An Effective Framework for Improving Performance of Daily Streamflow Estimation Using Statistical Methods Coupled with Artificial Neural Network

Abstract

This study presents an effective framework that combines artificial neural network (ANN) and statistical methods to more efficiently, consistently, and reliably estimate the daily streamflow in ungauged basins. First, two statistical methods, including drainage area ratio (DAR) and standardization with mean (SM), are used to transfer hydrological data from gauged (donor) to ungauged (target) basins, which is known as the regionalization process. Second, to get better estimation performance, an ensemble approach is applied, which is mainly based on a weighted combination of DAR and SM. Finally, a successful strategy with an optimized ANN structure is built using daily areal precipitation for the target basin, the daily streamflow of the selected donor basin, and the estimated daily streamflow for the target basin from the best-fit method as model inputs. Its performance is tested in a case study from the Coruh River Basin, Turkey, that involved using datasets from seven streamflow gauging stations on the mainstream of Coruh River. The proposed approach has indicated the best performance on both training and testing sets. The proposed approach proves to be one of the best available practical solutions in the streamflow estimation for ungauged basins.