Vol. 11, Issue 9 (2022)

Author(s):
Archana Majhi, Sagar Chandra Senapati, Ambika Prasad Sahu and Janaki Ballav Mohapatra

Abstract:
Flood is a severe natural hazard with potentially devastating consequences leading to huge loss of human life, agricultural production and property worldwide. Therefore, the use of flood forecasting and early warning systems is of utmost importance in order to reduce the economic losses and the risk for people. This study presents the application of artificial neural network (ANN) in forecasting lead-time Streamflow discharge at Khairmal station located in the middle reach of Mahanadi River basin using previous discharge data as well as discharge data from three upstream stations (Kantamal, Kesinga and Salebhata) as inputs to the model. Moreover, the study also investigated the effect of length of training data period on network architecture and model performance for attaining the best model efficiency. The findings of the study revealed about reasonable forecast of the one- and two-days ahead Streamflow discharge without relying on other information of the region. The 2003-2008 period trained model predicted the one-day ahead discharge accurately whereas, the 2003-2010 period trained model was found to be superior in predicting the two-days ahead Streamflow discharge. Further, the study inferred that the model trained with shorter training datasets requires complex architecture as compared to the model trained with longer training datasets.