Dynamic Adaptive Multimodal Graph Fusion Network for Multi-Hazard Recognition Using Remote Sensing, Social Media, IoT, and Seismic Observations
Abstract
Abstract Reliable recognition of natural hazards remains challenging because the relevance of remote sensing, environmental sensors, social media and seismic monitoring is heterogeneous and the derived evidence varies as events evolve. This study presents a Dynamic Adaptive Multimodal Graph Fusion Network (DAMGF-Net) for recognising earthquakes, floods, fires, smoke, and normal conditions by jointly exploiting these complementary information sources. Unlike fixed multimodal fusion, DAMGF-Net constructs dynamic cross-modal relationships and adaptively weights each modality according to the available evidence relevant to the prediction task. The resulting graph representation captures spatial, temporal, and semantic dependencies while enabling the contribution of individual modalities to be examined. Experiments comprised 68,308 remote-sensing images, 45,384 social-media posts, 59,276 Internet-of-Things sensor records, and 1,300 seismic events. DAMGF-Net achieved 96.27% test accuracy, a weighted F1-score of 0.9618, a Matthews correlation coefficient of 0.9497, and a mean multiclass ROC-AUC of 0.9944. The model further processed 88.6 images per second with 11.3 ms inference latency. These findings show that dynamically adapting cross-modal relationships and modality contributions provides an effective and interpretable approach to multi-hazard recognition, with potential for near-real-time disaster monitoring, early warning, and resilient decision support.
Read the paper
Related studies
- Current state and future directions for deep learning based automatic seismic fault interpretation: A systematic review
- Physics of Partially Saturated Porous Media: Residual Saturation and Seismic-Wave Propagation
- Modern Imaging Using Seismic Reflection Data
- Effects of marine seismic surveys on free-ranging fauna: a systematic literature review
- A review of the recent seismic activity in the southern part of egypt (upper egypt)
- A review of research on mechanism of seismic attenuation and energy compensation
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.