Enhancing Flood Management and Early Warning Systems in Mountain Rivers: An ANN-Based Approach

Abstract

The study focuses on developing an HEC-RAS 1D model to simulate flood hydrographs at desired locations, considering the inflows from two different rivers Rishiganga and Dhauliganga. The objective is to analyze the impact of inflow hydrographs from the aforementioned rivers, utilizing three sets of flood data: high flow (2400 m3/sec), low flow (15 m3/sec), and 1.5 times the high flow (3600 m3/sec) peaks. An Artificial Neural Network (ANN) model, employing the Feed-forward Backpropagation method, is utilized to train the neurons using the maximum flow hydrograph. The trained ANN model is subsequently tested with two unseen inflow hydrographs of low and 1.5 x high flow to evaluate its robustness. The findings reveal that the ANN model performs very well in flood estimation, providing efficient predictions based on the testing with the maximum flow and validating with the minimum and 1.5 times higher inflow hydrographs. Its primary advantage lies in saving time, enabling timely actions when flood warnings are issued. This research significantly contributes to flood management and shall enhance the lead time in disseminating early warning to local authorities and communities. Work shall help in providing timely and accurate flood predictions for proactive measures during any flood event. The study’s significance is rooted in its potential to enhance flood preparedness and response in areas affected by the Rishiganga and Dhauliganga rivers, ultimately ensuring the safety and well-being of local communities and infrastructures.