Policy learning pathways in disaster management: evolution of flood prevention systems in China’s megacity

How do urban emergency management systems evolve through policy learning, and what institutional mechanisms drive this evolution? This study addresses this question by examining three versions of megacity Changsha’s flood prevention emergency plans spanning nearly a decade (2012, 2017, 2020), supplemented by semi-structured interviews with 12 government officials and four official post-disaster assessment reports. We construct an integrated analytical framework combining large language model-based task segmentation (DeepSeek-V3), Chinese semantic embedding (bge-large-zh-v1.5) with BERTopic topic modeling, and bipartite social network analysis to systematically trace the structural evolution of flood emergency management. Qualitative evidence from thematic coding and documentary analysis is triangulated with computational findings to establish causal mechanisms underlying policy changes. Results reveal an evolutionary trajectory from single-department centralized management to multi-department division of labor and finally to cross-departmental coordinated response. Information Planning, Monitoring and Early Warning, and Material Support emerged as focal points by 2020, driven by institutional incentives favoring anticipatory governance, while Engineering Emergency Response and Logistics Support maintained core positions throughout. Publicity and Reporting, Social Security, and Material Support developed as critical bridging nodes for information flow and resource coordination. We identify three interacting policy learning pathways—crisis-driven learning mediated through hierarchical accountability, top-down diffusion of national directives with local adaptive implementation, and horizontal cross-jurisdictional policy borrowing—that jointly produce bounded experimentation under hierarchical constraints. These findings extend crisis-driven policy learning theory to authoritarian governance contexts and provide new methodological tools for understanding adaptive changes in urban emergency management systems.