Multi-Source Disaster Monitoring and Early Warning System Using Risk Scoring

Disaster information in Maharashtra is fragmented across government alerts, news reports and social media, which differ widely in speed and reliability, so District Disaster Management Authorities lack a single real-time view of an evolving event. This paper presents a multi-source disaster monitoring and early-warning system that ingests News APIs, social media and government alert feeds through an Apache Kafka producer–consumer pipeline, removes duplicate, spam and irrelevant content, and extracts the disaster type and affected city or village using keyword- and dictionary-based entity recognition. A risk-scoring engine weights each event by source reliability and recency and classifies every district as Low, Medium or High risk. A Safe Zone Identification module then applies disaster-specific safety criteria and Haversine distance to recommend the nearest suitable safe location, shown to the user as a photograph rather than text directions alone. Results are stored for retrieval and presented on a color-coded district map with a live news feed and threshold-based web and SMS alerts. The system is implemented as a working prototype; this paper describes its design and specifies the protocol for its quantitative evaluation, which is ongoing.