Application of artificial neural networks in flow discharge prediction for the Fitzroy River, Australia

JOORABCHI, A. ZHANG, H. and BLUMENSTEIN, M, 2007. Application of artificial neural networks in flow discharge prediction for the Fitzroy River, Australia. Journal of Coastal Research, SI 50 (Proceedings of the 9th International Coastal Symposium), 287 – 291. Gold Coast, Australia, ISSN 0749.0208 Prediction of flow discharge, and in particular floods, in rivers is one of the basic and key information in regards to operation and management of the river systems. The Fitzroy River, one of the largest Australian river systems, has a historical recording of heavy floods and there is a concern for the people of that area to have a clear prediction of the stream discharge to avoid damages. In this paper a feed-forward artificial neural network (ANN) model has been used to forecast the daily flow discharge of the Fitzroy River up to four days ahead. The feed-forward neural network uses error Back-propagation learning algorithm. A cross validation method is applied to prevent the over-fitting problem. The network uses multiple inputs including the daily values of discharge. The network output consists of four neurons in respect to the number of forecasted days. A suitable number of inputs for time-series data were selected by trial and error. Two different multi-layer networks were compared to find the optimised network. The results show an accurate forecasting of flow discharge during flood events. However, the neural network overestimates during low discharge with a mean value of 80 (m/s). ADDITIONAL INDEX WORDS: Flood, Back-propagation, Time-series prediction INTRODUCTION The application of artificial neural networks (ANNs) has been widely applied to the various areas. Numerous ANN models have been used as alternatives to the traditional numerical models. For example, in water resource engineering a neural network model was developed for river flood forecasting by CAMPOLO, et al (1997). FRENCH et al. (1992) applied ANNs to forecast the rainfall intensity. Most recently, SAHOO et al. (2006) established a neural network to predict the flash flood and attendant water qualities of a mountainous stream. Fitzroy catchment is the second largest catchment in Australia after the Murray-Darling Basin. It covers twice the size of Tasmania. The Fitzroy River (Figure 1) is one of the main rivers in this catchment that pass the city of Rockhampton. It has a number of dams and weirs to provide the fresh water to the city and its surrounding area. The importance of this area for scientific study is due to its significant loads of sediment and nutrients that are transported through this river and frequent flood events. In the last few years Coastal Cooperative Research Centres (Coastal CRC) studied this area on a number of projects. A one-dimensional hydrodynamic, sediment transport and biochemistry model is available for the Fitzroy estuary (MARGVELASHVILI et al., 2003 and 2005) based on conceptual model (WEBSTER et al., 2003). KELLY and WANG (1996) study the sediment transport in the Fitzroy River during flood events. Nutrient dynamics and sediment budgets for the estuary during a flood event are examined by FORD et al., (2006). Prediction of river flow and in particular flood forecasting is an important element of flood control systems. The early prediction helps us to minimise the flood damage. The results of floods in low lying areas are loss of communication and transport system problems. Due to low annual precipitation, runoff and elevated evaporation rates, Australia has a highly variable flow with large peaks and annual floods (WARNER, 1986). The Fitzroy River has had a number of historical floods caused by heavy rains with extended periods of low flow. It usually happens between January and April (Table 1). A flood warning system has been installed by the Bureau of Meteorology (BOM, 2005) which measures the water elevation and has a warning time of up to 60 hours for floods coming from the hinterland to the Rockhampton city. The Department of Natural Resources and Water operates a number of stations along the Fitzroy River to control a few parameters including stream water level and discharge (NRW, 2006). In this study a neural network model was developed and trained based on the 64 years of daily discharge measured since 1964 up to the end of 2005 in The Gap station (23° 5' 18" S and 150° 6' 28" E). Figure 2 shows the maximum measured value of discharge happened each year (from 1964 to 2006) at The Gap station. The presented neural network model, due to independency of the physical parameters such as boundary conditions, initial condition and bathymetry as well as reliable results and real-time prediction of flow discharge, could outper-