Daily streamflow forecasting by Artificial Neural Network in a large-scale basin

Abstract

Artificial Neural Network (ANN) models have been successfully applied to daily stream flow forecasting in many basins. However, most of them are designed for small or meso-scale basins rather than large-scale basins. One of aims in this work is to develop an ANN model with an optimized combination of input variables and a more accurate architecture for daily stream flow forecasting. Another aim is to compare the performance of ANN models and a rainfall-runoff model—XXT, which is a new efficient hybrid model of Xinanjiang model and TOPMODEL, in one day in advance stream flow forecasting. Yingluoxia basin, with a drainage area of 10009 km2, is chosen as a large-scale basin. The results show that the stream flow, precipitation and evaporation are all necessary to ANN modeling for this basin. The ANN model with an appropriate combination of stream flow, precipitation and evaporation as input vector performs much better than XXT in terms of Nash-Sutcliffe efficiency. Even if only using antecedent stream flow data as inputs ANN models are still better than XXT model for one day in advance flow forecasting.