DEVELOPMENT OF A FOG COMPUTING-BASED REAL-TIME FLOOD PREDICTION AND EARLY WARNING SYSTEM USING MACHINE LEARNING AND REMOTE SENSING DATA

Abstract

This study introduces a novel, fog computing-based real-time flood prediction and early warning system that integrates advanced machine learning algorithms with remote sensing technologies to address the limitations of traditional flood monitoring infrastructures. Existing systems often struggle with latency, over-reliance on centralized cloud computing, and fragmented data integration, which collectively hinder their effectiveness during critical flood events. To overcome these challenges, this research employed a mixed methods research design, combining quantitative experimentation with qualitative inquiry to comprehensively evaluate the system’s predictive performance, operational resilience, and stakeholder usability. The quantitative component involved the deployment of an optimized wireless sensor network, designed using Low Energy Adaptive Clustering Hierarchy (LEACH) and Particle Swarm Optimization (PSO), to collect real-time hydrological data including rainfall, river levels, and soil moisture. These data streams were fused with high-resolution remote sensing imagery from Sentinel-1 SAR and Sentinel-2 MSI, and processed using advanced machine learning models such as Long Short-Term Memory (LSTM) networks and Random Forest classifiers. The LSTM model achieved high predictive accuracy with an RMSE of 0.38 and NSE of 0.84, while fog computing nodes reduced latency by over 60%, enabling localized, real-time alert dissemination even during internet outages. The qualitative component involved 23 semi-structured interviews with disaster management authorities, field technicians, and community representatives to assess the system’s usability, trustworthiness, and practical challenges. The triangulation of findings confirmed the technical validity, operational reliability, and end-user relevance of the system. By integrating decentralized processing, intelligent forecasting, and human-centered design, this research contributes a scalable, adaptive, and field-tested framework for intelligent flood early warning systems. It holds significant potential for deployment in climate-vulnerable, infrastructure-constrained regions globally, advancing both technological innovation and disaster resilience in the face of increasing hydrometeorological extremes.