Prediction of Stage and Discharge by Hybrid Machine Learning Techniques and Flood Hazard Vulnerability and Risk Assessment of a Catchment using hec ras and gis based Multi criteria Decision Analysis

Abstract

Floods, an unpredictable natural disaster, recur with varying magnitude and frequencies which cause unexpected damage to human civilization across the globe. Teesta river basin is one such flood-prone basin in India which has experienced several historical floods. This study attempts to conduct a detailed analysis of the basin,considering all the aspects of a flood-prone basin. As discharge data is unavailable at the outlet of the basin, it is simulated by developing a hydrological model using HEC-HMS. Depending upon the existing correlation between the simulated and observed discharge at the two upstream gauging stations, the simulated discharge at the outlet is accepted. Artificial neural network (ANN) models and wavelet-based ANN (WANN) models were developed to predict stage at six gauging stations for one, three, and five-day lead time. The WANN models proficiently captured the original pattern of the stage values and precisely predicted the high water levels. Similarly, wavelet-based ANN (WANN) models, long short-term memory (LSTM) models, and wavelet-based LSTM models were developed to predict discharge at three gauging stations, including the outlet for one to five-day lead time. Both WANN and WLSTM models performed equally well. The flood hazard zones were identified based on the flood depth for different boundary conditions evaluated by the two-dimensional hydraulic model, HEC-RAS. The inundated area was also estimated for 2015, 2016, and 2017. The flood vulnerability map was prepared by combining the thematic maps of seventeen parameters with the multicriteria decision analysis technique, analytic hierarchy process (AHP). Finally, the flood risk map for 2015, 2016, and 2017 for different boundary conditions were generated by overlaying the flood hazard and vulnerability map. This intensive study will help policymakers to apprehend the incoming flood situation and protect the common people.