Application of the Entropy Spectral Method for Streamflow and Flood-Affected Area Forecasting in the Brahmaputra River Basin

Abstract

Reliable streamflow and flood-affected area forecasting is vital for flood control and risk assessment in the Brahmaputra River basin. Based on the satellite remote sensing from four observation sites and ground observation at the Bahadurabad station, the Burg entropy spectral analysis (BESA), the configurational entropy spectral analysis (CESA), maximum likelihood (MLE), ordinary least squares (OLS), and the Yule–Walker (YW) method were developed for the spectral analysis and flood-season streamflow forecasting in the basin. The results indicated that the BESA model had a great advantage in the streamflow forecasting compared with the CESA and other traditional methods. Taking 20% as the allowable error, the forecast passing rate of the BESA model trained by the remote sensing data can reach 93% in flood seasons during 2003–2017, which was significantly higher than that trained by observed streamflow series at the Bahadurabad station. Furthermore, the segmented flood-affected area function with the input of the streamflow forecasted by the BESA model was able to forecast the annual trend of the flood-affected area of rice and tea but needed further improvement in extreme rainfall years. This paper provides a better flood-season streamflow forecasting method for the Brahmaputra River basin, which has the potential to be coupled with hydrological process models to enhance the forecasting accuracy.