Advanced Flood Prediction Using Federated Learning with 2D Convolutional Networks

Advanced Flood Prediction Using Federated Learning with 2D Convolutional Networks P.Arun Reddy1, Pemmanaboina Rohith Karthikeya2, Ettaboina Shiva Krishna3, Dadige Shiva Kumar4, Navapet Sharan Kumar Yadav5 03 06 Floods are one of the most destructive natural disasters causing serious damage to lives agriculture and infrastructure Predicting floods in advance is a difficult task due to changing environmental conditions and complex water systems This study presents a flood forecasting model based on federated learning which allows multiple locations to train models without sharing raw data This approach helps reduce network delays while maintaining data privacy and security Instead of sending large datasets to a central server local models are trained and only the learned parameters are shared The proposed system analyzes data from 18 different stations to identify areas at risk and provides flood alerts up to five days in advance A feedforward neural network FFNN is used to predict water levels using factors such as rainfall snowmelt flow routing and hydrodynamics The model was tested on data from 2010 to 2015 and achieved an accuracy of 84

2026