A Comprehensive Survey on IoT and Machine Learning for Flood Early Warning Systems

Flood Early Warning Systems (FEWSs) increasingly combine heterogeneous observation streams, computational models, and interoperable information infrastructures. This survey reviews how Internet of Things (IoT)-based sensing, communication protocols, data pre-processing, machine learning, data assimilation, and operational platforms contribute to flood monitoring and early warning. Rather than presenting a new forecasting model or claiming end-to-end experimental validation, the paper synthesizes cross-layer integration patterns between observation networks, computational methods, uncertainty-aware decision-support, and standards-based middleware. Particular attention is given to the interface between in situ IoT observations and data-driven or physics-informed modelling approaches, including data assimilation and hyper-local flood simulation. A FIWARE/NGSI-LD hydrometeorological monitoring deployment is included as an infrastructure and interoperability case study, demonstrating the continuous collection and ingestion of sensor observations into the platform, semantic normalization, temporal persistence, and standards-based access to monitoring data.