Flood forecasting in similar catchments using neural networks

Authors: GHOLAM REZA RAKHSHANDEHROO, MOHAMMAD VAGHEFI, MOHAMMAD MEHDI SHAFEEI

Abstract: Flood forecasting is an essential requirement for solving a wide range of scientific and/or management problems. Neural networks have become attractive models in many practical applications, including flood forecasting. In this paper, 4 similar catchments in Iran, with high quality rainfall-runoff records, were studied. An artificial neural network (ANN) was built and trained as an event-based modeling tool utilizing data from only 2 of the catchments. The flood forecasting ability of the model was then evaluated for all catchments. Results showed that during the training, the model simulated observed peak flows very closely. Based on a good simulation of the response from unseen but similar catchments, it was concluded that the model may be utilized for flood forecasting in catchments that lie on the same cluster. However, as such catchments become less similar to the cluster; the simulation error would increase accordingly.

Keywords: Neural network, flood forecasting, similar catchments, ANN

Full Text: PDF