Conventional models of emergency science popularization, characterized by linear knowledge production, homogenized information dissemination, and reliance on single physical settings, are increasingly inadequate for addressing the evolving nature of disaster risks, heterogeneous public perception, and limited scenario coverage. Systematic optimization of emergency science popularization is therefore urgently required. Drawing on socio-technical systems theory, this study employs Latent Dirichlet Allocation (LDA) topic modeling to analyze emergency science popularization texts, extracting a novel development paradigm driven by “technology-institution-space”. Research indicates that the technology-enabled dimension achieves real-time adaptation to evolving disaster risks through dynamic knowledge production; the institutional innovation dimension resolves the challenges of fragmented resource allocation and homogenized information dissemination via a standardised precision communication system; while the spatial integration dimension establishes a comprehensive coverage network by leveraging the convergence of‘physical-social-informational’ triadic spaces. These three dimensions collectively form a multidimensional, comprehensive, and efficient emergency science popularization paradigm. Based on these findings, this study proposes several optimization pathways, including strengthening the application of digital and intelligent technologies, improving the resource support system for emergency science popularization, and promoting comprehensive, cross-regional, and multi-scale dissemination. The study provides a theoretical framework and practical insights for addressing the imbalance between the supply and demand of emergency science popularization and for enhancing public disaster risk awareness and preparedness, thereby supporting the modernization of China’s emergency management system and governance capacity.