AR), pure moving average (MA), combined AR and MA (ARMA) and Auto Regressive Integrated Moving Average with eXogenous input (ARIMAX) series, their major limitation is the pre-assumed linear form of the model.

Recently, Artificial Intelligence (AI) techniques have shown great ability in modeling and forecasting non-linear hydrological time series. AI techniques offer an effective approach for handling large amounts of dynamic, non-linear and noisy data, especially when the underlying physical relationships are not fully understood. In general, the application of AI technique does not require a prior knowledge of the process. One of black box modeling tools that has lately found application in a variety of areas, including hydrology and environmental engineering, is Artificial Neural Network (ANN). Furthermore, capability of ANN to establish nonlinear links between inputs and outputs makes it a useful tool for modeling hydraulic and hydrological phenomena. It usually produces results faster than its physical counterparts and as accurately or more so, but only within a range of observed values in the data used to build the model. With ANN, the model is adaptively formed based on the features presented from the data. Numerous papers have already been presented on the successful application of ANN for modeling the hydrological process as a non–linear complex phenomenon (e.g., Hsu et al., 1995; Dawson & Wilby, 1998; Tokar & Johson, 1999; Sajikumara & Thandaveswara, 1999; Sudheer et al., 2000; Lallahem & Maina, 2003; Senthil Kumar et al., 2004; Jain et al., 2004; Antar et al., 2006).

Complete Article List

Search this Journal:
Volume 14: 1 Issue (2023)
Volume 13: 4 Issues (2022): 2 Released, 2 Forthcoming
Volume 12: 4 Issues (2021)
Volume 11: 4 Issues (2020)
Volume 10: 4 Issues (2019)
Volume 9: 4 Issues (2018)
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing