An Edge-to-Cloud Data Processing Framework for Real-Time Emergency Situational Awareness in Multi-Cloud Environments

Abstract

Effective emergency management demands realtime processing and fusion of massive, heterogeneous data streams from IoT sensors, video surveillance, social media, and wireless platforms distributed across disaster-affected regions. Existing cloud-centric architectures suffer from prohibitive latency, while static edge-cloud deployments fail to exploit the complementary nature of multi-source emergency data for coherent situational awareness. This paper presents Emerald, an edgeto-cloud data pipeline with integrated multi-source fusion, architected over JointCloud infrastructure for real-time emergency situational awareness. Emerald introduces three key components: (1) an urgency-aware adaptive computation offloading strategy that dynamically redistributes workloads between edge, fog, and cloud layers according to a multi-dimensional disaster urgency model; (2) a fault-tolerant data transmission protocol with breakpoint resume that preserves critical data integrity under the evaluated degraded and intermittent network conditions common in disaster zones; and (3) a conflict-penalized quality-aware data fusion method at the fog layer featuring context-adaptive dynamic credibility assessment and adaptive conflict-penalized belief aggregation based on modified Dempster-Shafer evidence theory. Experimental evaluation using the iFogSim2 simulator with parameters drawn from the 2021 Henan floods and the 2023 Turkey-Syria earthquake demonstrates that Emerald achieves 11.1% lower end-to-end latency, 99.2% system availability, 7.7% higher fusion F1-score, and preserves all critical data items under the evaluated failure settings. Scalability experiments further show that Emerald's fusion F1-score improves from 68.3% to 87.8% as infrastructure scales from 50 to 200 edge nodes.